Open method · Reference
Oper8 is a way to redesign a process with AI and keep improving it after the first pilot. It gives teams a structure for choosing the work, deciding what people and agents should do, and learning from the results.
tap a node to read more
Outer loop · the customer teaches
What the AI puts in front of the customer at the moment of decision. Not a single answer, but an action — with its reasoning and the alternatives left open.
Read the articleWhat the customer does after a recommendation: accepts it, corrects it or leaves it unused. The team needs to check what those responses mean before changing the system.
Read the articleAn event from a system such as CRM, telemetry or billing that helps show what happened after a decision.
Read the articleCore · Decision Log
The record of each decision and its outcome. This is where both loops meet: the outer loop captures what happened with the customer, and the inner loop uses it to improve the next decision.
Read the articleInner loop · the system learns
Checks that show whether a change is good enough to use. The team decides what to monitor and when to pause or roll back.
Read the articleSix business metrics for an AI process, each paired with a counter-metric. The counter-metric catches the local maximum — the number goes up while the outcome gets worse.
Read the articleThe rules, patterns and context the AI applies on the next turn. Not a static document — a snapshot of where the flywheel currently stands.
Read the articleRead the method
Where to start
Start with the decision in front of you: which process to choose, how to test a pilot, what an agent may do, or how to review the outcome. The articles below explain the parts of Oper8 in plain English.
The full book is still being prepared in English. The reference articles are available now.
Reference
Read short articles on the ideas and practices behind Oper8. Each one links to related topics, so you can follow the questions that matter to your team.
The ideas behind the method and the decisions they help you make.
Ways to choose, test and run an AI-supported process.
Common reasons pilots stall or fail to improve the work.
The fifth shift · Hypothesis
The method proposes a move from defending a copyable product interface to building service shaped by repeated decisions with customers. Memory records those decisions with observed outcomes; reviewed recurring patterns enter the third layer of the Knowledge Base.
Product-Service Blend is the service form offered now. The Data Flywheel is the operating feedback mechanism. Service as Moat is a possible strategic effect over time, if those learning loops stay active and their quality is reviewed.
A hypothesis from approximately 12 projects in 2024–2026.
Read the reference article →A two-day practical workshop. Prepare a pilot plan for your own task and practise building an agent using a sample case.
About the workshopA one-process pilot in 2–3 weeks, or implementation across one or two processes in 4–6 weeks.
See the approachTalk through your starting point and whether a workshop or diagnostic would help.
Ask for a callThe method was written by Sergei Lipchanskii (founder of Kiberry) and Askhat Urazbaev (founder of ScrumTrek). It draws on 18 years of product transformations and three years of AI projects — among them Haleon, Coca-Cola HBC, Bizerba, Chery and Astana Hub.
More about the authorsTell us what your team needs to learn, build or change. We’ll help you choose a training programme or plan implementation.
See the work
Read how a US bid firm chooses which bids to chase from its own track record, and how an executive search firm works with AI agents.
Read the cases
© 2026 Kiberry — the Cyprus-based practice of ScrumTrek
Operated by Skill Tech Consulting Limited, Hong Kong SAR · Registration no. 76164705
