Our offerings
Getting started with AI-assisted engineering
A short session, in person, by invitation
How to tell sound agent output from the kind that passes review and should not have, and which module would take your team further.
- In person
- By invitation
- Nothing to install
Modular training for your team
Three one-day modules, each sold on its own
Foundations, Verification, and Platform. Each is a full day with its own job, so a team can buy the one where it needs the step up. Verification and Platform end with work committed to your own repository, such as guardrails and a verification standard.
- One day per module
- On site, or remote for Foundations
- Your code for Verification and Platform
Governed AI Agent Platform pilot
Five weeks, with two of your own workflows
Once the team is trained, we run two of your agent workflows on your own non-production Kubernetes, with restricted access and a signed record of every task, and measure review effort and cost before you decide on production.
- 5 weeks, after a 2-day assessment
- Two of your own workflows
- A decision on production
The free session in detail
Getting started with AI-assisted engineering
Where we got to, what the tools are, and what good looks like.
- Format
- In person, by invitation
- Level
- Beginner to intermediate
- Your code
- No
- Price
- Free
Most teams are now using coding agents, but have no agreed way of working with them. Closing that gap is the purpose of this session. We will cover what the tools have turned into, and then spend the bulk of the time on how to judge what an agent hands back to you. When most people think of generative AI failures, they still picture a hallucinated API call, which a compiler can now catch in seconds. The costly failures are now the ones which pass review, and are written with no more hesitation than correct implementations. For cloud native teams, these failures land in YAML and templates, where a mistake will show up as silence rather than an error. A talk-through with worked examples on screen, not a lab, given in person to your team or at your event. No prep or prior experience is assumed.
You will be able to
Place the tool your team uses on a ladder that runs from a web chat window to a harness working unattended
Explain why an agent missed something you were sure it knew
Recognise the kind of failure a normal code review waves straight through
Take away a handful of checks you can apply to agent-written work the same afternoon
Setting it up
Nothing at all, not even an install
Ask for an invitation
We give the free session in person. Tell us who you are and we will invite you to the next one, or arrange one for your team or your event.
The modules in detail
| Module | Length | Where | Your code | Who it is for | Price |
|---|---|---|---|---|---|
Foundations AI Coding Agents for Engineering Teams | 1 day | On site / remote | No | Engineers using an agent most days | On request |
Verification Reviewing and Testing AI-Generated Code | 1 day | On site | Afternoon | A delivery team | On request |
Platform AI for Platform Engineering | 1 day | On site | Afternoon | Platform or infrastructure teams | On request |
The free session is the starting point, and assumes no experience with an agent at all. The modules pick up from there, and are built for teams already working with agents day to day. The vendors’ own free courses cover the basics of each tool, so the modules start from how your team already uses it. Foundations and Verification are the pair most engineering teams will want. For a delivered pilot on your own Kubernetes, see Governed AI Agent Platform; its included two-day readiness assessment is separate from this training.
AI Coding Agents for Engineering Teams
Context, cost, and control.
- Length
- 1 day
- Format
- On site / remote
- Hands-on
- 4 hrs
- Level
- Intermediate
- Your code
- No
- Price
- On request
Most engineers using an agent are using it as a chat window that happens to edit files. They cannot say what is in its context, they have no idea what a run costs, they use the same model for renaming a variable and for designing an interface, and when it does something stupid they conclude the tool is stupid. Every one of those is an operating error, and every one of them is fixable. This is also the only module that needs nothing from you: no repository, no cluster, no credentials, and no security review, so it can be booked without a procurement conversation.
You will be able to
Name everything your agent knows at any given moment, and spot what it is guessing
Work out why a run went wrong
Bring down agent usage costs by restructuring the work and matching each task to the right model
Say where your models are allowed to run, from a laptop to your own cloud account
Leave with an instruction file, a skill, and a hook to port to your own repository the next morning
Setting it up
Everything runs on our sandbox
Reviewing and Testing AI-Generated Code
Guardrails, the delivery loop, and wiring agents into systems that are not AI.
- Length
- 1 day
- Format
- On site
- Hands-on
- 5 hrs
- Level
- Intermediate
- Your code
- Afternoon
- Price
- On request
The bottleneck in AI-assisted engineering is trusting the output, rather than the workflows generating output. If a team cannot review agent output quickly, it either merges work nobody understands, or slows down to the point where the agent was pointless. Both of these outcomes are common. The day covers the security side of that too, in the language your security function already uses, so what you take back to them needs no translation.
You will be able to
Put limits around an agent that automatically hold
Write down what your team will never hand to an agent, and defend it
Check agent-produced work step by step, starting before the agent has written anything
Spot instructions that have erroneously reached your agent, and decide what is allowed to reach it
Give an agent permissions to systems it needs which do not have damaging over-reach
Move a change from backlog to merge-ready using safe and repeatable processes
Setting it up
We will work in the branch of a non-production repository
One system you already run, with read-only permissions
AI for Platform Engineering
Kubernetes, operators, and access control.
- Length
- 1 day
- Format
- On site
- Hands-on
- 4 hrs
- Level
- Advanced
- Your code
- Afternoon
- Price
- On request
This day is for the team that has agents working well on application code and has found that almost none of it carries over to the platform. The scaffolding for running AI which assists in managing platforms is materially different: correctness in infrastructure is proved by a cluster reconciling rather than by a test run. An agent holding cluster credentials has a blast radius that an agent editing a Go file does not.
You will be able to
Catch infrastructure mistakes which an agent can make silently before reaching a cluster
Let an agent investigate a live workload
Decide what the agent process itself runs inside, instead of trusting the harness to bound it
Judge if an infrastructure change is correct
Know what your teams spend on models, who spent it, and what data left the network
Set the terms before an agent touches a service owned by another team
Setting it up
We will work in a non-production cluster or a namespace
Once the team has agents
Most engineering teams now have coding agents and no agreed way of working with them. The vendors teach the tools for free. This training is about your team's own work, such as the guardrail that stops an agent before a mistake reaches review, or the check that catches a change which passes its tests and does nothing.
How we teach it
Verification and Platform start on our sandbox and spend the afternoon on your own repository and non-production systems, so the team leaves with work committed to it, such as guardrails, a written verification standard, and a gateway in front of your model provider. Foundations runs on our sandbox all day, so it needs no access and no security review.
Your engineers use their own agents throughout, as they would at work. Our engineers spend their time on judging what the agent hands back: the failures a normal review misses, and the trade-offs the agent cannot see.
We use coding agents on our own client delivery. We have been delivering Kubernetes since 2015, including the multi-cloud platform we designed and supported for Siemens, which grew to 200+ clusters across AWS, Azure, and VMware, and we have run Kubernetes training for engineering teams at companies including Maersk, JCDecaux, AJ Bell, Vonage, and Numatic.
Does this sound like your team?
Teams already given AI tooling
Individual wins and processes which never coalesce into a shared way of working
Teams that tried it and stopped
Worth understanding what went wrong before spending more on licences and tokens
Platform teams holding the line
Happy to review agent work in a repository, but not sure how agents may work in a cluster
Engineering leads who need a standard
A worked example the team can point at
How the modules run
Six to ten engineers per module, so everybody works at their own keyboard
Your own laptop and agent, with access to your model provider
Nothing runs against production
The modules are designed for teams already using an agent day to day. Start with the free session if you are not quite there yet
Talk to a specialist
Interested in running AI-Assisted Engineering with your team? Book a quick chat with one of our team to find out more.
Talk to a specialist