EN · FRGet in touch
An abstract dark navy graphic: a loose tangle of thin pale threads rising from the center and straightening into an even coral grid on the right.

Maestro AI, by CS Maestro

AI onboarding and process automation. Built where it pays off.

We find where AI saves your team the most time and build it into the software they already use. Then we stay until the work runs without us.

Start with an AI roadmap · See the worked example

"We know we should be using AI. Everyone is trying something different, and nothing has really changed yet."

That is almost word for word what we hear. Someone bought licenses, a few people found a use for them, and the work the company runs on looks exactly the same as it did last year.

We start from your processes. The tools come after.

Where we map

We follow the work as it runs: who touches what, where it waits, which steps come back every week. That gives us two lists: where AI would save real hours, and where it would not. The second list matters as much as the first, and it is usually longer.

What we build

The automation lives inside the software your team already opens: SharePoint, Power BI, the shared inbox. Nothing new to remember. First the data underneath gets cleaned and named, so every answer points at the right information.

What your team keeps

Every judgment call. AI collects, reads and does the first pass; a person on your team approves what gets kept and decides what happens next. We train your own people on your own documents, so the knowledge stays inside the company.

What we do

AI roadmap

A written plan, in plain language, that you can act on with us or without us.

Which processes AI can take over or speed up, and which are better left alone.

When each one happens: what to do in the next one to three months, what comes at six to twelve, and what waits.

Which tools and licenses to buy: Microsoft Copilot, Claude and others, and who on your team gets them.

It is yours either way, whoever builds it.

An example roadmap: five processes with the hours a week each takes and a verdict beside it, from automate now to leave it alone.
An example plan: what to automate now, what waits, and what not to touch.

Process automation

Three kinds of work, all of which we already do for clients and for ourselves.

Manual work turned into automation. Many files into one report, a shared inbox read so nothing gets missed.

The process improved first, so we automate how the work runs, not how the procedure says it runs.

A small tool built for one job, where nothing on the market fits.

The same patterns cover orders, invoices, quotes and weekly reporting.

One process before and after: nine struck-through circles over the words "9 gone", and two coral circles over "2 kept".
Eleven steps by hand became two.

Getting your data ready for AI

AI gives confident wrong answers on messy data. Three things fix that.

One name per thing, so the same customer is not five different customers in five systems.

One meaning per field, written into a plain dictionary anyone in the business can read.

The history lined up, so last year’s figures still add up to what you reported last year.

Skip this step and the AI keeps answering confidently with the wrong figures.

The same customer field under five different names on the left, and on the right a written dictionary giving each field one meaning and one example.
One name per thing, one meaning per field.

Getting the team using it

Hands-on sessions built on your own tools, documents and processes.

People work on their own files, so they leave the session with something real already done.

Each team gets what fits its job, instead of one generic course for the whole company.

The automation keeps running after we are gone, because the people who use it helped shape it.

We leave when your team no longer needs us in the room to get it done.

An onboarding session: a question the operations lead asked on day one, answered from the team’s own files, with the three source files named under the answer.
A question the team asked, answered from their own files.

How many hours is your team losing to this?

Add them up here.

A week of it looks small until you price forty-eight of them.

What the hour costs you, not what it pays: wages, benefits and your share of the payroll contributions.

Across everyone who touches it

We have never taken over all of a process, and would not claim to

 

 

A year of that time, back.

Hours back each week 
Hours back each year 
Full-time people, at 40 hours a week 

What it is worth depends on what those people do instead. Working that out is what the free half hour is for.

Get in touch

AI investigates every number someone questions.

Every time someone questioned a figure in a Power BI report, a specialist traced it by hand, from the report down to the warehouse. Now AI does the digging, and a person on the team decides.

150–250 ha week back to the team, the equivalent of four to six full-time people, without hiring anyone.
Follow one taskEXAMPLE
“Q3 revenue on the regional dashboard shows $4.21M. Finance reports $4.35M. Find out why.”
AIOpened the report with the same filters. It shows $4.21M. Confirmed.
AIFollowed revenue through the data model to the sales table in the warehouse.
AIThe warehouse holds $4.35M, so the gap comes from the report itself.
AIThree invoices booked on September 30 fall outside the report’s date filter. Gap: $140K. Reconciled.
Your teamA person on your team confirms the cause and answers whoever asked.
AI investigatingSigned off by your team · Example task. The figures are invented.

AI does the investigation. Your experts decide.

Where should AI take your company next?

Most teams are trying something, and nobody has decided where it all leads. In thirty minutes we will look at where you are, and whether a roadmap would give you the direction that is missing.

Get in touch