AI.vc?
I had dinner last night with a few other VCs. We got talking - as one tends to quite quickly these days - about AI. We covered the usual: hype, overvaluation, negative gross margins, differentiation, etc. I think we all felt very smart - and also quite dumb/bewildered at the same time.
I then pitched an idea I’ve been mulling - or at least, a question I’ve been asking myself: what would a truly AI native VC look like?
To caveat, I think it’s unlikely that AI could do every aspect of the VC job because the human-to-human connection and relationship is an important component. Ask a top tier founder to only speak to an AI during a deal process, and quite quickly I think you’d find yourself with fairly sparse deal flow.
That said, I am interested in how far you could push AI across all aspects of the investment “job”. To illustrate, if I break down the five key stages of VC as I see them, I’m pretty sure AI could play a major role across all of them.
Sourcing
Separate agents could take different roles a VC team would e.g. thesis development, thematic research, searching the corners of the Internet for emergent themes, scouring Product Hunt and Reddit forums, and so on.
Integrate more structured data - from Sensor Tower, Similar Web, et al - and set the agents wild to find interesting niches that are less covered by the rest of the market.
Once found, agents could do customised outreach to founders to set up initial intro calls, prepare pre-call agendas and briefs (for both founder and VC) so the humans on the call take up as little of their time as possible, answer the “basic” questions beforehand (VC to founder: fundraise history, founding team, etc; founder to VC: cheque size, lead Y/N, etc.).
Picking
Next, an AI agent could sit on/in intro calls/meetings alongside the human VC - listening and analysing verbal and non-verbal cues to help assess the founder dynamics. As well as taking notes, the AI could suggest questions for the human VC to pick up - or not - in real time or as a follow-up for further DD.
This could get particularly interesting when speaking to co-founders to assess their compatibility and working styles.
In my discussion with the other VCs over dinner, I had a lot of pushback here: “It’s gut feel. It’s subjective. You can’t make it formulaic.” Do I agree with that? I’m not sure. I think gut feel is ultimately developed through exposure to data and knowledge of outcomes - which can be developed by a human over the course of a long career, or embedded in training of an AI. I think the “spidey sense” a lot of VCs think they have is actually just an inability to articulate the detail of the criteria on which they are assessing founders and teams.
Winning
This is probably the most human-heavy part of the job. Partnering with a VC is a 10-year journey, and you want to know the human behind the decision is with you for the long haul.
Still, I think AI could help prepare the “case” for a VC - why us? How can we help? Who is in our network? What do other founders think of us? All of this takes manual effort that could be automated into a single click of a button, leaving the human time to focus on relationship building.
Deal Execution
I dream of a world without lawyers… (just kidding, sort of). But seriously, so much time and money is wasted in back-and-forth emails, procedural requests, information documentation, and so on. Not far off all of this could be automated, with a very small handful of minutes spent in human-to-human contact (both commercial and legal) to tackle big points or edge cases.
Let’s move away from the ludicrousness of £20, 30, 50k of capital finding its way away from the productive use in the hands of a founder and into the pockets of law firms.
Portfolio Management
Similar to the “Winning” stage, nobody wants an AI on their board (or do they? Separate post to come), so AI’s role here is probably mostly in the mechanics: meeting prep, call notes, proactive support, automated follow-ups.
On the reporting side, all calls, decks, email updates, metrics dashboards etc. could be ingested into a centralised system, with LP updates auto generated each quarter (or even on demand).
All this would free up more time for VC partners to spend with founders, actually trying to do the mythical “value add” instead of wading through admin.
All of the above sounds overly simplistic and yet at the same time, quite far fetched. But I think it might be closer and more plausible than people think.
In that construct, the economics of being a smaller fund or even a “solo GP” start to stack up much better. Could an AI native play change the structure and lead to more decentralisation, less groupthink, less hype and ultimately better outcomes?
It’s an interesting thought experiment that I want to dig into further. I simply don’t believe that an industry that hasn’t really changed that much in fundamental structure can/should go on uninterrupted across the most significant platform shift since the Internet.
More to follow.