Comp plans tell people what matters. When AI changes the work, plans that stay the same either waste the gain or create unfairness.
What changes
- Capacity: if AI frees seller time, the same headcount can cover more — but only if territories, quotas or coverage change too.
- Attribution: who gets credit when an agent sources a meeting and a human closes the deal?
- Activity metrics: targets like "emails sent" or "calls made" become meaningless when machines can send unlimited volume.
- Role mix: new roles (agent supervisors, AI-enabled SDRs) need their own pay design.
A useful benchmark point
The Bridge Group's 2025 SDR report found a global median monthly quota of 10 meetings held, down 40% since 2018, and a median on-target earnings of $80K with a 68:32 base-to-variable split, unchanged since 2022. Expectations of what one human SDR delivers have been falling — before AI agents arrived at scale. Hybrid roles, where one person runs an agent, will need very different targets.
Principles for redesign
- Pay for outcomes, not activity: meetings held, qualified pipeline, closed revenue, retention.
- Write crediting rules before agents go live: e.g. the human who works an agent-sourced meeting gets credit; agent-sourced pipeline is tracked separately for ROI.
- Adjust quotas with evidence: change quotas when pilots show measured productivity changes, not on vendor claims. Explain the method.
- Pay supervisors for team output: an agent supervisor's variable pay should reflect quality and pipeline from the agent fleet, plus guardrail metrics.
- Protect against gaming: if AI makes an activity cheap, don't reward it.
Phasing
Run one plan year with AI metrics tracked but not paid on. Use the data to set the next plan's quotas and crediting. Communicate early; comp surprises cost more than any tool.