The playbook

Most firms don't have an AI problem. They have an AI operating problem.

A field guide that takes your firm from scattered personal use to AI running real work reliably. You don't read it front to back: you find where you actually are and start there.

The maturity ladder

Where is your firm today?

Stage 1
Personal leverage
Win back hours, starting this week.
You're here if…
  • A few people use AI, in their own way, when they remember to
  • Nothing is written down or shared
  • The value is real but scattered and personal
Read Stage 1
Stage 2
First business win
Bring AI into one real workflow.
You're here if…
  • Individuals are fluent, but no process runs on AI yet
  • You want a shared quality bar, not personal hacks
  • You're asking "what should we automate first?"
Read Stage 2
Stage 3
Expansion
Let automation and agents carry more.
You're here if…
  • One workflow works; you want the next five
  • You're weighing your first agent, and want it done safely
  • Things run unattended and you need guardrails
Read Stage 3
Stage 4
Operating engine
Make AI a dependable part of how the firm runs.
You're here if…
  • AI is in several workflows but nobody owns the whole picture
  • Clients, insurers, or regulators are starting to ask
  • You want it to compound, not sprawl
Read Stage 4
This is not a ladder you owe anyone a climb. Many firms operate well at Stage 2 or 3 indefinitely, and each stage says so at its close.
Stage 1 · Personal leverage

Win back hours, starting this week

Which recurring tasks actually drain your week? You have the basics; the value so far is real but scattered. That is the norm, not a failing, and this altitude is where it stops being scattered.

The reading

Most people meet AI as a curiosity: a poem on demand, a recipe from leftovers, a quick answer to a trivia question. The working version is less flashy and far more valuable. This section shows the shape of the opportunity, with one rule of thumb: if the work is made of language, a model can carry part of it today.

What it can carry, and what stays yours

Work made of language

Work made of language is reading, drafting, comparing, extracting, and summarizing.

This is where a model is strong today: it does in minutes what costs a person hours, and it does not tire on page two hundred.

Example

Reading forty contracts and returning every notice period and renewal date in one table.

Work made of judgment

Work made of judgment is decisions, advice, and the final word.

This stays yours. The model prepares the decision: it gathers, drafts, and compares, so you decide with better material in less time.

Example

The model summarizes the settlement options overnight; which one to recommend is still your call.

Most professional jobs are a braid of the two, which is why the gains land everywhere once you look for them.

The same tool, across one firm

  1. The core work. A lawyer hands over a stack of agreements and gets back every deadline, obligation, and unusual clause, matched and tabled for checking.
  2. Intake and admin. Messy inquiry emails become clean, structured entries for the case or project system, drafted and ready to file.
  3. Finance. Two exports that never quite agree get reconciled line by line, with the variances explained in plain language.
  4. Communications. One announcement gets drafted in the firm's tone, then reshaped for the newsletter, the website, and a client email.
  5. Analysis. A raw spreadsheet export comes back as the finding: what moved, what did not, and what is worth a closer look.
  6. Research. A week of industry and regulatory news becomes a briefing note, with the two developments that matter on top.
These are patterns, not products. None of them needs special software: the mainstream AI tools can do a version of all six today, in a browser, from a plain-language instruction. The examples are conceptual on purpose; your own week supplies the real ones.
If the work is made of language, part of it can be handed over. The rest of this stage teaches how: talking to the model like a colleague, giving it your context, and delegating whole tasks instead of asking small questions.

Your first automation, this week

  1. Pick a small recurring task. The weekly update email is a good one: gather your notes and sent messages from the week.
  2. Hand it over with a real instruction. Summarize what was achieved, in your voice, addressed to the person who actually reads it.
  3. Review, fix, send. The first draft will be close; your corrections make next week's better.

This needs no setup, no connected systems, and no permission from anyone: the account you already have is enough, and free plans are enough to feel the gain. One honest caveat before you rely on it: the model is weakest where a single exact fact or figure must be right without checking, which is why the working habits in the next section matter.

Takeaway

If the work is made of language, a model can carry part of it today, across every desk in the firm. Start with one recurring task this week; the account you already have is enough.

The difference between a party trick and a working tool is how you talk to it. Getting access takes minutes: any of the mainstream tools, in a browser, and a free account is enough to learn on. This section teaches the working habit, with one rule of thumb: brief it like a colleague, review it like a junior's draft.

A question is not a brief

A question

A question asks the model to produce an answer from what it already knows.

You get something generic, because it knows nothing about your situation beyond the sentence you typed.

Example

Write a follow-up email to a client.

A brief

A brief gives the model the situation, the material, and the standard to hit.

You get something usable, because the model works from your facts and your constraints instead of guessing them.

Example

Here is the thread and our proposal; draft a follow-up that answers their two objections, in my tone, under 150 words.

Most disappointing first experiences with AI are questions that should have been briefs.

The working loop: instruct, review, improve

  1. Instruct. State the task, hand over the material, name the audience and the standard to hit.
  2. Review. Read the draft the way you would a junior colleague's: the structure is usually sound, and the specific facts, names, and numbers are where you check.
  3. Improve. Say what is wrong and ask for the revision; the second pass with real feedback is where the quality jumps. Keep the briefs that worked, because a later section turns them into assets.
Make "look it up" your reflex instead of "what is". When a fact matters, have the model search and cite its source rather than answer from memory, and open the source yourself when the stakes are real. You stay accountable for what you send.
AI executes, you certify. Nothing the model produces goes to a client, a colleague, or a decision without your read. That rule is what makes delegating safe, here and at every later altitude.

Two session habits

  1. Give it your context deliberately. Within a chat it knows only what you put there, so paste the relevant background instead of assuming it remembers your business. The next section makes this permanent.
  2. Start fresh when a thread drifts. A long chat full of corrections drags old misunderstandings forward; take the good output and open a clean session.

The loop gets far more useful once the model stops being a stranger to your work, which is the next section.

Takeaway

Brief it like a colleague, review it like a junior's draft, and make "look it up" your reflex. The loop is the skill; the tool is just where you run it.

A model that knows nothing about your business starts every chat as a stranger. This section shows you how to give it your world once, with one rule of thumb: set up what it should always know, connect what it may see, and keep both under your control.

What it knows, and what it can reach

Persistent context

Persistent context is what the model knows about you before the chat starts.

Standing instructions, project spaces, and memory carry your background from session to session, and you can read and edit what is stored.

Example

It knows your firm's services and your tone rules without being told again every morning.

Connections

A connection lets the model reach into a system you use, live.

Through connectors it can read or act in your calendar, files, or inbox, always inside the access you granted.

Example

It reads this week's calendar and drafts the Friday summary from what actually happened.

Context is what it knows; connections are what it can see and do. With both set up, whole categories of asking disappear.

Set it up in three layers

  1. Standing instructions. Who you are, what the firm does, how you like output. Written once, applied everywhere.
  2. Project spaces. One space per ongoing matter, loaded with its documents, so every chat inside it starts informed.
  3. Connections, narrowest first. Connect the systems the work needs, one at a time, each with the smallest access that does the job.
Grant access the way you would hand over a key card: to named rooms, not the whole building. Approve actions per use rather than choosing "always allow", and disconnect what you stop using. The model works for you; the access decisions stay yours.
You stay accountable for what the system reads and does. Narrow scopes, per-use approvals, and an occasional look at what memory has stored keep that accountability real rather than theoretical.

Three checks on your setup

  1. What does it know? Read the stored memory and standing instructions; fix what has gone stale.
  2. What can it reach? List the connections; disconnect what the work no longer needs.
  3. What can it do alone? At this altitude, nothing irreversible without your approval.

With your world attached, you can stop asking questions and start handing over whole tasks, which is the next section.

Takeaway

Give it your world once: standing instructions, project spaces, and the narrowest connections that do the job. What it knows and what it may touch stay decisions you make, not defaults you inherit.

Questions save minutes; delegated tasks save hours. This section shows you how to hand over a whole piece of work, with one rule of thumb: delegate the task, keep the judgment.

The dividing line: dabbling or delegating

Dabbling

Dabbling uses the model for fragments of work you are already doing.

You ask for a sentence here and an idea there, so the task's shape and its hours stay yours.

Example

Polishing one paragraph of the proposal you spent the afternoon writing.

Delegating

Delegating hands the model a whole task with a brief, and makes you the reviewer.

The model produces the full first version; your time moves from producing to directing and certifying.

Example

Handing over the whole proposal first draft: the brief, the client notes, the outline, then reviewing what comes back.

The real hours live on the delegating side of that line.

What to hand over first

  1. Work with a known good shape. Proposals, summaries, reports, first-pass reviews: you can describe what good looks like, so you can check what comes back.
  2. First versions, not final calls. Drafts, options, and analyses to react to; decisions stay with you.
  3. Inside its strengths. Language-shaped work first; hold back the tasks where you cannot verify the output until your verification habits are solid.
Delegation changes where your time goes, not how much care the work gets. A short, real brief plus a serious review still puts your judgment on the page; what disappears is the hours of production in between.
Brief, review, iterate: the same loop, on a bigger unit. The habits from the earlier sections carry over unchanged; what grows is the size of what you hand over.

A good first delegation has three parts

  1. The brief holds the standard. Material, audience, tone, and what good looks like, stated up front.
  2. The review is real. Facts, names, and numbers checked; structure judged; nothing forwarded unread.
  3. The second round is planned. One revision with pointed feedback is part of the job, not a failure of the tool.

Tasks you delegate well once are worth making repeatable, which is the last section of this stage.

Takeaway

Delegate whole tasks with a real brief, and review like the work still carries your name, because it does. The hours come back on the delegating side of the line.

The first delegation is an experiment; the fifth should be a routine. This section shows you how to keep what works, with one rule of thumb: when something works twice, capture it so it works every time.

A result is not an asset until it is captured

A one-off prompt

A one-off prompt gets a good result and then disappears into the chat history.

Next month someone rebuilds it from memory, a little differently, with a little less of what made it work.

Example

The brief that produced the best proposal draft so far, somewhere in March's chats.

A captured asset

A captured asset is the working brief saved as a reusable instruction.

Saved as a skill or a template, it runs the same way every time and can be handed to a colleague.

Example

A saved "proposal first draft" instruction that anyone at the firm can run on a new client brief.

Capture is the difference between being personally faster and building something the firm keeps.

Three assets worth building

  1. Templates and standing briefs. Your best instructions, kept where you can rerun them, named so you can find them.
  2. Skills. Multi-step instructions the model applies on demand: how your firm writes summaries, structures reviews, formats deliverables. Reusable, versionable, and shareable.
  3. Scheduled tasks. Recurring work set to run on a rhythm, like a Monday pipeline summary that now arrives instead of being produced. Review what arrives; improve the instruction, not each output.

This is also the point where paying for the tool earns its keep. A free plan carries you through the working loop; the machinery of capture and context, project spaces, connections, saved skills, and scheduled tasks, mostly lives on the paid plans, from about twenty dollars a month per person. You are not paying for a smarter chatbot; you are paying for the features that turn a chatbot into a delegation system.

Treat these assets the way developers treat code: kept in one agreed place, improved in the master copy rather than in scattered private variants, and reviewed now and then for what has gone stale.
Every asset you capture makes the next delegation cheaper. This library is also the seed of something larger: Stage 4 turns it into a firm-wide asset, but it starts here, personal and small.

The starter library, concretely

  1. Five briefs that work. Your most-repeated tasks, captured and named.
  2. One skill. The task you delegate most, written as a reusable instruction.
  3. One scheduled task. A piece of recurring work that now arrives on its own.

That library completes this altitude's outcome; the altitude close below says what operating well here looks like.

Takeaway

When it works twice, capture it: templates, skills, scheduled tasks. Personal speed fades with the person; a library compounds.

Before you paste client work

One professional habit as the library grows: know what your tool does with what you paste. Two minutes in the data settings tells you whether conversations are used for training and how long they are kept; check once, choose deliberately, and the question is settled. Client material deserves the same care in an AI tool as in any other system, and when colleagues start following you onto the tool, business terms are the natural next step. The later stages pick that up.

Operating well at this altitude

Most readers should run at this altitude for a while, and staying here is fine. Operating well at Stage 1 looks like: the setup is trusted, the delegation habit is real, and the starter library grows on its own. Stage 2 is there for the day a shared workflow needs more than personal leverage, not because a next stage exists.

Stage 2 · First business win

Bring AI into one real workflow

This is where AI stops being a personal trick and becomes something the firm does: one shared workflow, with a quality bar that survives the enthusiast leaving.

The reading

Stage 2 starts with a choice: which workflow gets your firm's first real AI effort. The scarce thing is not opportunity; a working firm trips over automatable work daily. The scarce thing is choosing well. This section turns the pile into one commitment, with one rule of thumb: list broadly, filter hard, win one.

Thirty minutes of inventory

  1. Where work waits. Walk the firm's queues and handoffs: quotes waiting to be written, tickets waiting to be answered, invoices waiting to be checked. Waiting work is the loudest signal.
  2. Where people retype and reconcile. Every place someone copies between systems, cross-checks two lists, or re-enters what already exists somewhere is a candidate.
  3. Where the same shape recurs. The reports, replies, summaries, and first drafts that get produced to the same pattern every week.

If Stage 1 ran its course, part of the list writes itself: the tasks people already hand to a model on their own are votes for what the firm should automate properly.

The pick that impresses, and the pick that wins

The impressive pick

The impressive pick is the one that would look best in a demo.

It is usually judgment-heavy, exception-rich, and tangled in half the firm's systems, which is exactly why it fails as a first build.

Example

An assistant that handles the whole client relationship, from first enquiry to renewal.

The winning pick

The winning pick is boring, frequent, bounded, and easy to check.

It runs often enough to judge, has a clear start and finish, and a mistake is cheap to catch, which is exactly what a first build needs.

Example

Drafting the first version of every proposal from the intake notes.

The filter below turns that instinct into a checklist.

Seven tests before you commit

  1. Frequent. The task happens daily or weekly, so there is enough volume to judge the result and the win pays back quickly.
  2. Bounded. It has a clear start and finish, with defined inputs and outputs.
  3. Painful. It costs real hours or real delay today, so winning it is worth something.
  4. Measurable. You can put a number on how it runs today; the next section captures that number as your baseline.
  5. Low error-stakes. A mistake is cheap to catch and cheap to fix, because early versions will make some.
  6. Mature. The process is stable and understood, not something the firm is still redesigning.
  7. Fed by reachable data. The inputs it needs exist in systems you can connect, current and complete, not on paper or in one person's head.

Rank what survives

  1. Impact. The hours it drains, how often it runs, what delay costs, what an error costs. The step that clogs first when volume rises and the step that quietly eats your most qualified people's week both score high here.
  2. Effort. The systems it touches, the exceptions it sprouts, the judgment it needs. More of each means a heavier build.

Score each survivor on both, and take the first workflow from the high-impact, low-effort corner. Impressive-but-heavy candidates keep their place on the list; they are what Stage 3 is for, after this win has built the muscle.

The constraint is where work waits, which is not always where people complain. A step everyone grumbles about may cost the firm little, while a quiet step nobody mentions holds up every job that passes through it. Follow the waiting work, not the noise.
The workflow to win first is high-impact, low-effort, and passes the seven tests. And it is one bounded task, not a department: if proposals are the clog, the first win is drafting the first version of one, not reinventing how the firm sells.

Typical first picks include proposal drafting at an agency, first-line support replies at a software business, the seasonal crunch at an accounting practice, and after-hours enquiries at a local service firm. Treat these as places to look, and test whatever you find against the seven tests.

Once the workflow is chosen, the next move is to map it as it really happens.

Takeaway

Candidates are everywhere; choosing is the skill. List broadly, filter with the seven tests, rank by impact against effort, and commit to one bounded task.

You have chosen the workflow; before anything gets built, it needs a map. This section shows you how to make one in a week, with one rule of thumb: capture the work as it actually happens, together with the number it actually costs.

Every workflow exists twice

The official process

The official process is how the workflow is supposed to run.

It lives in handbooks and job descriptions, written from memory, often by someone who no longer does the work.

Example

The handbook says every enquiry is logged in the system before anyone replies.

The real process

The real process is how the work actually gets done this week.

It includes the workarounds, exceptions, and shortcuts people have built over time, and those are exactly what an automation trips over.

Example

In practice, urgent enquiries go straight to a senior's inbox and get logged afterwards, when there is time.

Map the real one. The gap between the two is where the surprises live.

Record the work, don't write it from memory

  1. Record one full pass. The person who does the work records their screen and talks through what they are doing, once, the way they would show a new colleague. If several people touch the workflow, each records their part; the handoffs between them are the most important footage.
  2. Let a model draft the map. Hand the transcript to a model with your instructions and it returns a step-by-step description in minutes instead of an afternoon.
  3. The person corrects the draft. They fix what the recording missed and mark the exceptions. A person certifies the map, not the model.

Writing a process down from memory produces the official version. Recording it produces the real one, including the steps nobody thinks to mention.

Check the inputs, not only the steps

An automation is only as good as the data it reads, and people repair bad data without noticing they do it: they spot the stale price, remember that this client's address is wrong in the CRM, fill in the field the form left empty. That silent repair is part of the real process, and it disappears the moment a machine runs the step. The old rule "garbage in, garbage out" gets sharper with AI in the loop, because a model handed wrong data does not stall the way a script would; it produces something plausible from it.

  1. Where each input lives. Name the system every step reads from and writes to; this is also what the route choice in the next section is decided on.
  2. Whether it is right. Spot-check real records for the four failures that matter: wrong, out of date, incomplete, inconsistent between systems.
  3. Where people quietly fix it. Mark every point where someone corrects data on the way through. Each one either gets cleaned at the source or written into the build; what it cannot do is keep happening silently.

Attach the before-number

The baseline is one real week of the workflow, counted, not estimated. A tally in a shared sheet is enough; the point is that the week is counted while it happens, not remembered afterwards.

  1. How often it runs. The actual number of passes that week, not the average anyone remembers.
  2. How long a pass takes. From the moment work arrives to the moment it is finished, including the waiting.
  3. Where it waits. The handoffs and queues, because delay often hides in the waiting, not in the doing.
The baseline is not paperwork for its own sake. It is the number the pilot in move four will be judged against; without it, nobody can say whether the new way is better, including you. People's sense of time saved is not a reliable measure, so count the week rather than asking about it.
No baseline, no build. The map says what to build; the baseline is what proves, later, whether it worked.

Formal mapping traditions exist for all of this, from SIPOC tables to swimlane diagrams to the ISO-standardized process notation, and you do not need any of them. Borrow their three good ideas, listed below as parts of the finished map, and hold the whole thing to one page in plain language. The bar: a new colleague could run the workflow from it.

Five things the finished map contains

  1. The boundaries. What starts a pass and what ends one, what comes in and what goes out. One line each.
  2. The steps as they really happen, each with its person. Who does what, in order, with the handoffs between people marked; exceptions and workarounds included, labeled as such.
  3. The decision points. Every place the path forks and what decides the fork, because those forks are where rules end and judgment begins in the next section.
  4. The data behind each step. Which system each step reads and writes, and the input checks from above.
  5. The before-number and one named owner. Passes per week, time per pass, where work waits; and a single person who owns the map, the baseline, and the decisions the next moves ask for. Named now, not after something goes wrong.

With the map and its baseline in hand, the next move is choosing the simplest way to automate what you just mapped.

Takeaway

The map is an asset, not admin. Record the real process, check the data it runs on, count a real week, name one owner; everything you build next stands on it.

Your map from the last section lists the workflow step by step. This section shows you how to pick the right amount of automation for it, with one rule of thumb: give every step the simplest tool that can handle it.

Look at your map: two kinds of steps

Mechanical steps

The same situation always gets the same action.

These steps run identically every time, so the rule can be written down and handed over.

Example

When an invoice arrives, it goes into the client folder and the amount is entered in the sheet, the same way every time.

Judgment steps

The right action depends on what is in front of you.

Something has to be read, weighed, or worded before the step can happen.

Example

Answering a customer complaint, summarizing a call, or deciding whether a request is urgent.

Mark each step as one or the other. That split does most of the choosing for you.

Match the work to the simplest tool

  1. Fixed rules for the mechanical steps. If this happens, do that, every time. Rules are predictable and fast, and they should carry as much of the workflow as possible.
  2. An AI step for a single judgment step. You hand it one task with your instructions and guidelines: draft the reply in your tone, summarize the report, pull the key figures from the contract. It does that task and stops; what happens before and after stays with you.
  3. An agent for a stretch that varies. You give it a goal, boundaries, and sign-off points; it carries out several steps in a row, choosing the order to fit the case in front of it. The big automation platforms now offer agents as a built-in option, so getting one is not the hard part; deciding how much of the route to hand over is.
An automated workflow runs the exact route you laid, every time. An agent follows your instructions but chooses the turns as it goes. You still decide which systems it may use, which rules it follows, and where it pauses for your sign-off.

When is the agent right? If the steps in your map hold for nearly every case, automate them as written. If the rules keep sprouting exceptions until nobody can hold them all, that stretch is agent territory, as in a support inbox where the right response depends on who is writing and what has already happened.

Most well-automated workflows are a combination: fixed steps carry the process, AI handles the judgment steps inside it, an agent runs the stretches that vary. The simplest combination that covers your map is the right one.

Two ways to get it

  1. Ready-made. A vendor has already built the workflow as a finished product, the way bookkeeping tools read receipts on their own. A product has one shape, so the adapting is done by your process, not by the software; that trade is worth it when your workflow looks like everyone else's.
  2. Custom-made. Built around your workflow as you actually run it, so the map you drew stays the map instead of bending to a product's shape. This route used to mean a software project; today's AI infrastructure, standard connectors, models that can operate tools, and mature automation platforms, means a custom workflow is assembled from proven parts and hardened in weeks, not built from scratch in months.

The choice follows the map. Ready-made earns its place where your workflow looks like everyone else's; everywhere the workflow is part of how you win and keep clients, building to fit is the stronger route, and it has never been more attainable.

Where it runs is a separate choice: a vendor's cloud, or systems you run yourself. Data sensitivity, what you already run on, cost at your volume, and who can maintain it all weigh in.

Three checks before you build

  1. What it may see. Access to the data the workflow needs, and nothing beyond it.
  2. What it may do alone. Which results a person approves before they count; anything a client sees stays behind that line.
  3. Who owns it. An alert when it fails, and one named person who responds.

How you prove the new setup beats the old one is the next move: the pilot.

Takeaway

Rules for the steps that never change, AI for the judgment steps, an agent for the stretches that vary. The simplest combination that covers your map is the right one.

The new setup exists; now it has to prove itself against the baseline from move two. This section shows you how to run that proof, with one rule of thumb: run the new way beside the old way, and decide before you start what result means go.

Run it beside, not instead

A parallel run

A parallel run tries the new way while the old way keeps working as before.

Nothing breaks while you learn, because the old route still produces the real output.

Example

The model drafts the reply, and this month someone still writes theirs as usual; you compare the two.

A hard switch

A hard switch replaces the old way on day one.

Everything now depends on a setup nobody has watched under real load, and the way back is improvised.

Example

The old drafting routine is dropped the same week the automation goes live.

Pilot in parallel; switch only after the pilot has passed. The comparison is the whole point of the exercise.

Set the pilot up in three decisions

  1. The pass mark, decided in advance. Write down what result means go before the first run: the time a pass may take, the share of outputs accepted without rework, whatever your baseline measures. A pass mark chosen afterwards will bend toward whatever result you got.
  2. The time box. Give the pilot a fixed end date: long enough for the workflow to run often enough to judge, short enough that a verdict must come. For most weekly workflows that is about a month; a fixed date beats an open-ended trial that nobody ever closes.
  3. The judges. The people who do the work daily run the pilot, not the person who built the setup. A demo run by the builder always looks good; the real test is a normal Tuesday under real load.
The baseline from move two is the referee. The pilot's numbers are compared against the counted week, not against how the old way felt; the map told you what to build, and the baseline now tells you whether it worked.
Stopping is a result, not a failure. A pilot that ends in a clean no after five weeks has done its job; it cost you weeks, where a workflow kept alive on hope costs quarters.

Three questions the pilot must answer

  1. Did it beat the baseline? The counted result against move two's number, not an impression.
  2. Did the team accept it? Outputs used as they came out count; outputs quietly redone do not.
  3. What broke, and why? The pilot's exception list is the work order for the rollout. If nothing broke, the pilot was probably too gentle; feed it the awkward cases on purpose.

A pilot that passes earns the rollout: making the new way the default is the next move.

Takeaway

Run the new way beside the old, judged by the people who do the work against the counted baseline, with the pass mark set before you start. Stopping is a legitimate result.

The pilot passed; now the new way has to become the way the work is done. This section covers the rollout, with one rule of thumb: a workflow is rolled out when it is the default, not when it is available.

Rebuild the routine, don't decorate it

Bolted on

A bolted-on rollout adds the new tool on top of the old routine.

Every old step survives and new ones are added around the tool, so the workflow ends up with more steps than before.

Example

The model drafts the reply, and the draft then passes through every step the manual version had, plus a new review of the draft itself.

Rebuilt

A rebuilt rollout redesigns the steps around what now runs on its own.

Steps that existed only because a person had to do the work by hand are removed rather than preserved.

Example

The meeting where drafts used to be assigned disappears, because the drafts now exist before it would have started.

Rebuild rather than bolt on. If the old process survives untouched beside the new one, the gains stay cosmetic.

Three things that make it stick

  1. A default, not an option. From a named date, the new way is how the workflow runs; the old way remains only as the documented fallback. As long as both routes are open, the busy week picks the familiar one.
  2. A champion beside the owner. One person on the team who uses the workflow daily, answers colleagues' questions, and collects what annoys them. It is a lighter job than it sounds, and it is how new routines survive contact with a busy week.
  3. A measure that rewards the new way. If people are still judged by numbers that reward the old way of working, the old way returns quietly. Judge the workflow by its measured outcome, the number move six tracks.
The first weeks after rollout run slower than the pilot suggested, because the whole team is now learning what the pilot group already knew. Budget for that dip instead of reading it as failure; it is the cost of the last mile, not a verdict on the workflow.
Watch it, and keep the way back. Someone checks the workflow's output on a schedule, and the owner can switch to the documented old way at any time, without a meeting. Automated work drifts as inputs, tools, and models change; unwatched, it degrades quietly.

Three checks before you call it rolled out

  1. The default test. A new case entering the workflow today goes down the new route without anyone deciding that it should.
  2. The absence test. The workflow still runs when the champion is on holiday, because the written runbook, not one person's memory, carries it.
  3. The rollback test. The owner knows exactly how to switch back, and the old way is documented well enough that switching back actually works.

Whether the rolled-out workflow keeps earning its place is the last move: measuring the win.

Takeaway

Rolled out means the new way is the default, someone owns its health, and the way back exists. Rebuild the workflow around what runs on its own; don't bolt the new tool onto the old routine.

The workflow runs; the last move decides what it earned. This section shows you how to measure one workflow's return, with one rule of thumb: count verified outcomes against the baseline, and let that number make the keep, revise, or retire call.

Count outcomes, not activity

Adoption

Adoption measures how much the new way is used.

It is easy to count and comfortable to report, but a busy tool and a better workflow are not the same thing.

Example

The team runs the drafting step on almost every proposal now.

Verified outcomes

A verified outcome is a result that was accepted and used.

It counts only when the work product passed the person who signs off on it, at the quality the firm actually ships.

Example

Most drafts go to clients with light edits; the ones rewritten from scratch are counted separately.

Report outcomes. Adoption tells you the tool is alive; outcomes tell you whether the workflow won.

Put one number on it: cost per success

  1. Count the successes. Verified outcomes over a set period, taken from the workflow's own records, not from memory.
  2. Count the full cost. The tool's price for the period, plus the human minutes still inside the workflow: preparing, reviewing, correcting.
  3. Divide, then compare. Cost divided by successes, set against the same number computed from the baseline week. The workflow wins if a good result now costs less, in money or in scarce hours.
Ask the team how it feels and you will get an answer; it just will not be a measurement. People's sense of saved time is unreliable in both directions, so the verdict comes from the counted numbers, with the team's experience kept as context for the revise decision.
Keep, revise, or retire: decided on a date. Put the review in the calendar when the rollout ends. Keep what beats its baseline; revise what almost does, the exception list usually says how; retire what does not. A retired workflow is not a loss: it taught you what the next one needs.

What goes in the dossier

  1. The before and after. The baseline week and the measured period, side by side.
  2. The verdict and its date. Keep, revise, or retire, signed by the owner, with the next review date already set.
  3. What you would do differently. Two or three lines, written while it is fresh. This is what makes the second workflow cheaper than the first.

With the dossier complete you have run the whole loop once; operating well at this altitude is a destination in itself, and the next constraint is where the loop runs again.

Takeaway

Count verified outcomes, not usage, and price them against the baseline. The number, not the mood, makes the keep, revise, or retire call.

Operating well at this altitude

Many firms should operate here, and some should stay. Operating well at Stage 2 looks like: the workflow runs as the default, its dossier is current, its number gets checked monthly, and no heroics are involved. Stage 3 is there when the next constraint genuinely needs machines acting unattended, not because a next stage exists.

Stage 3 · Expansion

Let automation and agents carry more of each workflow

You have won once. Now the loop runs again and again, and the route ladder extends up into automation and agents, with the rule that saves the most pain: rules before agents.

The reading

Stage 3 lets machines act without someone watching each run, and what they act on is your systems and your data. This section gets both ready, with one rule of thumb: grant the least access that does the job, and fix unfit data before you build on it.

Connections grow up at this altitude

A personal connection

A personal connection acts as you, inside your own accounts.

Fine at Stage 1: the blast radius is your own inbox, and you see everything it does.

Example

The assistant that reads your calendar to draft your week's plan.

A business connection

A business connection acts on shared systems, often with nobody watching the run.

It gets its own login and its own narrow permissions, so what it can touch is decided, recorded, and revocable.

Example

The invoice automation reads the accounting system every night under its own account, with invoice access only.

The difference is not the tool; it is the accountability around it. Stage 3 runs on business connections.

Grant access like it will be audited

Three habits do the work here, and none of them needs a security team. Give each automation the least access that does its job: the invoice workflow reads invoices, not the whole accounting system. Give it its own login rather than borrowing whoever built it, so the access survives that person's departure and can be switched off without touching anyone's account. And never let one system hold all three of: access to private data, exposure to content outsiders can write, and a way to send data out; drop any one of those legs and the worst case collapses.

Example

The proposal drafter reads the CRM and the template folder under its own login, and nothing else. When its builder changes jobs a year later, nothing about the automation changes, and nobody has to remember what was connected through her account.

Build the context the firm owns

Stage 1 gave your personal AI its world; this altitude gives the firm one. The firm-scale version is deliberately low-tech: a small set of plain documents the firm owns, in a folder the firm controls: what the firm does and for whom, how it writes, its service conventions, the decisions people keep re-explaining. One routing file on top tells any tool where to look first. Every workflow in this stage starts informed instead of briefed from scratch, and because the files are ordinary documents rather than any vendor's memory, the asset survives every change of tool and model underneath it.

Example

A "how we write" file, a services-and-pricing file, and one page of conventions per major client. The routing file says which applies when. The new intake automation reads it on day one and drafts in the firm's voice from its first run.

Context ages the way data does, so it gets the same discipline as everything else that runs unattended: a scheduled job folds each week's new decisions in and retires what has expired, a periodic check catches duplicates and contradictions, and one named person approves what changes. A firm that adopts a shared context layer without that maintenance owns a contradictory, expensive mess within a quarter.

Sort your data into three piles

  1. Clean. Accurate, current, consistently structured. Build on it now.
  2. Improvable. The right data in a messy shape, like client records where half the phone numbers carry country codes and half do not. Standardize it as part of the build.
  3. Unfit. Wrong, stale, or contradictory, like two systems that disagree about which contact is current. Fix it first.
Automation does not clean data; it industrializes it. The Stage 2 map marked every point where a person quietly repairs bad records on the way through; here that marking becomes law, because a machine repeats the error faithfully, at speed, in every run. That is why unfit data blocks the build instead of being fixed later.
Ready means: least access, its own identity, sorted data, and context the firm owns. An afternoon spent here saves the weeks that follow a build on sand.

Four questions before any build in this stage

  1. What can it see? The scopes are listed and no wider than the workflow needs.
  2. Who is it? Its own credentials, revocable without touching anyone's personal account.
  3. What is it standing on? The data it touches is sorted, and nothing runs on the unfit pile.
  4. What does it know? The context files exist, are current, and someone owns them.

With access scoped and data triaged, start automating, and start with the predictable work: the next section.

Takeaway

Least access that does the job, an identity of its own, no builds on unfit data, and a context layer the firm owns. Machines act at machine speed; readiness is what makes that a feature.

The fastest wins in this stage are not the cleverest ones. This section is about the predictable work, with one rule of thumb: automate the steps that never change before the ones that think.

The boring steps go first

Three things make them the right start. The rule already exists, on a checklist or in someone's head, so automating it is transcription rather than invention. They fail loudly and fix cheaply: a file in the wrong folder is visible and reversible in a way a wrong judgment is not. And every step a plain rule handles costs almost nothing per run, which starts to matter once the platforms below begin metering your volume.

Example · a real one

A large delivery company automated one boring internal task: account-unlock requests. Around eight hundred a month, each dropping from thirty-five minutes of a person's time to twenty, roughly two hundred hours a month recovered, and the build took five hours on a standard automation platform. No cleverness anywhere: plain rules, with manager approval kept exactly where it was.

Know your billing unit

  1. Per task. Some platforms charge for each step executed: cheap for short workflows, adds up on long ones. Built-in helper steps are often free.
  2. Per credit. Others charge in credits, where a standard step costs one credit but AI steps consume by size and complexity, so the bill tracks what the AI actually does.
  3. Per execution. Others charge per workflow run, however many steps it contains: predictable for long workflows.

The unit matters more than the sticker price: the same workflow can be cheap on one meter and expensive on another. Model a month of your real volume on each before committing.

Measure the blast radius before you build

  1. Read-only. Dashboards, digests, morning summaries: if it breaks, someone sees a stale number. Build these freely; they are how the firm learns.
  2. Internal records. Trackers, bookings, handover notes: a mistake costs an internal correction. Build with the harness from section four in place.
  3. Client-facing. Anything that leaves the building runs in parallel against the current way first, exactly per the Stage 2 pilot method, before a client ever sees it.

Two practical companions to that ladder. First, keep the familiar front end: if the team lives in a spreadsheet, let the new workflow read and write that same spreadsheet instead of forcing a new screen; the machinery changes underneath while the habit survives. Second, know the bulk-work boundary: working through a list inside a chat tool, with a person reviewing, is comfortable up to somewhere around one or two hundred items a run, at real usage cost; past that, or the moment nobody reviews each item, the job belongs on an automation platform built for volume. Treat both figures as practitioners' rules of thumb, not physics.

You have outgrown a platform when you spend more time working around it than in it: workflows split in two to fit its limits, exception handling that fights the tool, costs that climb faster than volume. That is a normal milestone, not a mistake, and the map you keep from Stage 2 moves with you.
Named callout · Generative production

One brief, many executions, one brand gate.

The safe first taste of generative AI inside automation is production work: one approved brief becomes many drafts, variants, or formats, produced by a model inside the workflow. What makes it safe is the brand gate: a person or a hard checklist that everything passes before it leaves the building. The shape is predictable even though the content varies, which is why it belongs in this section.

Automate the predictable first. It pays back immediately, it teaches you the platform cheaply, and it leaves a clean skeleton for the judgment steps and agents that come next.

Three checks before scaling an automation

  1. The volume is real. It runs often enough that the build pays back.
  2. The meter is understood. You know what unit you pay in and roughly what a month costs at your volume.
  3. Failure is visible. When it breaks, something tells someone. Silent failure is the expensive kind.

Once the predictable work runs itself, the interesting question is the varying stretch in the middle: your first agent, next section.

Takeaway

Automate the steps that never change first: they are specified, cheap to run, and cheap to fix. Know your billing unit before the bill teaches it to you, and size every build by what happens if it breaks.

An agent is the first thing you will run that chooses its own path through the work. This section is about the first one, with one rule of thumb: give it a narrow job, real boundaries, and a suggest mode to prove itself in.

First, one word that means two different things. The popular beginner courses use "agent" for a chat persona with standing instructions, and if you have built one of those, that was real work; keep it. This stage means something with more consequence: software that acts in your systems, toward a goal you set, choosing the order of its steps within your rules, without you watching each pass. The caution in the next three sections is about the second kind, and it would be overwrought for the first.

Example · the bridge between the two

Give a chat persona a mandatory stop phrase and make it end every session with a short report of what it did and where it struggled. You have just rehearsed ceilings, off-switches, and logging, the whole grammar of the next section, at zero stakes.

What a good first agent job looks like

  1. Narrow. One kind of work in one place: triaging the support inbox, not "handling support". Agents with sprawling toolkits choose badly; small and focused wins.
  2. Frequent and checkable. It runs daily, and a person can tell at a glance whether a given decision was right.
  3. Low-stakes on error. A mislabeled ticket costs a correction, not a client.

Inbox and ticket triage is the classic first agent for exactly these reasons: constant volume, quick to check, cheap to correct.

Two modes, one promotion path

Suggest mode

In suggest mode the agent drafts the decision and a person confirms it.

It does the reading and the routing work, while every outcome still passes a human hand.

Example

The agent labels each ticket and proposes a reply; the team sends it or fixes it.

Act mode

In act mode the agent carries out the decision itself, inside its boundaries.

Act mode is earned, not assumed: a category moves up when its suggestions have stopped needing correction.

Example

Routine password-reset tickets route themselves and get the standard reply; anything unusual still goes to a person.

Start every agent in suggest mode. The path from suggest to act is the whole subject of the next two sections.

You no longer need to build much to get an agent; the platforms you already use offer them as a built-in option. So the first design question is not which agent, but whether: work you cannot check, and work a fixed rule already handles, should not be delegated to an agent at all. Past that gate, the real work is the definition: the goal, the boundaries, the sign-off points, and examples of right answers. An agent is a job description, not a science project.
Write the job description before the agent exists: what it may read, what it may decide, what it must hand to a person, and what a right answer looks like, with real examples attached.

Three things in place before the first run

  1. Boundaries. The systems it may use and the data it may see, scoped in the platform itself, not just described in the prompt.
  2. Sign-off points. The decisions that always go to a person, however confident the agent is.
  3. A scorecard. A simple tally of right, wrong, and escalated, because the promotion decisions ahead will need it.

The harness that makes this safe to leave running is the next section.

Takeaway

First agent: one narrow, frequent, checkable, low-stakes job, launched in suggest mode with its job description written first. Capability is easy to get; the definition is the work.

Everything in this stage that runs unattended needs a structure that keeps it dependable. This section builds that structure, with one rule of thumb: give every unattended run a ceiling, a logbook, and an off-switch.

The harness, in four parts

  1. Ceilings. A spend it cannot exceed, a time limit per run, a cap on steps and retries. Errors in autonomous runs compound, and ceilings turn a bad night into a bounded cost instead of an open one.
  2. An off-switch. One known way to stop it now, that works, that more than one person can reach.
  3. A logbook. Every run recorded: what it read, what it decided, what it did, attributed to the automation that did it. When something looks wrong, the log answers instead of memory.
  4. A runbook. One page: what this is, what normal looks like, what to do when it is not, who owns it. Written for the colleague who is not you.
A health check is a habit, not a system. On a schedule, a person reads a sample of runs and the week's log and asks one question: is this still doing what we meant? Drift arrives quietly, as inputs, tools, and models change underneath the workflow.

Make agents auditable by design

Four working practices make the harness real for agents specifically. Have the agent produce structured output, a table or a labeled list an owner can audit line by line, rather than prose that can hide a wrong call inside a nice paragraph. Require it to report obstacles instead of improvising around them: "could not find the May invoice" is a good result, while a guessed invoice number is not. Keep its toolkit minimal, because every tool it holds is something it can use at 2 a.m. with nobody watching. And treat shared skills and plugins as what they are, software from the internet: read what one does before installing it, because skills can run tools and execute code.

Example

The triage agent ends every night with a table: ticket, label, action taken, and one plain line at the bottom, "3 tickets skipped: sender not in the client list." That skipped line is the practice working. Obstacles reported, not improvised around.

One special case deserves its own fence: an agent that operates a computer directly, screen, mouse, and keyboard. If you experiment there, isolate it like a contractor on day one: its own user account, a machine that holds nothing else, access granted surface by surface, and a hard payment ceiling if a card goes anywhere near it.

Decide who operates it

Someone runs this machinery, and deciding who is part of hardening. On a managed platform the vendor keeps it alive while you own the definitions and the checks; that is the practical answer when volume is modest and nobody wants server duties. Running it yourself buys full control of your data and of cost at volume, and it is a real operating job: updates, backups, monitoring. Data sensitivity, what you already run, volume, and who would honestly maintain it make the call; the full where-should-it-run question, including the middle ground between those two poles, gets its own treatment in Stage 4.

The harness is the product. Ceilings, logs, a runbook, and an off-switch are what turn something that works in a demo into something the firm can lean on.

Three tests of a hardened build

  1. The overnight test. If it misbehaves at 2 a.m., the ceilings cap the cost and the log shows what happened.
  2. The absence test. Someone other than the builder can read the runbook and act, including switching it off.
  3. The drift test. The next health check is in a calendar, and the last one actually happened.

Hardened, an automation can start earning more autonomy, one step at a time: the next section.

Takeaway

Ceilings, off-switch, logbook, runbook, and a scheduled look at what it is actually doing. The harness is what makes unattended work dependable.

Autonomy is not a setting you switch on; it is trust the system earns, category by category. This section builds the ladder, with one rule of thumb: promote on measured performance, and keep some doors permanently human.

The promotion ladder

  1. Suggest. The agent proposes; a person decides. Every new capability starts here, however good the demo looked.
  2. Act with review. The agent acts inside its boundaries; a person reviews shortly after, and undoing is easy.
  3. Act and report. The agent acts on its own and the log says what it did; scheduled spot checks replace per-case review.

Promote one category at a time, when its correction rate has been low and stable for long enough to mean something, against a bar you set before you started. If corrections rise after a promotion, the category moves back down; demotion is maintenance, not drama.

Example

Password-reset tickets ran six weeks in suggest mode with corrections near zero, earned a month of act-with-review, and now act and report. Refund requests live in the same inbox and will never leave suggest mode: money is a permanent door.

Worth saying plainly: this ladder is more ambitious than what the model makers teach their own customers, whose courses keep a person in the loop at every tier and stop there. The difference is deliberate, and it is not bravado; it is the gate. Autonomy here is never promised and never assumed. It is graduated category by category on correction rates you can show, and revoked the same way; the graduation bar, and the demotion that mirrors it, is the part most advice skips.

Doors that stay human, permanently

  1. Anything a client sees. Sent under your name means certified by a person, at every rung of this ladder.
  2. Money. Payments, refunds, pricing: a person approves, whatever the system's confidence.
  3. The irreversible. Deletions, contracts, and commitments that cannot be walked back.

These gates are not training wheels that come off later; they are the design.

A designed handoff is not a compromise. An agent that resolves what it can and routes the rest to a person, on purpose, is the mature shape of automated work, and the same idea scales from one inbox to a zone model for the whole firm.
Trust is per category, earned by numbers, and revocable. The scorecard from your first agent is the promotion file; without it, every autonomy decision is a mood.

Before any promotion

  1. The scorecard says so. A low, stable correction rate over a real period, for that category specifically.
  2. The harness held. No ceiling breaches or silent failures in the period.
  3. The gates stand. The permanent doors are untouched by the promotion.

With trust building workflow by workflow, the remaining question is how the portfolio grows: the last section.

Takeaway

Autonomy is earned per category, against a correction-rate bar set in advance, and some doors stay human by design. Demotion is maintenance, not drama.

Expansion is not one big program; it is the same loop, run again on the next constraint. This section is about growing well, with one rule of thumb: every workflow gets the full loop, and everything that runs gets watched.

The map from Stage 2 is the brief for every build in this stage: watch the work as it really happens, count a real week, name one owner. Two sentences of that method carry the whole discipline: map the real process, and no baseline, no build.

The growth discipline

  1. One constraint at a time. The next workflow is chosen like the first: where work waits now. Wins compound; scattershot builds compound maintenance instead.
  2. The loop, uncut. Route, pilot against the baseline, harden, measure. The steps that feel skippable on workflow five are the ones that fail on workflow six.
  3. Build, buy, or hire, per workflow. Some workflows a standard platform covers, some arrive as finished products, and for some the right answer is a specialist who builds and hardens it to fit. The route logic from Stage 2 decides, workflow by workflow, never as a firm-wide policy.
Example

The Friday pipeline digest broke in March and nobody noticed until May, because a missing nice-to-have looks exactly like a quiet week. The fix took an hour. The two months of decisions made without it did not.

Growth changes the failure mode. One workflow that breaks is an incident; ten automations without monitoring are ten quiet degradations racing to be discovered by a client. The monitoring habit from the harness section is what makes ten manageable.
Monitoring on everything, with no exception for the small ones. Unmonitored automations degrade quietly, and the cheap ones degrade quietest, because nobody is looking.

Three checks on everything that runs

  1. Every automation has an owner and a runbook. No orphans; the list of what runs exists and is current.
  2. Everything is watched. An alert or a scheduled look, sized to its stakes.
  3. Everything paid for itself or was retired. The measuring move keeps running per workflow; the portfolio version of the discipline is Stage 4's job.

When enough runs that keeping the list current is itself a job, you have arrived at the operating engine: Stage 4.

Takeaway

Grow one constraint at a time, run the whole loop each time, and watch everything you leave running. The current, owned list of what runs is the door into Stage 4.

Operating well at this altitude

Operating well here is a complete destination. It looks like: several workflows running under real harnesses, an autonomy log that says what earned trust, and monitoring that catches drift before clients do. Stage 4 is there when keeping the whole picture current becomes its own job, not because a next stage exists.

Stage 4 · Operating engine

Make AI a dependable part of how the firm runs

What changes when AI stops being a project and becomes something you rely on. What this stage builds is light: documents and routines, not platforms, and far lighter than the governance genre implies. Most firms have none of this, which is exactly the opportunity.

If you self-located straight to this stage, here is what the engine gathers up: the personal collections of prompts and instructions become one shared library, the notes on each automation become one list, the guardrails around each build become one page of rules, and the measuring of each workflow becomes one review of the whole. Each earlier stage teaches its piece in full; this stage only assumes they exist somewhere, however informally.

The reading

The operating engine starts with two short documents: the rules, and the list of what runs. This section builds both, with one rule of thumb: write down what is allowed and what is running, and keep both current enough to trust.

The policy and the register

The one-page policy

The policy says what your firm does and does not do with AI.

Six short sections cover it: scope, approved tools, data that never gets pasted, where a person signs off, who fields questions, and when this page is reviewed next.

Example

Client financials never go into consumer tools; the approved tools are listed below; the operations lead fields questions.

The register

The register is the single list of everything that runs.

One entry per automation: what it does, who is responsible for it, what data it touches, where it runs, where its instructions live, and whether it is still in use. Keep it wherever the firm keeps its working documents; what matters is that there is exactly one.

Example

Anyone can open it and know what runs, who looks after each one, and which ones were retired.

The policy is the rules; the register is the reality. Together they answer the two questions every serious client, insurer, or regulator asks: what do you allow, and what actually runs?

Every automation has one responsible person

The most important thing on the list is the name next to each entry. An automation nobody is responsible for runs until it breaks, and then fixing it is nobody's job, which in practice means it stays broken or quietly keeps producing the wrong thing. One person per automation, someone who knows the workflow and answers for it; when that person leaves, the entry gets a new name the same week or the automation gets retired. The same goes for the documents your automations rely on as their source of truth, such as a price list or a description of your services: name who may change those, because a quiet edit there changes what everything built on them does.

Example

The "how we price" document feeds three automations, and the register names the operations lead as the person who owns it. When a partner wants a rate changed, the change goes through her, and all three automations change behavior on the same deliberate day instead of drifting one by one.

Why a simple list is enough

The biggest software vendors have started building this same kind of inventory into their own products, because even they found that nobody could reliably say what was running. Yours can be far simpler and do the same job. It is also the answer sheet you will be asked for: when an insurer or a client wants to know what touches their data, the list already says. And it makes updates calm: when a tool or model changes, the list tells you exactly what to re-check.

Most firms have none of this, and not only small ones: ask a leadership team anywhere what AI actually runs in their business today, and few can answer with confidence. The governance literature makes the fix sound like a program with a steering committee. It is an afternoon: six sections on one page, one list, one review date. Light is the point, because heavy governance dies of neglect.
If it runs, it is on the list. No automation without an entry, and no entry without a person responsible for it.

Three signs this is working

  1. Someone owns the documents. The policy's review date has been honored at least once.
  2. The list matches reality. Pick an entry, find the automation; pick an automation, find its entry.
  3. They get used. Questions actually reach the named person, and new builds start by adding an entry.

With the rules written and the list current, one question remains that most firms never ask deliberately: where does all of this actually run? The next section decides.

Takeaway

One page of rules, one list of what runs, both owned and current. This is an afternoon's work, and it puts you ahead of most firms of any size.

Every AI tool your firm uses does its work somewhere: on a provider's computers, or on your own. By this stage those choices add up to how much control, cost, and dependency the firm carries, so this section makes them deliberate, with one rule of thumb: for everything that runs, you can say where it runs and why.

A choice you already make elsewhere in the business

Your email almost certainly runs at a provider. Your files might too, or they might sit on a machine in the office. Nobody agonized over those choices; each followed a real need. AI is the same decision in new clothing: most tools are services you rent, some can run on computers you own, and the right answer depends on the work, not on fashion.

Using it as a service

The provider runs everything, and you use it through an account.

You start in minutes and always have the newest version. Your protection is the contract: the plan you are on decides what the provider may and may not do with your data.

Example

The firm's AI assistant and its automation tool, both subscriptions, both the provider's problem when something breaks at night.

Running it yourself

The software runs on computers your firm controls.

It takes real effort to set up and someone has to keep it healthy. In return, nothing leaves machines you own, and the cost stays the same however much you use it.

Example

A firm runs a freely available model on its own server for the documents that may never leave the building.

Between those two ends sit real middle options, and most firms end up using several at once. You can rent a private space at a large cloud provider, where the software runs on their machines but inside a walled-off area only your firm can reach. You can use a provider's tool configured so that your data stays in your own systems, or in your own country. The point is not to pick a camp; it is to match each automation to the protection it actually needs. Everyday work runs comfortably as a service. Work with a hard constraint, such as client confidentiality or a rule about where data may be stored, earns a more controlled option. And paying for maximum control everywhere means paying for a constraint you do not have.

Decide where the firm's knowledge lives

Many firms build up a small set of documents their AI tools read before doing any work: what the firm does and for whom, how it writes, what its services cost. If yours has these, they deserve their own decision, because every tool depends on them. The best home is usually the simplest: ordinary files in a folder the firm controls, such as the shared drive it already uses. Ordinary files can be read by any tool, moved anywhere, and survive every change of provider. Keeping that knowledge inside one tool instead, in its settings or workspace, is more convenient day to day, and it quietly ties you to that tool: switch providers, and the firm's accumulated knowledge has to be rebuilt from scratch.

Example

The firm keeps these documents in the same cloud folder as its client files. When it later changes AI assistants, the new tool reads the same folder and knows the firm on day one.

Two questions that protect you

First: when this breaks, whose job is it? For a service, the provider's; for anything you run yourself, someone at your firm, and that someone should be named before the breakage, not during it. Second: how would we leave? Providers raise prices, retire products, and get bought. Leaving calmly takes three things you can arrange today: the firm's knowledge stays in files you own, you have tried exporting your work from each tool once so you know it works, and each automation's instructions say how the work would get done by hand for a week if it had to be.

Example

The instructions for the invoice automation end with one short paragraph: how to export it, and how invoices get processed by hand from the same checklist it was built from, if the tool is ever down.

A look ahead · Many agents working together

Fleets of AI agents that coordinate each other are coming, and most firms should not start there.

You will hear more and more about systems where many agents pass work between themselves. The idea is real, and it multiplies everything this playbook taught you to keep in check: what the system can reach, what it can spend, and what can go wrong while nobody watches. The honest readiness test is simple: your individual agents run reliably, they rarely need correcting, and your list of what runs is current. Until then, several simple agents doing one job each beat one clever fleet.

Make it a decision, not an accident. For everything that runs, the firm can say where it runs, who takes care of it, and how it would move if it had to.

Three questions to ask about what you run

  1. Where does it run? Pick any automation on the register; it says where the work happens and who takes care of it.
  2. Could you leave? You have tried getting your work out of each tool, and the firm's knowledge files would move untouched.
  3. Could you change calmly? When a provider updates or retires something, the register tells you what to re-check.

That settles where things run. What keeps all of it alive is people, and that is the next section.

Takeaway

Match each automation to the protection it actually needs, keep the firm's knowledge in files you own, and know for everything that runs where it runs, who cares for it, and how you would leave. These are decisions, not projects.

The engine is documents and routines, and documents and routines only stay alive when people look after them. This section covers the people side, with one rule of thumb: give every standing piece one named person, and lead adoption by example rather than by announcement.

One name per piece, not new jobs

This stage leaves the firm with a few things that need ongoing attention: the list of what runs needs its reviews to actually happen, the shared collection of proven prompts and templates needs occasional pruning, and people's questions need somewhere to go. Give each of these one named person. These are not new positions and this is not an org chart; in most firms it is a few hours a month folded into jobs people already have, and in a small firm one person can carry all of it. The reason for the names is simple: a responsibility held by "the team" is held by nobody.

Example

In a twelve-person advisory firm, the operations lead makes sure the quarterly review happens, a senior consultant tidies the shared collection now and then, and questions go to the partner who wrote the policy. Nothing changed on paper; three things stopped falling through the cracks.

A team does not out-adopt its leader

Adoption follows what leaders visibly do, not what they approve. A leader who uses the tools in the open, drafts with them, shares the instructions that worked, brings their own automation to the review, gives everyone else permission by example. The opposite is just as visible: a leader who announces AI and never touches it teaches the firm that the tools are for other people. This never stops mattering; the steady state is simply the point where working this way has become ordinary.

Example

A partner opens the Monday meeting with the brief she used to draft a client memo, including the two corrections she had to make before it was right. That five-minute habit does more for adoption than the training budget.

Grow people who build, not just use

Three habits do more than any training program. Hold a build day: each person automates the task they like least, self-chosen, so nobody's work is automated at them, and whatever proves itself goes into the shared collection with the builder's name on it. Ask it in interviews: "show me how you work with AI" says more than any line on a resume, and it tells candidates what kind of firm they are joining. And point the firm's own tools at its own people work, starting with onboarding.

Example

The firm's onboarding manual, converted by a model into short modules with a quick check after each one and a first-day practical task marked against a checklist. A new hire's gaps show up as flagged answers to talk through together, and the manual finally gets read.

Adoption is cultural before it is technical. Announced tools get quietly ignored; demonstrated tools get copied. The habits in this section are deliberately cheap, because the expensive version, a rollout program with mandatory training, mostly produces attendance.
One name per piece, and leaders who go first. That is the entire people plan, and it is what keeps everything in this stage from going stale.

Three signs this is working

  1. Every piece has its person. Ask who looks after the list, the shared collection, and the questions, and get an answer without a pause.
  2. The last build day left something behind. Something built at it still runs, and its builder gets the credit.
  3. The leaders' own use is visible. Anyone in the firm can name something the leadership actually does with the tools.

People keep the engine alive; numbers prove it is worth keeping alive. What the whole portfolio returns is the next section.

Takeaway

A named person for every standing piece, leaders who visibly use the tools, and people who learn by building. The people side costs hours a month, and without it the documents go stale.

Earlier in the playbook you measured one workflow; the engine measures all of them. This section sets the rhythm, with one rule of thumb: every automation on the list shows what it delivered, on a schedule it cannot skip.

The discipline is the same one that proved your first workflow: compare what the automation delivers against what the work cost before, and decide from the number, not the mood. What changes at this stage is only the scope: you are now looking at everything the firm runs, together.

The review, one automation at a time

  1. What did it deliver? Results and costs since last time, taken from the automation's own records. The responsible person brings the numbers; the meeting does not reconstruct them.
  2. Is it fading? Compare against its best stretch, not just its starting point. Automations fade as inputs, tools, and models shift around them, and the review is where fading gets caught on a schedule instead of by accident.
  3. Keep it, fix it, or retire it? Decide, and write the decision down with a date. Retiring things is normal; a firm that never retires anything is not measuring, it is admiring.
Example

The person responsible for the intake summaries arrives with this quarter's count of summaries the team actually used, compares it with the best quarter so far, and writes "fix": the new document type it mishandles is already named in its own logs.

What each decision leads to

Each decision feeds something useful. A keep with a strong number is the argument for taking on the next workflow, with evidence attached instead of enthusiasm. A fix leaves the review as a clear work order, because the automation's own records usually say what went wrong. And a retirement feeds the shared library: what the workflow taught goes onto the shelf before it is switched off, so even a retirement leaves the firm faster than it found it.

Example

A meeting-notes automation whose summaries nobody actually uses gets retired at its second review. Its process notes and its brief go to the shared library first, so the next attempt at the same problem starts warm instead of from zero.

These numbers are what earn the engine its budget. Measured wins argue for the next build better than any enthusiasm, and the retired automations are what make the wins credible; a list where everything supposedly works convinces nobody.
Keep the measuring simple, and the schedule fixed. For each automation, three plain questions: what did it deliver, what did it cost, and is that worth keeping? Quarterly works for most firms; look monthly at anything new or busy, and twice a year is enough for the stable and quiet. Set the dates in advance, let nothing skip its turn, and if the review takes more than a morning, the measuring has grown too clever.

Three signs the review is honest

  1. Nothing gets skipped. Every running automation shows fresh numbers.
  2. Fading gets noticed. Anything flat or slipping is compared against its best stretch.
  3. Decisions get written down. Dated, in the register, retirements included.

The numbers keep the engine honest. The next section prepares you for the people outside the firm who will ask about all of this.

Takeaway

On a fixed schedule, for every automation: what did it deliver, what did it cost, keep it or not. A firm that never retires anything is admiring its automations, not measuring them.

Sooner or later someone will ask what your firm does with AI: a client before they sign, an insurer at renewal, eventually a rule. This section turns that question into a strength, with one rule of thumb: whoever asks, the honest answer already exists in writing.

The question is coming, and it is an opportunity

Firms rarely lose work for using AI; they lose it for being vague about it. A clear, unhesitating answer signals exactly what the person asking wants to know: that someone here is in charge of this. And because most firms cannot yet answer the question well, answering it plainly sets you apart more than the tools themselves do.

The two answers worth preparing

The engagement-letter line

Clients are told, in the letter, which tools touch their work.

Name the tools and what they are used for. A vague "we may use AI" line protects nobody and reads like it is hiding something.

Example

Drafting and document review are assisted by [named tool] under our review; your data is not used to train it.

The answers, ready in advance

The questions about data and AI are answered before anyone asks them.

Clients and insurers increasingly send question lists: what do you use, what does it touch, who checks the output. The list of what runs already holds those answers; a short summary kept ready turns a week of scrambling into a same-day reply.

Example

The insurer's AI questions get answered from the current list of what runs, the same week, without a meeting.

If your people picked up the habit early of noting what a model touched in each piece of work and who checked it, none of this is new effort: the engagement letter simply says at firm level what the work already says internally, and the honest sentence writes itself from practice.

Example

A client's procurement team sends a twelve-question list about AI use. The answers come from the policy page, the summary of what runs, and the engagement letter already on file, and the reply goes out the same week.

What the law actually asks of a firm like yours

Less than the headlines suggest. For a typical services firm using general-purpose tools, the current rules come down to two things: your people know how to use the tools responsibly, and nobody is misled about dealing with AI, so chatbots say they are chatbots and AI-made content is labeled as such. The heavy obligations target uses like automated hiring decisions, and they sit mostly with the companies that build the models, not with firms that use them. A short internal training and the one-page policy from the start of this stage cover most of what is asked.

This is the one part of the playbook that touches law, and law moves. Treat this as orientation as of mid-2026, not as advice; your counsel has the final word for your firm.
The honest answer already exists in writing. The letter names the tools, the summary of what runs is current, and your people can say what they use and how it is checked. That is the whole preparation, and it fits in an afternoon.

Three signs you are ready

  1. The letter says it. A client can learn from their engagement letter which tools touch their work, without having to ask.
  2. A question list holds no fear. An AI questionnaire arriving today would be answered this week, from documents that already exist.
  3. The answers match. Ask anyone at the firm what it does with AI, and the answer agrees with the policy.

The last section turns everything the firm has learned into its most compounding asset: the shared library.

Takeaway

An honest letter, ready answers, and people who can say what they use. Firms rarely lose work for using AI; they lose it for being vague about it.

The first section of this stage built the list of what runs; this one builds its companion: the collection of what works. They are different things. The register exists so you stay in control; the library exists so the firm gets faster, and its rule of thumb is: treat proven prompts, templates, and instructions like any other thing of value the firm owns.

Personal speed vs firm speed

Personal collections

A personal collection makes one person faster.

It lives in someone's account, improves when they remember, and leaves when they do.

Example

The best proposal brief in the firm, known to exactly one senior.

The shared library

The shared library makes the firm faster, permanently.

The proven pieces live in one agreed place, each with a note on what it is for and who brought it in.

Example

A new hire uses the firm's proposal template in week one and produces a house-standard draft.

Promotion is the habit: when something personal proves itself, it moves to the shared shelf, with credit. And the shelves are organized by the work, not by the tool: a proposal shelf, an intake shelf, a reporting shelf. People reach for the library with a task in front of them, and a library sorted by tool names goes stale with the first change of provider.

Example

A new consultant with a proposal due opens the proposal shelf and finds the brief, the tone guide, and the checklist the firm's best proposals already use. She never has to ask which tool the firm prefers; the shelf answers it in passing.

Keep it healthy, lightly

Three habits are the entire maintenance plan. One good copy of each thing, improved in place rather than copied into private variants, because variants are how quality quietly forks. A one-line note when something changes, so a bad change can be undone. And an occasional tidy-up by whoever looks after the collection: retire what has gone stale, and re-try the important pieces when the tools underneath them change.

The library is why the fifth workflow costs a fraction of the first. Every documented process, proven brief, and set of instructions is a head start on the next build. Firms that keep them compound; firms that keep starting over do not.
Once something works, capture it so someone else can run it without its inventor. That principle built your personal collection back in Stage 1; here it becomes how the whole firm keeps what it learns.

The shared shelf, concretely

  1. One agreed home. Wherever the firm already keeps shared documents; the place matters less than there being exactly one.
  2. A promotion habit. Personal wins get moved, named, and credited.
  3. An occasional tidy. Someone prunes and re-tries the important pieces now and then. A small standing habit, not a committee.

That closes the engine: rules, a clear view of what runs and where, people who keep it alive, honest numbers, ready answers, and a library that compounds. The close below is the steady state.

Takeaway

Promote what works from personal to shared, shelve it by the work it does, and keep it tidy. Documented work compounds; undocumented work walks out the door with its owner.

Operating well at this altitude

The engine is a steady state, and it is deliberately light. Operating well here looks like: the list of what runs is current, the numbers are honest, the answers are ready, the library keeps growing, and none of it needs a committee. This is the destination for firms with enough running; there is nothing above it to climb to.

How it works

One method, run at rising scale

The first three stages run the same six-move loop at rising scale: on your own recurring work, then on one shared workflow, then across many workflows with automation and agents. The fourth stage is different in kind: the operating engine, the light set of standing rules and shared documents that keeps everything you've built running and compounding. Six principles sit underneath all of it.

The loop · six movesRun in order, one workflow at a time
1

Find the constraint

What actually breaks first, or drains the most hours. One bounded task, not a whole function.

2

Map the work

Watch the process as it really happens, capture a baseline, and name one owner. You can't improve or prove anything without the before-number.

3

Choose the route

Pick the simplest option that clears the bar: a fixed rule, a single model call, a set-up assistant, an automation platform, or an agent. Move up a rung only when the one below genuinely runs out.

4

Run the pilot

Run it beside the current way, decide the go/no-go rule in advance, put a time limit on it, and treat "stop" as a legitimate result.

5

Harden and own

Add the guardrails, name the owner, and make it the default way the work gets done.

6

Measure the return

Track verified outcomes, not how many people are using it. Retire whatever stops paying its way.

Move 6 hands back to move 1, pointed at the next constraint.
The six principlesUnderneath every stage

Problem before tool

Start with the bottleneck you actually want to clear, and let that choose the tool, not the other way around.

Rules before agents

An AI that runs on its own is only as reliable as the process behind it, so define the workflow as clear steps and rules first, then let it take over.

AI executes, humans certify

Let the system do the work, but a person reviews and signs off on anything a client sees or that would be costly to get wrong.

The harness is the product

What makes an automation safe to trust is the structure around it: spending limits, clear boundaries, an off-switch, and a record of what it did.

Ground every answer

When a model gives you an answer, make it cite and check its sources, because you stay accountable for the result.

Codify and delegate

Once something works, capture it as a repeatable process, so the next time is faster and someone else can run it without you.

Book a working session

Forty-five minutes on one of your workflows

Bring the workflow that bothers you most. We map where the work leaks, tell you what we would run the loop on first, and you keep the notes either way. No deck, no obligation.

The AI Operating Playbook, by Sprint Assembly
Link copied