AI engineering velocity

Your team has the licences. What it does not have is an agreed way to work.

This is a hands-on programme for software teams, run on site in small cohorts, and the exercises are written from your own code after we have read it. Your developers work on files they recognise, which is why the habits survive contact with a real sprint.

Everyone leaves with working code they built themselves, not a certificate.

Who it is for

Three situations where this is the right kind of help.

Developers and QA engineers on teams that already have the tooling in place and no shared practice around it. If none of the three below sounds like your team, say so on the call and we will tell you to save the money.

✓

Adoption is uneven: a few developers are far ahead and the rest have barely started.

✓

Senior engineers are sceptical, and their scepticism is mostly well founded.

✓

Output quality is inconsistent, and nobody agrees on what good looks like yet.

Before the room

The exercises get written after we read your code.

Every engagement starts with three pieces of discovery. Nothing comes off a shelf, which is also why we cannot quote a day rate without the call.

01
Leadership session

Stack, tooling configuration, and what leadership needs the team to be able to do afterwards.

02
Anonymous questionnaire

Participants describe where the friction actually is, in their own words. Anonymity is what makes the answers usable.

03
Code pattern review

A read of the team’s real patterns and conventions. Exercises get built from this, not from a template.

Three shapes

Scoped to the team rather than to a calendar.

Length follows what the team needs. All three share the curriculum below.

01 / DISCOVER

Discovery and assessment

The pre-engagement work, sold on its own. Useful when you want to know where the team stands before committing to anything longer.

YOU GET BACK
✓A written report on current practice and gaps
✓A scoped programme proposal
✓The anonymised themes from the questionnaire
02 / TEAM

Team programme

On-site, hands-on, small cohorts of developers and QA engineers working through the full curriculum against your own code.

IN THE ROOM
✓Exercises built from your codebase
✓Cohort work, not lecture
✓Capstone build in longer engagements
03 / COACH

Individual coaching

One developer at a time, on their own tickets. Fits either side of a team programme, or on its own for a specific gap.

HOW IT RUNS
✓Recurring sessions pairing on live work
✓Targeted sessions for one gap: review habits, agent workflows
✓Support for a lead carrying the practice back to the team
✓Follow-up for developers who need more time
The curriculum

Three areas, taught against your own stack.

Depth and pace vary by format. The shape does not. Open an area to see what sits inside it.

·Prompting that produces verifiable output
·Model selection, and when a cheaper model needs a stronger one checking its work
·Tool and context configuration
·Coding standards enforced through the assistant
·Where AI belongs in the SDLC, and where it does not
·Data sovereignty
·Reading generated code critically

capstone build · full UI, API and database application, built by the teams · included in longer engagements

What you keep

Four things, and all of them stay with the team.

Nothing here depends on us coming back, and nothing expires when the engagement ends.

Working code

Built by the team during the engagement, running against your own stack.

A prompt library

Shared, reusable, and owned by the team after everyone goes home.

A measurement baseline

Agreed up front so impact can be assessed against something real later.

Shared standards

Review expectations and conventions for AI-assisted work, written down.

What it is not

Worth saying plainly, before anyone signs.

×
Not a certification.

Nobody leaves with a certificate. They leave with code they wrote and standards they agreed to.

×
Not a tooling procurement exercise.

We work with the licences you already hold and configure them properly. Nothing here depends on buying more.

×
Not a promise of a specific velocity number.

Anyone quoting one before reading your codebase is guessing. The baseline is established during the engagement so you can judge for yourself.