AI Engineering Training — Teams that deliver better with AI
For engineering teams already using AI — Copilot, Cursor, Claude Code, or others — who want a consistent and measurable methodology.
AI Engineering Training is a hands-on, cohort-based training methodology that works on your engineering team's real repository, with delivery, adoption, competence, and business metrics measured before and after the program.
First cohort closed
First cohort closed — case in publication Q3. The methodology continues measuring before and after; the delta report is published once its disclosure is approved.
Who this is for
AI Engineering Training is designed for in-house engineering teams that have already adopted AI tools — but lack a shared methodology and a system to measure impact.
Teams already using AI, but with inconsistent adoption
Your engineers use Copilot, Cursor, Claude Code, or other AI agents — but each person does it their own way. Code quality and speed vary across squads.
Engineering leaders who need to show ROI
You've been asked to demonstrate the return on your AI investment — and today you don't have metrics to back it up. The program measures the delivery, adoption, and competence delta on your own repo.
Teams seeing the hidden cost of AI in their code
Quality incidents have been attributed to AI-generated code, or your security team is overloaded with PR-review volume. The methodology integrates review and governance practices.
Not for
- Teams that haven't adopted AI tools yet and need a generic introduction.
- Individual training or self-paced courses. We work with full teams in cohort format.
- Teams looking for a specific AI tool more than a methodology.
How it works: from Pulse to Craft
Two stages, focused on your team, your repository, and your way of working.
1 to 2-week assessment
We capture a baseline (T0) of delivery, AI adoption, and competence metrics on your own repository. We deliver a gaps report with concrete recommendations.
6 to 8-week cohort
We train your team in cohort format, with hands-on sessions on your real code. We close with a re-measurement (T1) and a delta report across the four metric families.
Methodology: measure before and after, on your repository
The program is built on a simple premise: if we don't move the needle, the data shows it. That's why we use a T0 → T1 measurement system covering four metric families, captured on your own repository and your real team.
Four metric families
Delivery
DORA metrics: lead time, deploy frequency, time to change, and failure rate — captured on the team's real workflows.
Adoption
SPACE metrics: satisfaction, performance, activity, communication, and efficiency — captured at team level, with no individual ranking.
Competence
Technical assessment of AI tool use per squad, focused on prompting, review, and governance — compared against the initial assessment.
Business
Business indicators your team already tracks: hours per feature, cycle time, backlog burn-down. Same method, your metrics.
We measure the system and the team — not individuals.
We train on your repository, not generic exercises
Sessions and the closing project operate on the team's real codebase, with the same stack and AI tools already running in production. It works with Java, .NET, Python, TypeScript, polyglot repositories, and any modern stack.
Three formats depending on where your team is
One methodology, three ways to start. You choose the depth — and we adjust it after Pulse.
Initial assessment
1 to 2 weeks
We capture the T0 baseline on your repo and deliver a gaps report with concrete recommendations. Useful to understand where you stand before committing to a full program.
Full program
6 to 8 weeks
Full cohort, hands-on sessions on your repo, methodology for prompting, review, governance, and delivery. Closes with a T1 re-measurement and a delta report per metric family.
Ongoing reinforcement
Quarterly
Successive cohorts, continuous measurement, and office hours to sustain the methodology over time. For teams that have completed Craft and want to keep the consistency.
How we work — the principles that apply to this service
Methodology, not tool
The program is agnostic to the AI tool your team already uses. It works with Copilot, Cursor, Claude Code, Codeium, or any combination. The value is in the methodology, not the vendor.
Cohort, not individuals
We work with full teams in cohort format. Shared accountability is what sustains the change.
Measure first, promise later
We start by measuring. We close by measuring. We don't guarantee uplift — we publish the delta as it stands. If we don't move the needle, the data shows it.
Frequently asked questions
Is this a course or a consulting engagement?
It's a cohort-based training program with hands-on components on your repository. It's not a generic course and it's not an implementation consulting engagement: the main deliverable is that your team internalizes the methodology.
Do we need a specific AI tool?
No. The methodology is tool-agnostic. We work with Copilot, Cursor, Claude Code, Codeium, or whatever stack the team already uses. The key is that the team is already experimenting with AI — this is not an introductory program.
How long is the full program?
Craft runs 6 to 8 weeks, with sessions distributed across that period. Pulse, the initial assessment phase, runs 1 to 2 weeks. Craft+ is a quarterly reinforcement scheme for teams that have already completed Craft.
When do we see the first results?
First signals appear between weeks 4 and 5, in the middle of the program. The formal T1 re-measurement happens at the close of Craft. Stabilization and business-attributable ROI are observed in the 2 to 3 months that follow.
How do you measure impact without ranking my team?
We measure the system and the team, not individuals. DORA and SPACE metrics are captured at the squad and team level. The competence assessment is anonymized and delivered in aggregate. We publish the delta as it stands, without rankings.
Do you work on our stack and our tools?
Yes. Sessions and the closing project run on the team's real repository, with the same stack and AI tools already in use. Java, .NET, Python, TypeScript, polyglot — all supported.
Related services
Your team needs more than training — it needs the right team. These services complement AI Engineering Training.
Start with Pulse
The first step is a 1 to 2-week assessment on your repo, with a concrete gaps report. No commitment to continue.
or email us: hello@23people.io