The training

The AI Delivery Training

After the workshop, the practice.

The workshop gives your teams a level, a bottleneck and a plan. The training is how the plan becomes normal working practice. We teach in cohorts, in your repositories, on your real tickets, and we shape the content around the gaps the assessment actually found rather than a fixed curriculum.

Comes after the workshop6 to 12 per cohortOn site, remote or hybrid
Prerequisite
The workshop first. It defines what your cohorts need.
Cohort size
6 to 12 people. Small enough that everybody is seen.
Format
On site, remote or hybrid. Your choice per cohort.
Taught on
Your repositories, your backlog, your conventions.
Content
Composed per cohort from the areas below.
Price
Per cohort. The assessment beforehand is free.
Why in your codebase

A skill learned on a sample project stays on the sample project

We have watched teams come back from good, well-taught courses and change nothing. Not because the teaching was wrong, but because the gap between the exercise and the real system is where the habit dies.

Your conventions are the hard part

Prompting is quickly learned. Getting AI to respect your error handling, your middleware pattern and your naming is the part that takes practice, and it can only be practised on your code.

The exercises are your tickets

People work through real items from your backlog. What they produce in the training is work you would have done anyway.

Competency, not attendance

A certificate for sitting in a room proves nothing. We assess against defined competencies, so you know who can actually do it afterwards.

Cohorts keep the standard alive

Six to twelve people learning the same practice at the same time gives you a critical mass that outlives the training.

What we cover

Three areas, mixed to fit the cohort

These are the areas we teach in. Which of them a cohort gets, and how deep, comes out of the assessment. A team whose bottleneck is testing does not need the same sessions as a team whose bottleneck is requirements.

Area 01

Working with AI, properly

The honest version of AI fluency. What the models are good and bad at, where they quietly fail, and how to tell a plausible answer from a correct one. Claude Code and the other tools your team already pays for, used the way people who do this daily use them.

Most teams believe they are past this. Then we look at how much AI output actually ships without rework, and the number says otherwise.

What people can do afterwards
  • A shared vocabulary across the team
  • Honest judgement about AI output
  • Personal habits that stop contradicting each other
Area 02

The team standard

The context system and the discipline of keeping it alive: the files the AI reads before every task, who owns them, and what happens when they go stale. Specifications precise enough that an agent can build from them. A review method that everyone applies the same way.

This is the part that moves a team rather than a person. It is also the part that quietly decides whether anything from the workshop survives the next quarter.

What people can do afterwards
  • Context files the team wrote and will maintain
  • Specs practised on your own tickets
  • One review method instead of ten
Area 03

Agents, security and leading it

For teams heading further: running agents on real work, scoping what they may touch, reading what they did and knowing when to stop them. Alongside it the boundaries, what may go into a prompt and what may not, and how policy holds from the organisation down to the individual.

For leads and architects, the same material from the other side: what to ask for, what to measure, and what not to approve.

What people can do afterwards
  • Agent workflows on your own repositories
  • Clear rules on data and prompts
  • Leads who can tell progress from activity
What gets assessed

What we score against

At the end of a cohort people are assessed on what they can do, not on whether they turned up. You get the results per person and per team.

Context engineering

Can they write and maintain the files the AI reads before every task?

Specification quality

Can they write a spec precise enough for an agent to execute?

Review judgement

Can they tell correct output from output that merely looks correct?

Test quality

Can they get meaningful tests rather than coverage padding?

Security boundaries

Do they know what may go into a prompt and what may not?

Agent supervision

Can they scope, run and stop an agent on real work?

What you get

What exists when a cohort finishes

Not a certificate and a slide deck. Artefacts in your repositories and a clear picture of who can do what.

How it runs

The practical setup

We fit the delivery around your sprints rather than pulling everyone out for a week.

Cohort

6 to 12 engineers per cohort, mixed roles where it helps.

Cadence

Sessions spread across weeks, with practice on real work in between.

Format

On site, remote or hybrid. Remote works if cameras stay on.

Assessment

At the end of a cohort, against the competencies above.

The practical questions

What people ask before booking

Do we need the workshop first?

Yes. The workshop is where every team gets its level, its bottleneck and its plan, and that is what tells us what a cohort actually needs. Without it we would be guessing at a curriculum, which is exactly the generic training we are trying to avoid.

Can we skip ahead to the agent material?

We would advise against it, and here is the honest reason: agents amplify whatever your conventions are. If those conventions are not written down yet, starting there just makes the mess faster.

Does remote work?

Yes, if cameras stay on and people have their own repositories open. The exercises are hands-on, so passive attendance shows up immediately in the assessment.

What if someone does not pass?

Nobody fails. We tell you who is ready and who needs another round on which competency. That is the point of assessing at all.

Can we mix roles in a cohort?

Often yes, and for the team standard area we recommend it. Requirements and testing improve fastest when the people from those phases learn together with the engineers.

What does it cost?

Per cohort, depending on scope and format. The assessment beforehand is free and tells us both which tracks you actually need.

Next step

Start with the free assessment

Everything starts with the 30-minute assessment. It tells us where each team stands, and the workshop that follows tells us what the cohorts need, so you are not paying for training your people do not need.

Assessment · Free

Delivery-Level Assessment for your organisation

  • 30-minute conversation with a senior practitioner
  • Evaluation of your six delivery phases
  • Placement on DL1–DL5
  • One bottleneck, one next step - in writing
Please enter your first name.
Please enter your last name.
Please enter your company name.
Please enter a valid email address.
Fields marked with * are required

Free, 30 minutes, no system access. We use your details only to arrange the assessment.