A well-run agent can still be the wrong worker to deploy first. Agent strategy is the portfolio decision above the runbook: selection, sequence, aggregate exposure, and exit.
What is an AI agent? owns the worker. Using AI agents effectively owns one worker’s charter, tools, ceiling, and log. This page owns which workers enter the portfolio and in what order.
Four portfolio choices
This is an editorial decision card, not a validated scoring instrument.
- Selection. Which recurring jobs justify a stateful, tool-using worker, and which remain scripts or human judgment?
- Sequence. Which dependency must work before the next worker is added?
- Aggregate exposure. Across all workers, what can send, spend, delete, or speak as the company? Several small ceilings can still create one large exposure.
- Exit. What evidence pauses, narrows, or retires a worker?
Missing an exit rule, an experiment becomes permanent infrastructure.
Collision control
Not a model pick. Models change. Portfolio choices survive the swap.
Not a single-worker runbook. The charter, tool allowlist, step ceiling, and trace stay on the operating page.
Not an orchestration graph. Why multi-agent systems fail owns coordination after several workers exist.
Not SENSE restaged. SENSE is the strategic-intelligence layer. This card decides which agent jobs enter the operating portfolio.
What this page is not
It is not a vendor shortlist, charter template, or runtime bake-off.
AI Agents for Startup Strategy is live. Name the first job, its dependency, the portfolio exposure, and the exit rule. If the only filled field is a model name, you do not have an agent strategy.