Why Q1 agent roadmaps need explicit kill criteria
A roadmap that only lists launches is a wish list. Kill criteria turn Q1 agent bets into decisions you can defend in March when budget and patience run thin.
This week's headlines roundup opened with CIOs publishing Q1 agent roadmaps with explicit kill criteria. That phrase sounds harsh. It is kinder than letting three half-working bots limp through the year because nobody wants to be the person who "killed AI."
December already asked leaders to set retire-by dates on pilots. January is when you operationalize that discipline on the plan everyone will actually fund.
What belongs on each roadmap line item
- Owner with authority to shut the workflow down.
- Success metric tied to time, money, or risk (not message volume).
- Budget cap for tokens, seats, and contractor hours.
- Kill date or review gate (for example, end of Q1).
- Kill criteria written as observable thresholds, not vibes.
Example kill rule: "If cost per resolved ticket does not beat the human baseline by 15% after eight weeks of stable traffic, decommission and document lessons." That is a decision, not a mood.
Graduate, pivot, or kill
Every Q1 bet should end in one of three outcomes. Graduate to production with guardrails and a line in the budget. Pivot to a narrower scope with a new eval baseline. Kill and free the team for the next bet. "Keep evaluating" is not a fourth option unless you attach a new date and metric.
Tie graduation to eval gates and continuous evals so promotion is evidence-based, not political.
How to keep kill criteria honest
Write criteria before the pilot goes live, when optimism is high but still specific. Review them in the same meeting where you review agent spend. If the metric is missing because instrumentation was skipped, that is a kill signal on its own.
Escalate to leadership when an owner wants a third extension. Extensions without new criteria are how sprawl becomes permanent.
Template you can paste into a steering deck
- Workflow name and business sponsor.
- Baseline metric from pre-agent operations.
- Target by kill date (with tolerance band).
- Monthly spend cap and current burn.
- Decision on [date]: graduate / pivot / kill.
One page per agent bet. If it does not fit on one page, the scope is probably too big for Q1.
Connect roadmaps to the rest of the stack
Q1 plans should reference security work already on the calendar: RFP security appendices, backup drills, and refreshed use policies. A roadmap that ignores compliance and ops is a marketing slide, not an operating plan.
Want help turning agent bets into roadmaps your team can execute? Talk with Swift Media about agents, governance, and shipping discipline.
Matt Potter · Swift Media