Prove the fit. Measure the result. AI automation consulting that gives your team time to grow the business. We find where the hours go, then build around how you work.
How an engagement runsSix steps, in the same order, whatever we build.
Two systems that need connecting or a team of AI agents running your prospecting, the order holds. We study the work before we design, agree the result before we build, and prove a first build on your own data before the rest. At every step you hold something you can check, and the next decision is yours.
- 1Study and measure
- 2Agree the result
- 3Prove a first build
- 4Build beside your current work, then switch over
- 5Hand over
- 6Watch, and read the result
Step 1 of 6 · Proof SprintFirst, we find where the time and money go.
We sit with the people who do the work and follow real jobs from arrival to close: who handles each one, in which system, and where it waits. We look at people, process and tools together, because the cause of a slow job is rarely where it shows.
Then we measure it from your own records: how long a job takes, how often it comes back, what it costs to handle, and what it costs when it is late or wrong.
You get a map of the work as it runs today, with the measurements on it. For most owners it is the first time the whole job has been on one page.
Early in our founder’s years at a Fortune 250 enterprise, one everyday question, how is this client doing, took an analyst two weeks to answer, because only that analyst knew where everything was kept. He built the system that put each client’s full record in one live view. The study finds that kind of work in your business.
- You hold
The map of your work today, measured.
Step 2 of 6 · Proof SprintWe agree the result and its measure before anything is built.
We take you and your leaders through the map: where the time goes, where errors start, what it costs, and where growth is held up. Then you get a plan that ranks the jobs worth building, what we would build for each, and what each should return.
Together we pick where to start and put three things on one page: the result the build must deliver, the number that will show it, and what that number reads today. The finished build is read against that page.
- You decide
Where to start, and the result we build for.
- You hold
The plan, and the agreed result on one page.
Inside the plan · How we choose what to buildEach step of the job gets the kind of build it needs.
We go through your job one step at a time and ask five questions. The answers choose the build.
Does software already do this well?
Then we use it: we set up what you own or recommend what to buy, and connect it to everything around it. Where nothing on the market fits how you work, we build it.
On the mapA won quote books the job and raises the invoice, with no retyping.
Is the right answer always the same?
Then a fixed rule does it. It gives the same answer every time, costs almost nothing to run, and anyone can read why it acted.
On the mapStandard jobs are priced from your price list.
Does the step take reading or judgment?
Then AI does that step inside a workflow whose rules stay fixed. It reads the email, photo or contract and tells you how sure it is of each answer.
On the mapA customer’s email, photos and attachments become one complete request.
Does the next move depend on what turns up?
Then an agent does it. An agent gets a goal instead of a script, works out the steps itself, and takes only the actions you allow. That includes the always-on agents the AI labs sell, which we set up and customise, and agents we build for you.
On the mapFor an unusual job, an agent pulls similar past jobs, supplier prices and the customer’s history, and prepares the quote for your estimator.
Is the work too big for one agent, or too important to go unchecked?
Then a team of agents does it. Each has one job, and one does nothing but check the others against the source.
On the mapOn a large quote, a second agent checks every figure against the price list and the customer’s terms before it reaches you.
Most builds mix several of these, and none ranks above another. Each is chosen for the best result and return, never for complexity that adds nothing. Ask for something specific and we build it, with a result worth paying for behind it.
The first plan for a sales team’s prospecting system called for ten tools and a custom approval screen. After studying how the sales lead worked, we built it on a handful and put the approvals in the chat app already on the lead’s phone. It was delivered in 25 days, and the morning review took five to ten minutes.
Step 3 of 6 · Proof SprintA first build proves itself on your own data.
We build the first piece and run it on real work from your business: your documents, records and requests. What you see working is what you get.
We test it on past work where the right answer is known and show you the results item by item: what it got right, what it sent to a person and why, and how it compares with today. Where there is no past work to test on, we run the new way beside the old on live work.
That ends the Proof Sprint. The measured map, the agreed result and a working first build are in front of you when you decide.
- You decide
Whether to go on to the full build.
- You hold
A first build that works on your own data, and its test results.
See this on your own work. Start with the Operations X-ray, our free first look at your operation, or book a call about the job.
Step 4 of 6 · The buildYour business keeps running while we build.
Your team works as they do now while each part is built and tested on real work beside your current process. For a time the same work runs both ways, until the new way matches or beats the old on the measures we agreed.
A system nobody uses returns nothing, so we are deliberate about what changes for your people: we keep what works, change what the result needs, and train your team before anything new goes live.
The switch happens when the build is complete and you say go. Until then, nothing in your day depends on it.
- You decide
When to switch over.
Inside the build · How much runs on its ownYou decide what runs alone and what waits for a person.
We set it with you, step by step, from what a wrong step would cost, the rules of your industry, and how you like to run the business. Each step gets one of three settings.
Runs on its own
For steps a fixed rule decides, or where a mistake is cheap and easy to undo. You see them in the record, not your inbox.
Runs, and flags what it is unsure of
The system does the work and rates how sure it is. Anything below the line you set goes to a person, with the reason attached.
Waits for a yes
The step is prepared and held until the person you name approves it. We recommend it for anything that moves money, touches regulated data or cannot be undone.
The line is set on evidence: in testing you see how often each step was right on your own past work with known answers.
We start with more checking than the build will need. As results come in, you can move a step from checked every time to checked only when the system is unsure. That switch is yours, and so is the one that turns the system off.
Every step leaves a record: what came in, what the system did, which rule or check applied, who approved it and when. Each build gets only the access its job needs.
The legal intake we built rates its confidence in every detail it reads from a scanned report. In testing, the low ratings landed on details that were hard for anyone to read, such as partly legible handwriting on an insurance field: exactly the ones a person should look at.
We run our own company this way: new work runs on a small batch and is checked against real output before it scales, every change is recorded and reversible, and a person makes the calls with real consequences.
Step 5 of 6 · The buildWe hand it over working, with a team that knows how to run it.
We train your people on the live build and agree who owns it on your side.
You also get the record: how the work ran before, how it runs now, what each step does, which checks apply, and what to do when something needs a person. When someone leaves, the work carries on. When you hire, the record is the training. When you grow or change platforms, you start from a clear picture.
You own everything we build, including the code.
When the enterprise where our founder led analytics moved its service operation to a new platform, it was a continuation, not a restart, because the work was already mapped and documented.
- You hold
The working system, the record of before and after, and the code.
Step 6 of 6 · The buildSixty days on, the result is read against where you started.
For sixty days after hand-over we watch the build on live work: we see a failure when it happens, fix what breaks, and tune the checks as real volume comes through.
Then we take the step 1 measurements again, the same way, on the same work, and show you before and after against the result we agreed.
With the first result in, we show you where the business could go next. Your plan already ranks the jobs that follow, and the next build starts further along.
The prospecting system we built for a sales team reported any failure in the pipeline to us within a minute, with its cause named, and sent the sales lead the day’s numbers every evening.
- You hold
The before-and-after reading, measured the same way both times.
Questions about working with us
What do you need from us?
Time with the people who do the work, access to the systems it runs through, and past work with known answers to test against. One person on your side owns the decisions, and we do the building.
We already know what we want built. Do you still study the work first?
Yes, and it goes faster when you know what you want. The study gives the starting measurements the result is read against, and makes sure the build fits how the job really runs.
How long does an engagement take?
The Proof Sprint is short. The full build depends on the job, and your plan gives you the schedule before you commit. For scale, one of our builds was working in 48 hours and the prospecting system above was delivered in 25 days.
What happens after the sixty days?
The system keeps running. We can keep watching it, build the next job on your plan, or lead AI across the whole business as your Fractional AI Officer, an ongoing role whose builds follow these same six steps.
The first look at your operation is free.
Get your Operations X-ray, worth having whether or not we work together. Or book a call, tell us about the job, and we will tell you what a Proof Sprint on it would cover.
Want to see the systems first? See what we’ve built
Book a 30-minute call Opens Calendly in a new tab