Your releases keep going out while every team learns to work with AI agents.
Your engineers learn to work with AI agents in waves of 8 to 12, role by role, with mentoring on real work, until your platform team runs the practice itself.
Book a call (opens in a new tab)- Waves of 8 to 12
- measured per team only
- your platform team owns it at the end
A pilot team first, then waves, then your platform team
One team trains while the others keep shipping, so delivery keeps running the whole time.
-
01
Agree the measurement plan first
Before the first wave, we agree what we'll measure, per team only, and when we'll review it together.
-
02
Start with one pilot team
One team goes first. What it learns sets the agenda and the order for every team after it.
-
03
Then train in waves of 8 to 12
Each team trains on its own, 8 to 12 people at a time, in waves we plan with you. 60 developers means 5 to 8 team trainings; 400 means about 34 to 50.
-
04
Hand it to your platform team
Your platform team owns the shared setup and its standards. Every new wave starts from them, and your own people carry the practice from there.
Mentoring is on call after each two-day team training while that team applies it to its own work.
Every role gets an AI training built for its work
Each wave takes the training for its role. Most organizations start with engineering, and we plan the order of the waves with you.
For engineering teams
-
Your whole team working with coding agents the same way.
- Software Engineers
- QA
- Analysts
-
From each engineer's own setup to one pipeline the team shares.
- Software Engineers
- QA
-
Routine platform work goes to agents, and your team approves every change.
- DevOps
- SRE
- Platform Engineers
-
Harden your agents, then use them to harden your code.
- AppSec
- Architects
- Software Engineers
For product and leadership
-
Specs your engineers' agents can build from.
- Product Managers
- Product Owners
- UX
-
What to steer, what to measure, and what to leave alone.
- CTOs
- VPs
- Engineering Managers
For business teams
-
Teams outside engineering hand recurring work to agents, and a person still makes the call.
- Marketing
- Finance
- Sales
- HR
- Service
From new habits to lasting capability
From new habits to lasting capability
Training changes how people work. It lasts when team habits, decision rights, platform ownership, architecture, quality and measurement change with it.
New habits
People learn the techniques, but the system keeps working the old way.
- Experiments stay with individuals.
- The early enthusiasm fades.
- Team rituals stay the same.
- Practices drift without feedback.
- There's no path from first use to scale.
The organization knows what's possible but can't sustain it.
Lasting capability
The practice becomes part of everyday delivery and of how the organization steers.
- Agents work inside real delivery.
- Practices get better every sprint.
- Quality gates and context are owned centrally where that helps.
- Teams become more independent over time.
- A platform team can scale the system across the organization.
The capability keeps growing inside your organization.
Four measures, plus the business outcome you care about
Four measures, plus the business outcome you care about
Every measure has a baseline and a named data source.
The four measures back up the business outcome you name for your context. We measure per team only.
- Lead time from idea to first users
- DORA
- Changes redone before they reach users
- Rework rate
- Target users of what shipped, after 30 days
- Use
- Tokens, agent and review time
- Cost per change
Your platform team becomes how the practice scales
Once several teams work this way, an internal group usually takes ownership of the shared setup and its standards, so each new wave starts from them.
Runtime and access
Route requests to the right model, put MCP and tool access under governance, and run every agent in an isolated environment.
Shared registries
Keep reusable context, skills and agent setups in registries, each with a named owner, so one team can reuse what another built.
Guardrails by risk
Gate changes with evals, approval rules and security scans, scaled to the risk of each change.
Paved roads and onboarding
Make costs visible, ship sensible defaults teams can follow, and onboard people until your own team carries the system.
A shared platform each team can adapt
A shared platform each team can adapt
Delivery teams adapt the shared setup to their own context and feed what they learn back into it.
-
Shared platform
Context infrastructure, agent configurations, quality gates and reference patterns have one internal owner.
-
Pilot evidence
Baseline, outcome, constraints, roadmap and pilot results set the rollout priorities.
-
Delivery teams
The practices move into real refinement, planning, implementation, reviews and retrospectives.
-
Internal owners
Named people take over the technical system, the rollout model and day-to-day operations.
-
Measurement and review
Delivery evidence, the community and deliberate reviews keep the setup useful as constraints change.
Agents became part of how these teams deliver
-
Rollout across teams
We work with AI much more deliberately now. 'Agent first' has become our reflex: do we need to do this ourselves, or can an AI agent handle it? The support can help take away some of the developers' fear of working with AI.
Michael Mohring (opens in a new tab)
Head of CC Product Development, in|sure Workflow & Workplace adesso insurance solutions -
Training
The sessions got our engineers genuinely excited to try things on their own. That's the hardest part. They kept experimenting, sharing what they built, and asking questions long after the workshop ended.
Belma Ibrahimović (opens in a new tab)
Head of AI saas.group -
Training
The interactive session helped us plow through a mountain of questions and tactical decisions. Our product-engineering organization now fully embraces agentic engineering patterns.
Nino Ulsamer (opens in a new tab)
Co-Founder & CTO StashAway
Rollout across teams
We've streamlined our processes across the entire Scrum cycle, from cutting epics through sprint planning to deployment. What used to be manual and inconsistent now runs in a far more structured way. The team spends less time on alignment loops and more on creating real value. My advice to peers: just do it. You'll see the value quickly, and the earlier you start, the sooner you benefit.
CTO
Pricing scales with your waves
The scope comes down to how many team trainings you need and which training each role takes.
At the end, your platform team owns the shared setup
How is this different from booking a single training?
A single training brings one team to a shared practice. At scale, every team takes the training for its role in planned waves, with mentoring alongside, until your platform team owns the shared setup.
ClientsRead the retailer case studyWhat do we need in place before we start?
A mandate from management, a real project for each team, and basic access to suitable AI tools. Every team brings its own code.
What if people push back?
We start with the team's real pain points and its engineering standards. We show the system working on their own work, keep people in control of intent and quality, and let evidence do the convincing instead of promises.
How do you protect quality as AI use grows?
Every training covers quality gates, explicit context, architecture boundaries, independent review loops, tests and handoff to people. Agents get more autonomy only where the feedback system can support it.
Do you measure individual engineers?
No. We measure per team only. The trainings follow your approved tools, deployment model, data classification and security requirements.
AboutProcurement and legalWhat happens when the project ends?
The practices, configurations, documentation and measurement model stay with your organization. The work is complete when your platform team owns the shared setup.
Train your whole organization, one team at a time
Start with a pilot team, plan the waves with us, and hand the practice to your platform team.
Elroy drafts the email with you in Claude or ChatGPT, and only you can send it.
Book a call (opens in a new tab)