TOPIC 06 OF 06
A 90-day adoption roadmap
Rolling agents out without a messFor a team or organisation, adopting agents well is a phased plan, not a switch you flip. Days 0–30: pick one low-stakes, high-friction process, name an accountable owner, and set the boundaries and gates before anyone connects anything. Days 30–60: run it read-only or draft-only, log what it does, and review the logs with the people affected. Days 60–90: widen scope only where the logs earned trust, write down what an agent may and may not do, and decide what evidence you'll keep. The theme throughout: start narrow, keep a human accountable, and let real evidence — not enthusiasm — widen the leash.
A support team adopts an agent for ticket triage. Month 1: it only labels and routes (reversible), owned by the support lead, never replies to customers. Month 2: it drafts replies for human send, logged and spot-checked. Month 3: after clean logs, it auto-sends only the most routine reply type, still gated for anything unusual.
The failure mode here is widening the leash on enthusiasm instead of evidence: month 3's auto-send gets switched on because the demo felt good, not because the logs were clean. Let each expansion of scope be earned by what the previous phase actually showed — and if a phase surfaces a problem, the roadmap runs backwards too, not just forwards. (Who owns it and what records you keep is the governance side of the same move — see the Governance Studio. This is a planning aid, not legal advice; high-risk uses carry duties, see the EU risk lens, that a roadmap doesn't discharge.)
Name one process to pilot, its accountable owner, and the one thing the agent may NOT do in month one. That's your day-zero plan.