Dealflowhub.ai

AI in Deal Sourcing

Gartner projected that more than 75% of early-stage investor reviews would be informed by AI and data analytics by 2025. What that changes about how founders…

Key findings

  • Gartner projected more than 75% of VC and early-stage investor executive reviews would be informed by AI and data analytics by 2025.
  • Industry surveys since put AI use in deal sourcing research above 80% of firms.
  • Structured, machine-readable company data increasingly determines which deals surface at all.
  • Sourcing is being automated faster than judgement — conviction still forms in conversation.

A change in the top of the funnel

Gartner projected that by 2025, more than 75% of venture capital and early-stage investor executive reviews would be informed by AI and data analytics. Subsequent industry surveys put adoption for deal-sourcing research above 80% of firms.

Adoption figures from vendor and industry sources should be read with some caution — definitions of "using AI" vary widely, and the incentives of those publishing the numbers are not neutral. The direction is nonetheless consistent across sources.

What actually changes for founders

The practical consequence is that the first pass over a company increasingly happens without a human reading anything the founder wrote. Structured signals — sector, stage, geography, traction data, team background — determine whether a company enters a shortlist at all.

This cuts both ways. It weakens the historic advantage of warm introductions, since a company with strong structured signals can surface without a network path. It also means an incomplete or inconsistent public footprint can remove a company from consideration before any judgement is applied.

The implication is unglamorous but real: keeping company data accurate, current and consistent across the places investors look is now part of fundraising rather than adjacent to it.

  • Sector, stage and geography must be unambiguous and consistent
  • Traction data should be current wherever it is published
  • Team background needs to be verifiable from public sources
  • Inconsistency between sources is itself a negative signal

The risk nobody has solved

A screening model trained on which companies previously raised will reproduce whatever patterns sat in that history — including the ones the industry says it wants to move away from. If past funding skewed toward particular backgrounds, universities or geographies, a model fitted to that record will treat those attributes as predictive.

This is not a hypothetical objection to automation; it is a known property of systems trained on historic outcomes. It deserves more scrutiny than it currently gets, and founders outside the typical profile have a concrete interest in how much weight any given firm places on automated ranking.

The counter-argument is that structured screening can also widen access, because a company with strong signals no longer needs a warm introduction to be seen. Both effects are real. Which dominates depends on how a given firm builds and supervises the system, which is rarely disclosed.

What has not changed

Automation has moved fastest where the work is highest-volume and lowest-judgement: sourcing, filtering, initial research. It has moved slowest where conviction is formed.

No investor we are aware of writes a cheque because a model ranked a company highly. The model decides who gets the meeting. The meeting still decides the investment — which means the effect of AI sourcing is to change who reaches the conversation, not what happens in it.

For founders the practical conclusion is narrow but useful: the work of being discoverable and the work of being fundable are now genuinely different tasks, and doing the first well no longer follows from doing the second well.

Sources

Photo: Unsplash.