Two days: from the first “what even is an AI pilot” to a reasoned go/no-go and a prototype that proves the AI can handle it. The Oper8 method on your own processes, run by its authors in person.
Led by the authors of the method — Sergei Lipchanskii and Askhat Urazbaev
Most companies lay AI over the old process.
Oper8 does the opposite: it redesigns the process around AI.
Deploy is AI bolted onto the old process (+10–20% speed). Reshape rewrites the process around AI: roles disappear, the cycle shortens several times over. Invent is a product that could not have existed without AI. Sixty per cent of companies get stuck at Deploy and believe that is what AI transformation means.
The AI learns from customers (outer loop) and from post-mortems of its own errors (inner loop), with the Decision Log at the centre. Without this, a pilot never reaches production.
Economics, feasibility, metrics and counter-metrics, safe limits. You leave with a process choice and a defensible go/no-go decision, rather than a “let's try it”.
One process, five levels of AI autonomy: from suggesting to a human (A1) to running unsupervised (A5). The level is chosen per step, not once for the whole process. We build the autonomy map on day one.
That is the method. The workshop is its applied format — on your own processes, in two days.
Eight steps, eight artefacts. Day one is process discovery, the economics of a pilot, and a defensible go/no-go. Day two is the target architecture, a prototype inside our agent harness, and two shadow-run iterations. The in-house format runs either as two intensive days or as six two-hour sessions at your own pace — same programme either way.
Before the workshop
In-house
A 60-minute call with the sponsor one or two weeks before day one. We fix the process, gather 10–15 real cases from your own practice, and agree access to data and systems. Participants need no preparation — the method is taught at the workshop itself.
Open enrolment
A week ahead we send instructions: read the Oper8 book and sketch five to seven steps of your own process — optional, but useful. No need to install anything; the technical side sits with the facilitator.
On data: the work happens outside your perimeter. Sensitive and personal data is not needed — cases are anonymised beforehand, and the accounts are not integrated with anything.
01
Day one
Why AI pilots so often fail to pay back. Deploy, Reshape, Invent — the three modes of change. How a pilot lives after launch. Not a lecture — worked through with examples from your own industry.
→ What remainsA shared vocabulary: the five shifts, Deploy → Reshape → Invent, the Data Flywheel, the autonomy spectrum.
We take a process apart on a familiar case — employee onboarding, document handling, lead qualification. The odds are you have something similar, and by the end of the workshop we map the method onto your own process.
→ What remainsA teaching case chosen and mapped onto your process — the parallel between them is visible.
We investigate the process through three channels: what the logs show, what observation shows and what people say in interviews. Then we build an AI service blueprint, with maturity and AI autonomy tagged at every step, and mark where AI applies.
→ What remainsAn AI service blueprint of the process, plus an autonomy map tagged A1–A5 at every step.
A financial model of the pilot, in three blocks: leading metric, lagging metric, inference cost. A separate block converts gains in AI autonomy into money — customer touches plus labour saved. The model decides: go, change the step, or stop.
→ What remainsA financial model of the pilot, and a gate decision: go, change the step, or stop.
02
Day two
We design how the pilot has to live in production: UI, LLM layer, RAG over a vector store, Decision Log, the A1–A5 autonomy spectrum, metrics and counter-metrics. The architecture gives us the budget — GPU, tokens, salaries — and a sharper cost for the pilot. Without it, the economics are guesswork.
→ What remainsA target architecture — UI, LLM, RAG, Decision Log, A1–A5, metrics and counter-metrics — with a costed budget.
The prototype is built inside our own agent harness: a builder agent helps you write the prompt, assemble a starting Knowledge Base and escalation rules, and run the first cases. No programming required — your work is the rules, the knowledge and the logic, not the code.
→ What remainsA working prototype agent — prompt, starting Knowledge Base, escalation rules — run against the first cases.
The team works not as the expert but as the process owner: comparing the answers against real decisions, finding divergences in the Decision Log, correcting the rules and the Knowledge Base. By the evening, the team has completed two iterations and the quality is visibly climbing.
→ What remainsA Decision Log covering two shadow-run iterations, with the quality curve visible between them.
Behind the template sits an implementation radar: five axes, 23 areas — from observability and rollback plans to ownership and middle-management resistance. Your process profile — cost of error, reversibility, regulation — decides what comes first. Month one is what to do, which metrics, whom to involve; months two and three follow either the doing track or the exploring track.
→ What remainsA filled-in 90-day roadmap for every participant — the thing that closes the gap between the workshop and Monday.
The method isn’t delivered off slides — you work through it in conversation with the agents.
We built this harness for our own consulting work, then opened it up to workshop participants.
A scout helps pick the process, an analyst builds the AI service blueprint, an economist works out the ROI, an architect designs the flywheel, a builder assembles the prototype, and a sixth carries you to the roadmap. An orchestrator passes context between them, so a correction at any step is picked up by the rest.
It starts with onboarding: the agent learns who you are, which process you own, and whether you have the authority to change it. Everything after that is a set of questions about your process, your metrics and your economics.
At the end, one button assembles a deck that answers the method's key questions: the process, the bottleneck, the metrics, the economics, the architecture and the roadmap. It downloads as PowerPoint — a deck you can take to the budget holder when you need to defend the investment.
You can come back to your process after the workshop: refine the blueprint, redo the economics and extend the prototype. Nothing to install — just a web interface; we handle the technical side.
You can ask the agents anything about the method — they answer and carry you onward. Everyone tunes the tone and the depth of the answers to their own taste.
Formats
In-house runs for a single team inside your company, on one or several of your own processes: two intensive days, or six two-hour sessions at your own pace — same programme either way. Open enrolment is for individuals who come to live through the method on a teaching case and map it onto their own situation.
Dates to be confirmed
We’ll find you the next one
Ask about datesYour process, your team
Scheduled around you
Talk through a workshopSergei Lipchanskii and Askhat Urazbaev. Sergei co-founded the AI consulting practice in 2023. Askhat founded ScrumTrek in 2008.
Two layers of experience behind the method
18 years of product and Agile transformation — including Coca-Cola HBC, Haleon and Bizerba
3 years of AI projects — the pilots and rollouts the Oper8 method grew out of
The workshop is the method with your hands on it: what exactly to do when you take a process apart, how to work out the economics of a pilot, which safe limits you need from the start. Two days is the floor — one day gets you to a go/no-go, but without the second day that decision has no architecture behind it. In our experience teams cover the same ground in four to six weeks of two-hour meetings. The workshop compresses it.
The book
One company followed the method from start to finish, taking it apart step by step. If the workshop is the method with your hands on it, the book is the method through your eyes: how the process changes, how people react and what the process produces.
Get in touch
Within one working day: a 30-minute call with one of the people who runs this.
The open method: the Data Flywheel, the five elements of the flywheel, the autonomy spectrum.
Read the methodIf the process isn't chosen yet — a one- to two-week audit presented to the leadership team.
See the formatsThe full cycle, from spec to a running pilot, with our team. 8–12 weeks.
See the approachWe’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
