The Agentic Contract

Enterprises are no longer blocked by whether agents work in pilots. They’re blocked by whether agents can be owned, reviewed, monitored, and amended in production.

This whitepaper introduces the agentic contract — the shared operating model for Heads of AI, platform leads, and GRC reviewers shipping agents quickly without losing control over risk, quality, or cost.

• Why ticketing, GRC, IAM, and observability stacks silently fail past a few dozen agents
• Contract anatomy: scope, decision authority, ownership, risk envelope, quality bar, cost ceiling
• The six-stage lifecycle — and who has to show up at each gate
• Risk classification per run, quality binding across four signals, cost ceilings with defined response
• Where Prefactor fits: the runtime that keeps ~85% of runs silent and pages the named owner on the rest