Sourcing moved from network to signal
Funds used to see whatever their network sent them. Models now scan hiring data, product releases, repository activity, app rankings and public filings to surface companies before they announce a round.
For founders this is good news: a strong signal can reach a fund that has never met your investors.
Screening is now structured matching
The first filter is increasingly a mandate match — stage, sector, geography, cheque size and thesis language compared against your profile. Vague profiles lose here silently, because they never enter the shortlist.
Diligence gets faster, not softer
AI summarises data rooms, extracts contract terms, flags cohort anomalies and drafts reference-call questions. The result is fewer weeks spent reading and more time spent probing the two or three things that actually decide the investment.
Where humans still decide
Conviction about a founder, judgement about timing, and the willingness to be contrarian remain human. Models are strong at ranking the known and weak at pricing the unprecedented.
What founders should do about it
- Keep your public footprint current — models read it
- Make your profile machine-readable: explicit stage, sector, geography, ask
- Publish traction you are comfortable sharing; silence reads as no signal
- Use AI to rehearse the hard questions before the partner meeting