Vertical AI as VC Operating Layer
Alex Johnson, co-founder & CEO of Velvet, on vertical AI for venture capital
This is really a statement about implementation burden, not model quality. A horizontal agent can look strong when a fund already has its own CRM, permissions, deal data, and workflows wired together, but most VC firms do not. In venture, the winning product is the one that arrives with the venture specific workflow already built, reads decks, data rooms, emails, and cap tables, writes memos, updates CRM, and helps with co investor and LP workflows without asking the customer to assemble the stack first.
-
Velvet is built around the actual deck to decision workflow. Its product pages center on deal snapshots, inbox agents, AI dataroom analysis, memo generation, network mapping, manager underwriting, and LP monitoring. That makes it closer to an out of the box operating layer for a fund, not just a search box over documents.
-
The contrast with Hebbia shows the difference. Hebbia is powerful for finance teams that already have structured internal systems and want deep querying across firm data, and it now plugs directly into PitchBook. That fits PE and large institutional buy side teams especially well, where workflows are already standardized and technical setup is easier to justify.
-
That also explains why horizontal tools mainly matter for the top end of venture. The interview frames only about 10% of firms, such as SignalFire and Insight, as likely builders. Everyone else is trying to avoid stitching together Glean, Hebbia, CRM, and bespoke prompts, because a small investment team wants analyst output, not an internal AI integration project.
The market is likely to split cleanly. Large data rich firms will keep buying horizontal building blocks and shaping them into proprietary systems, while the broader VC market will consolidate around vertical products that package venture workflows end to end. As more funds expect AI to update records, draft memos, and surface warm intros automatically, the default winner will be the product that behaves like a venture team member on day one.