SAN FRANCISCO — Startup accelerators face a structural problem: the math no longer works.
Programs typically receive an average of 487 applications per cohort, which a team of three to five members must evaluate within two to three weeks. That translates to roughly 21 applications per person per day—a pace that degrades human judgment.
Research on decision fatigue shows quality deteriorates significantly after extended evaluation periods. An application reviewed at hour eight of a reviewer's day may receive a different assessment than the same application reviewed at hour two, regardless of merit. Reviewers develop shortcuts, relying on pattern matching rather than comprehensive analysis.
The supply side of the problem is clear. Global venture investment reached $368 billion in 2024 across 35,684 deals—capital concentration that pushes more founders into accelerator pipelines earlier and more aggressively. Techstars says it personally reviews every application despite receiving thousands. Accelerating Asia Ventures filters over 700 startups for its cohorts.
The bottleneck is real. When human capacity cannot scale with application volume, accelerators either reject strong founders by chance or accept weaker ones by fatigue. Neither outcome is efficient capital allocation.
Some firms are exploring AI-assisted initial screening to move beyond subjective judgment. The premise is straightforward: use technology to reduce decision fatigue in early filtering, letting human evaluators focus on a curated pool where their judgment matters most.
