Fifty hours in a hybrid format, a working prototype on your own process and a plan to scale it — in ten to twelve weeks.
AI changes processes. We rebuild companies.
Every participant picks a real working process before the programme starts. Every artefact and prototype is about that process, not about a teaching case. Six months of support follow, so the prototype survives to production.
Who it is for
Product and project managers, business analysts, L&D, legal, finance, HR strategists, PMO leads, internal consultants. No programming skills are needed — half the participants arrive with no development experience and leave with a working prototype.
What you need first
Why the order is this way
Overview
What an AI champion is, the levels of autonomy, choosing the process.
Technical foundation
Five modules back to back: n8n → RAG → MCP → AI-assisted development → AI engineering. Hands on, on your own process.
The path of an initiative
Discovery → pilot → architecture → economics on real data → from pilot to production.
Strategy and exit
Portfolio, organisational structure, the Rulebook. The bridge into six months of support.
The principle
Hands first, discussion after.
Until a champion has built a prototype and run into the limits of AI with their own hands, any discussion of adoption is a theoretical exercise. That is why the five technical modules come before the business blocks, not after.
The participant chooses a real process before the programme starts. Every artefact is about that process, not a teaching case.
A prototype in n8n (or built with a coding agent) wired into the corporate Knowledge Base and working systems, plus a real feel for where AI falls over.
System logs, observation of the work, expert interviews. Then marking each step against the five levels of AI autonomy.
A three-gate funnel, and protection against four standard mistakes: deciding from a demo, chasing small quick wins, deferring to the most senior voice in the room, and copying someone else’s success.
A parallel run in shadow mode, exit criteria, and three versions of the ROI case — for the financial, the operational and the strategic listener.
A three-phase rollout plan, six levers for working with resistance in middle management, and the design of the new roles an AI transformation creates.
Async · before the startAn entry assessment against the 5D model of AI readiness, a pre-read, and the choice of the process you will work on.
An individual read across five dimensions: people, process, data, technology, culture. The participant picks a real working process from their own function — every artefact and prototype afterwards is about that process, not a teaching case. The pre-read on the AI landscape puts everyone at the same starting point.
OutputAn individual readiness profile, plus a hypothesis about where the process should be in twelve months.
4 hoursFive shifts in thinking, five principles of an AI-native business, and the levels of both depth of change and AI autonomy.
Deploy, Reshape and Invent — the three modes of change — and five levels of AI autonomy. We work through why organisations get stuck at “we rolled ChatGPT out to a department” and how not to repeat it. A champion is not an enthusiast with ideas; they are a process owner with a mandate.
4 hoursn8n as a no-code orchestrator for AI processes: flows, triggers, error handling, integrations.
The difference between calling a model and building a process around one. Without that line drawn, any discussion of adoption stays a theoretical exercise.
OutputA first working agent for every participant.
4 hoursHow to give AI access to corporate data without handing that data to the model.
Embeddings, vector stores, retrieving the relevant context. Learning to spot hallucinations, and to prepare data so they do not happen.
OutputA prototype wired into the corporate Knowledge Base.
4 hoursTool calling as the way to give an agent hands: API calls, databases, operations inside corporate systems.
MCP as the emerging standard for connecting AI to outside services. From an agent that answers to an agent that acts.
OutputA prototype that does not only answer but acts.
4 hoursHow coding agents work, and how a non-programmer can assemble working code.
Framing the task, checking what was generated, iterating with Claude Code and Cursor. Each participant assembles their own coding agent.
OutputA prototype with code components that no-code cannot produce.
4 hoursEight architectural patterns for agents, each with its failure modes. The key shift: Evals First.
Tests not as a readiness check but as the specification. We build the test cases before the system, and watch the harness grow out of them — the frame the agent lives in.
4 hoursThe first step is not “where can we fit AI in” but “where does it produce a measurable effect”.
Hybrid discovery through three channels — logs, observation, interviews — marking steps by autonomy level, prioritising through a three-gate funnel, and guarding against four cognitive traps.
OutputA process map marked up by autonomy, plus a shortlist of initiatives that cleared all three gates.
4 hoursFrom the shortlist we take one initiative and design the pilot.
The pilot is designed so that it produces either grounds to scale or a well-argued no — and the second is also a success.
OutputA pilot spec plus an n8n flow running in shadow mode.
4 hoursDesigning the agent that runs in the pilot. Gate 0 as protection against automating chaos.
Once the pilot plan exists, the agent gets designed. If Gate 0 — process readiness and protection against a data swamp — does not pass, that is not a failure; it is three months saved.
OutputAn architecture spec for the agent, plus Eval Dataset V1.
4 hoursThe key shift: ROI is calculated from the shadow run, not from forecasts.
A different position in front of the CFO: not “we think we will save” but “over four weeks we would have saved X, here are the logs.” Six metrics of an AI-native organisation, leading versus lagging, Goodhart’s law and counter-metrics. FinOps as a lever: routing, caching, batching.
OutputThe ROI case in three versions, plus a FinOps plan.
4 hoursScaling is riskier than the pilot: a far bigger budget, eighteen months, and much more at stake.
Three phases with criteria for moving between them: Enable (2–4 months) → Embed (3–6 months) → Evolve (continuous). The standard failure modes of each. Resistance in middle management as rational behaviour, plus six levers for thawing it. Six new roles an AI transformation creates: what grows internally, and how long it takes.
OutputAn Enable → Embed → Evolve plan, a plan for working with resistance, and a map of the roles.
4 hoursWe move up from a single initiative to the transformation as a whole. Not a lecture — a discussion with artefacts on the table.
A portfolio with named owners (the 70/20/10 rule; automation plateaus after about six months). Organisational structure in three phases: centralised → federated → embedded, with the turning point around eight to twelve processes. The Rulebook as a living document: red lines, guardrails, operating practice. The provocation: a full redesign costs an organisation 30–40% of its middle-management positions, and companies that do not plan for that transition get sabotage instead. ML versus GenAI as a business decision.
OutputA strategic map of the transformation, plus three concrete actions for the next 30 days.
After the programme · six months
Skills in AI tooling go stale in three to six months. After the main programme ends, champions stay in the community and keep getting updates.
The shape of the programme depends on how many champions take part and how much support follows it. We work that out on the call.
Not certificates — working prototypes and things running in production.
15 champions with working prototypes
Over ten weeks, 15 champions from marketing, finance and R&D defended prototypes built on their own real processes in front of the sponsor. Each brought a working prototype and a plan to scale it. One of them — a 70/30 split in customer support — went to production and produced a measurable effect.
Throughput per agent
+110%
Attrition
18% → 5%
Payback
1.5 months
Some engagements are under NDA, so they are not named here. The rest are on the case studies page.
The main format is a corporate programme built around your company. If you cannot fill a group right now, send people to an open cohort instead.
Dates to be confirmed
We’ll find you the next one
Ask about datesYour function, your team
Modules tailored to your industry
Talk through a programmeRequest
If you need the programme shaped around your function and context, tell us what the problem is. Within one working day, a 30-minute call: who is in the cohort, what the support looks like, and the dates.
Six training options for different roles: the workshop, AI Champions, executive training, AI Literacy, AI-DLC and Vibe Coding.
Compare training optionsCoverage at scale: from 50 to 40,000 people, role-based tracks, industry-specific programmes.
Read moreBefore you train champions, work out which processes to hand them.
Book a diagnosticWe’ll look at your problem, pick the right format, and show where AI actually strengthens your operations.
50+
companies have gone through AI transformation with us — from diagnostics to working solutions.
© 2026 Kiberry — the European practice of ScrumTrek
