Agents at each step
Specification, architecture, code, tests and releases, with agents at each step. Each agent has a defined task, defined inputs and a person who checks its output. We start where the gain is clearest and the risk is lowest.
02 · AI for IT
AI in how your teams build software.
For the teams that build and run your systems. We bring AI agents into each step of the software lifecycle and redesign the work around them. Your engineers review every decision that matters.
Who It Is For
For IT leaders whose teams use AI in some steps and not in others.
In our experience, most engineering teams in large companies already use AI assistants to write code. Far fewer use it in specification, testing, review or operations, or have rules for how far agents may go. This work is for the people who own delivery and the people who sign off on change.
- CIOs and heads of engineering with AI assistants in use and no measured result
- Teams carrying legacy code that few people can still change safely
- Risk and compliance owners responsible for change control
What We Do
Agents at each step, and the work redesigned around them.
The work redesigned
Adding an agent to an unchanged process rarely changes delivery time. We redesign who does what, where reviews happen and what evidence each step leaves. Your engineers keep the decisions that matter.
Legacy modernisation
Legacy code modernised and documented, so teams can change it safely. Agents read the existing code, write down what it does and propose migrations. Engineers review every change before it is merged.
Measured on delivery
Measured on delivery time and quality, with the teams who use it. We record the baseline before anything changes. If a step does not improve, we say so and change the approach.
AI SDLC
AI across the eight steps of software delivery.
AI SDLC means AI across the software development lifecycle, not only in writing code. We cut delivery into eight steps, because that is where the human approval changes. Each step gets its own level.
01Intake and prioritisation
How work arrives, is sorted and is routed.
02Specification and planning
From a need to clear rules, acceptance criteria and tasks.
03Design
Architecture, interfaces and user experience.
04Implementation
Code, tests and documentation.
05Testing and checks
Automated tests, security analysis and the evidence they leave.
06Review and acceptance
Who reviews a change, and who decides to accept it.
07Release and change control
Change approval, deployment and rollback.
08Operations and maintenance
Monitoring, incidents, fixes and upgrades.
Six levels, from manual to self-improving.
LevelNameWhat it looks likeWho decides
L0Manual
No AI in this step.
People
L1Assisted
People use approved AI assistants. No whole task goes to an agent.
People
L2Delegated
Agents carry out whole tasks. A person checks every output.
A person, every time
L3Orchestrated
Agents work in a connected pipeline with automated checks.
People, at set points
L4Governed autonomy
In categories your rules allow, agents complete the work and evidence accepts each change.
A named owner per category; people handle exceptions
L5Self-improving
The system measures its own results and proposes improvements to itself.
A person approves every change to the system
The top level is not software delivery without people. At every level, a named person owns each rule and each approval, and no agent approves its own work. A company is a profile of eight levels, not a single score.
How an Engagement Runs
Assess, choose one step, build, measure.
Assess
The software delivery assessment gives each of the eight steps a level, with the evidence behind it.
Choose the first step
With your teams, we pick one step and one target level, and write down what changes.
Build and run
Agents built into your tools and pipelines, on your servers or your cloud, used by one team first.
Measure and extend
Delivery time and quality compared with the baseline, then the next step, in order.
What You Hold at the End
Agents in your delivery, and the rules that govern them.
- Agents running in your delivery pipeline, with code you own
- A level for each of the eight steps, before and after
- The rules for each step: what agents may do, and who approves
- Legacy code documented, so teams can change it safely
- Your engineers trained to run and extend it
Built for Regulated Companies
Your rules decide how far agents go.
- Change control mapped to DORA: every change recorded, tested, approved by someone independent, and verified
- For critical systems, review and release stay at L3 unless your internal audit agrees otherwise
- EU AI Act and GDPR reviewed with your compliance team
- Agents never assign, monitor or evaluate individual employees
- On your servers or your cloud, with the models you choose
Ways to Start
Three ways to start. The same team in each.
2–4 weeksStart here
AI strategy diagnostic
Where AI creates value in your business, what it is worth and what to do first. For software teams, the software delivery assessment gives each step of delivery a level and a next step.
- Interviews across the business
- Every AI idea sized in hours and euros
- Risk review with compliance
- A ranked, written plan
You hold at the endA plan your executive committee can approve
First system
One AI system in production
One use from the plan, built into your tools and live with the people who use it.
- An AI product for a business team, or AI in your software delivery
- On your servers or your cloud
- Every answer traced to its source
- Measured before and after
You hold at the endA system people use every day
Ongoing
Embedded AI team
Strategy and engineering alongside your teams, one use after another.
- Senior people working with your teams
- The plan kept up to date
- New uses built in order
- Your teams trained to run them
You hold at the endAn AI programme that keeps shipping
Not sure where to start? Thirty minutes will tell.
Tell us where your AI work stands. You will leave with a clear view of what to do next.