Use case
Agentic processes at work: the game development example
One game, twenty markets, new releases all the time. Each market has its own technical requirements, and the certified game mathematics must not be touched. Errors here cost not just money but market access. The five building blocks below show how agents speed up a studio's releases without touching what is certified.
Building block 1
RAG for market requirements
The challenge
Whether a release may enter a market depends on that market's technical requirements: game parameters, display obligations, stake and feature limits. And they change.
The approach
A RAG pipeline checks the game configuration and the release candidate against the stored requirements per jurisdiction, with source and date. The agent does not say "looks fine"; it shows which requirement it checked against.
Building block 2
Action-space whitelisting
The challenge
An agent that may do anything is a risk. An agent that may do nothing is useless.
The approach
The agent may prepare release packages, flag configuration deviations and produce test reports, but it never touches the certified game mathematics and never deploys to production itself. For example, asset and localisation changes run through the pipeline automatically, and anything that touches certified mathematics or regulated parameters goes to the human-in-the-loop gate.
Building block 3
Hash-chained audit trails
The challenge
Test labs and regulators accept no release decision that cannot be traced.
The approach
Every step is logged immutably: which build was tested, against which version of the requirements, who approved. The entries are hash-chained, so later changes are detectable. For certification or a query, the path of every release is fully documented.
Building block 4
Quality against golden sets
The challenge
No pipeline rework and no model change may shift test or simulation results unnoticed.
The approach
Fixed reference cases: known configurations with a known correct test result. Every change to models, prompts or the pipeline is measured against them before it goes live. Test quality never drops unnoticed.
Building block 5
Human control in the right place
The challenge
Studio and compliance teams fear two things: losing speed to control, and losing control to speed.
The approach
The system works on its own but stops at precisely defined gates: a new market, a change to regulated parameters, a new platform integration. There the team decides. The person corrects, and the system learns from it in a structured way. Assistance instead of loss of control: autonomy grows with proven quality, not the other way round.
Transferable
Game development is the example here. The building blocks are the same as in accounting and the tax firm: wherever errors cost money or market access. They are code, not concepts:
- local-agent-pipeline ↗ (whitelist, gates, hash-chained audit trail)
- rag-approval-blueprint ↗ (approving a RAG system with a control catalogue)