Skip to content

AI training for your team

Corporate AI training in three layers: everyone gets a standard, each area gets its own repeated work running differently, and a small internal team gets the twelve weeks that keep it alive. Not a course your people attend and forget.

The starting point

Four things are usually true before this starts. Say which ones you recognise, and ignore the rest.

  • AI became a meeting topic and the day to day stayed exactly the same.

  • Whoever tried it did it alone, each in their own way, with no shared standard and nothing written down.

  • The knowledge sits in the heads of two or three people, and it leaves the building when they do.

  • The processes stay manual: proposals that take days, support repeating the same answer, a report assembled by hand every week.

What the programme is for

To leave the operation running with AI on its own, at a standard you can hold, with the people you already have.

01

A single standard of context and use

02

A method the whole team follows

03

A trained team

04

Accompaniment until autonomy

Three ways in, by the size of the company

The same work is shaped differently depending on who is in the building. An owner with no team does not need a champions cohort, and a company with departments does not need a session with the founder every fortnight. Pick the row you are on.

01

CompanyAn owner, with a small team or none

Eight sessions, one at a time

You are the operation. The sessions run with you, in plain language, no terminal and no code. Session two builds the document that holds everything about the company, and from session three each one kills a task that eats your week.

ShapeEight sessions, about an hour each

02

CompanyA company with areas and teams

The programme, in three layers

A shared standard for everyone, then each area working on its own repeated tasks, then a small internal team through the twelve weeks that keeps it alive. This is where day to day productivity across the company actually comes from.

ShapeThree layers, twelve weeks on the third

You are reading this one
03

CompanyA company that wants the system built

Diagnosis, build, handoff

We find what the bottlenecks cost, build the modules that remove them, and hand the system over trained and documented. Training is inside it, at the handoff, rather than being the product.

ShapeAbout three months, four phases

They are not alternatives to argue between. Most companies start on one row and end up on another, and the conversation that decides which is always the same one.

Three layers, one company

Productivity with AI is per role, not generic. A single course for everyone produces exactly what you already have: a company that talks about AI while the day to day stays the same. So the programme runs in three layers, each for a different group, each with a different job.

01

The base

WhoEveryone

The jobOne shared standard, so the company stops improvising.

  • How it actually works, and where it stops being reliable.
  • What never goes into it. A company that puts hundreds of people on AI with no data rule has bought a leak.
  • Where your approval gates are and how to work them.
  • Your own way of using it, written down, so it stops depending on who taught whom.
02

By function

WhoEach area, in working groups

The jobWhere the productivity actually shows up.

  • Sales, finance, support, operations. Each group works on its own repeated tasks.
  • Not prompting exercises. The proposal they write, the report they assemble, the ticket they answer.
  • Each group leaves with the tasks it repeats most already running differently.
  • What is worth automating, and what is faster left as it is.
03

The multipliers

WhoA small internal team

The jobThe reason the other two layers do not decay.

  • Twelve weeks, one session a week, detailed below.
  • They build and maintain the shared assets: the second brain, the skills library, the agents.
  • They are who the company asks when something breaks, instead of asking us.
  • They carry it forward after the programme ends.

What holds the three together is the shared library. Without it, two hundred people invent two hundred prompts and quality is random. With it, one good skill built by the multipliers is used by everyone.

How you know it worked

Most corporate AI training sells attendance. This one is measured, using the same instrument as the diagnosis: pick real recurring work, time it, change it, time it again.

  1. 01Before we start, each area names three to five tasks it repeats every week, and we time them as they run today.
  2. 02The programme works on those exact tasks, not on examples.
  3. 03At the end we time the same tasks again, and the difference is the report.

No promised percentage. The number comes out of your operation, not out of a case study from somebody else's company.

Inside the twelve weeks

Layer three in detail. Twelve work fronts, organised into four blocks of three weeks. Each block ends with something running, not with a module marked complete.

01Weeks 1 to 3

Diagnosis and foundation

  • Diagnosis of the processes with the people who run them
  • The company second brain, structured with permissions per team
  • The company's context and rules, written by the team
02Weeks 4 to 6

Operation and standard

  • Operating the system on real work
  • A library of skills built from what the team repeats most
  • Connectors, so the AI reads your real data
03Weeks 7 to 9

Agents and architecture

  • The agent map for the operation
  • The harness scorecard and the gaps in priority order
  • Approval gates, each one tested by breaking it
04Weeks 10 to 12

Autonomy and governance

  • Deploy, so it runs off somebody's laptop
  • Sensitive data separated, and what the law asks of you
  • The weekly ritual, and the roadmap for what comes next

The twelve weeks

One session a week, with accompaniment between them. Every week ends with something your team can show, because in a programme this long a deliverable is the only honest way to promise progress.

  1. Week 01

    How AI actually works

    You leave withA shared read of where your operation sits today, and what is missing.

  2. Week 02

    The company second brain

    You leave withThe mother folder created and the team reading from the same source.

  3. Week 03

    Context beats prompts

    You leave withYour context and rules written, so nobody retypes them every session.

  4. Week 04

    Operating from the terminal

    You leave withOne real company task executed by the system.

  5. Week 05

    Skills

    You leave withThe first working skill, built from the task your team repeats most.

  6. Week 06

    Connectors

    You leave withAt least two connectors live, reading your real data.

  7. Week 07

    Agent architecture

    You leave withThe agent map drawn, and the first subagent running.

  8. Week 08

    Harness engineering

    You leave withYour harness scorecard, with the gaps in priority order.

  9. Week 09

    Hardening

    You leave withThe approval gates installed, each one tested by breaking it on purpose.

  10. Week 10

    Deploy

    You leave withOne job running off your machines and reporting back on its own.

  11. Week 11

    Security and compliance

    You leave withSensitive data separated and protected, plus the list of what the law asks of you.

  12. Week 12

    Cadence and demo day

    You leave withThe weekly ritual running, and the roadmap for the next agent written.

What it asks of you

Constant presence, without stopping the operation.

One session a week

Twelve of them. The same named people every week, the ones who run the process, not a rotating audience.

Accompaniment in between

Where the week's work gets applied to real jobs and the team gets unstuck. Training on its own disappears the moment the routine tightens.

One review a month

With whoever decides. Progress, blockers, and what needs a call. Three across the programme.

What stays with the company

Tools used in the daily work, not a consulting report that lives in a drawer.

  • The company second brain, structured and populated
  • The skills library, built from your own repeated work
  • The agent map, and what each one is allowed to do
  • The approval gates, documented and tested
  • Operating documentation, written for the people who run it
  • A trained team, operating the method

Out of scope

The programme structures the operation and trains the people. The work itself stays with your team, which is the point.

  • We do not run your operation or produce your work.
  • We do not build your production system here. That is an implementation project, and it is priced separately.
  • No unsupervised automation on anything risky, in writing.
  • Not permanent support. The programme ends in autonomy, not in a dependency.

Three rules that make it stick

01

The people who use it are in the room

We train the people who will actually run the process, not only a manager who will relay it. Training the wrong layer is the most common reason a system dies after handoff.

02

Approval gates are taught, not assumed

Nothing sensitive leaves the company without a person saying yes, and your team learns exactly where those gates are and how to work them.

03

You own what gets built

Your build, your data and your configuration are yours. The documentation stays with you, so the knowledge does not leave when we do.

How it starts

The three layers can run together or on their own, sized to the company. It also works as the way in, because by the time the base is done and the areas have been through their own work, you know where AI pays in your operation.

  1. 01A conversation, and the programme agreed
  2. 02Kickoff, and the team named
  3. 03Diagnosis of the processes with those people
  4. 04The twelve weekly cycles begin

Questions about the training

How do you measure whether the AI training worked?

With the same instrument as the diagnosis: pick real recurring work, time it, change it, time it again. Before we start, each area names three to five tasks it repeats every week and we time them as they run today. The programme works on those exact tasks, not on examples. At the end we time the same ones again, and the difference is the report. No promised percentage: the number comes out of your operation.

What stays with the company after the training?

Tools used in the daily work, not a consulting report that lives in a drawer: the company second brain structured and populated, the skills library built from your own repeated work, the agent map and what each one is allowed to do, the approval gates documented and tested, operating documentation written for the people who run it, and a trained team operating the method.

How much time does the team have to put in?

Constant presence, without stopping the operation. One session a week, twelve of them, with the same named people every week, the ones who run the process, not a rotating audience. Plus accompaniment in between, where the week's work gets applied to the real job, and one review a month.

Start with an AI diagnosis.

Thirty to forty five minutes. We find out whether there is a bottleneck worth removing, and I tell you honestly if I cannot help.

Book a diagnosis