AI Citizen Builder Programme
Each of 12 non-technical people brings a real business problem from their own team and works on it for the full 10 weeks. They learn by doing, and finish with an agent their team uses.
Value stream mapped with the team, what good looks like agreed, a target set
Instructions, knowledge and tools
Eval set, permissions and a named owner
Measured against the target
Builders don't practise on made-up examples. Each one brings a problem from their own team, and every session moves it forward. The value starts during the programme, not after it.
Each builder maps their team's work as a value stream, picks the problem worth solving, agrees what good looks like and sets a target to measure.
Every skill goes straight into their own agent: instructions, knowledge, tools, testing and security.
The agent is in use, measured against the target, and shown to leaders at the show and tell.
~500 hours
a month, from one citizen builder
For the last two years I've worked with a FinTech firm, supporting their AI journey. There we helped one citizen builder free up roughly 500 hours a month. They were allowed to build with AI, and they had the safety guardrails to do it properly, and they built a value multiplier.
Their agent takes about 3 minutes off each second-line support ticket. The team handles around 10,000 a month.
Jack Ridge, Refactor
The programme is for people who know the work, not developers. Each builder uses a real problem from their own team as the material.
Understanding how work flows, and where to apply AI and where not to.
Prompts, instructions, skills and evaluations in Microsoft Copilot Studio, so builders make agents that perform well, not just agents.
Which connectors and tools to use, version control, and looking after an agent once it's live.
Only 39% of people using AI at work have been trained by their employer (Microsoft and LinkedIn, 2024).
Live, online, one session a week. Each builder starts with their own team's work and finishes with an agent their team uses.
Value stream map how your team's work really moves. Pick the business problem worth solving, agree what good looks like, set a target to measure, and shape a first agent idea.
A tour of Copilot Studio, then building an agent with clear instructions: its role, its tone, what it should refuse and when it hands off.
Grounding the agent on the right sources, knowing which data it can and can't use, and handling the main paths of a conversation.
Getting the agent to do real work in other systems, using approved connectors, agent flows and the right level of access.
Writing an eval set, reading the results, and keeping the agent lean so it's affordable to run every day.
Permissions, versioning, publishing to Teams or Microsoft 365 Copilot, and who looks after the agent once it's live.
Builders show what they built and what it changed for their team.
Something not covered here? Book a call and ask.
People who know the work in their team but aren't developers, such as operations, finance, HR or customer service staff. You nominate 12 people for each cohort. Each person works to solve a real business problem that exists within their team.
No. The programme is designed for business staff, not developers. Copilot Studio is taught from scratch, starting in week 5.
An agent built on a real problem from their own team, taken from idea to working solution, with an eval set showing how well it works and a named owner to look after it. Their team uses what they build, so the value multiplies.
Workshops take around two hours a week. From week 5, we advise doubling that to allow time for building.
Copilot Studio licences, your internal guidance on connectors and tool policy for us to reference and follow, and a provisioned route to production across your environments, for example dev, test and prod.
Live and online, one session a week, in Microsoft Copilot Studio inside your own Microsoft 365 tenant.
£20,000 + VAT for a cohort of 12, and £1,000 + VAT for each extra builder.
Yes, you can run as many cohorts as you wish.
Tell us about your team and we'll tell you whether it's a fit.