Top-Performing Reps Now Carry $750K to $2.25M in Quota — and AI Pipeline Gains Are Part of Why

Drawing on ICONIQ’s 2026 GTM benchmark survey of over 150 B2B and AI software companies, SaaStr reports what high-performing reps at fast-growing, well-funded startups actually carry: roughly $750K annual quota at SMB with ~90% expected attainment, $1.35M at Mid-Market with ~90% expected attainment, and $2.25M at Enterprise with roughly 85% expected attainment. These are explicitly not universal averages — they describe companies that have hit real product-market fit, not the broader market.

Part of what’s pushing quotas up is directly attributable to AI tooling: a 10-11 point improvement in lead-to-MQL conversion tied to AI-driven pipeline quality, plus a structural shift in what reps own — 65% of high performers now also carry cross-sell motions rather than pure new-logo acquisition. Compensation structure is moving with it: pay tied to Net New Recurring Revenue rose from 25% to 33% of total comp mix, rewarding expansion alongside acquisition.

The methodology question underneath these numbers is the one Notion’s Global Sales Compensation Partner Brian Le addressed a week earlier: how do you actually set a quota that a comp plan can support, rather than reverse-engineering a number from a revenue target? Le’s guidance is to combine top-down and bottoms-up approaches and backtest targets against real historical conversion rates — the same discipline that makes ICONIQ’s benchmark data usable rather than aspirational. He also flags that compensation plans need to shift as companies move from seat-based to usage-based pricing, which is exactly the pricing transition AI features are forcing across B2B software right now.

The practical read for sales leaders: these quota figures only mean something next to your own pipeline-generation capacity and comp structure. A $2.25M enterprise quota built on someone else’s AI-driven MQL lift isn’t a target you can import — it’s a target you can only hit if your own pipeline math, not SaaStr’s benchmark, actually supports it.