A fund's shape becomes a founder's problem on the calendar. At a firm built for a handful of concentrated bets a year with no associates [5], the first real conversation is with someone who can say yes and who is also carrying the entire file himself. That shortens the path to a decision and raises the bar for what gets a meeting. A smaller fund does not mean a friendlier process. It means fewer meetings, taken more seriously, with no staff layer to absorb a half-finished story.
Pande's own numbers show why concentration in this sector is difficult rather than fashionable. He puts the odds of a drug surviving from the first trial to the end of the third at 20%, with a single trial still costing hundreds of millions of dollars [8][9]. Invert the survival rate and about five programs have to enter trials for one to come out the far end, so the pooled trial spend behind a single success runs to roughly five times what one program costs [15]. A fund writing a few checks a year cannot average that down. The fund is underwriting the handoff: who pays for phase 3, and on what evidence. It is worth noting Pande's stated reason things fail there is not sloppy biology but animal models that do not predict humans [10].
The second change is what counts as a moat. TechCrunch frames the AI-biotech conundrum plainly: unlike text, biological data cannot be scraped off the internet, so nearly every company ends up building its own walled-off dataset [13]. That is awkward for anyone whose pitch is a model. If the data has to be generated, the capital plan is a wet-lab plan, and diligence becomes a question about instruments and sample access rather than benchmark scores. Pande's claim for AI is modest by design: not perfect, but better than any animal model, and interesting once it clears that bar [11]. He also declined the easy line about synthetic data thinning out trial enrollment, calling it an aspiration [12], while saying the cost and time to reach trials has been shrinking [14].
For your own raise, sort the funds on two axes: new checks per year, and who owns your file after the first meeting. Many checks with a staffed team is the machine most founders have already met, where process is fast and partner attention is rationed after the wire. Many checks with a thin team is where diligence stalls without anyone telling you. Few checks with a partner owning the file gives you a decision-maker and a much higher chance the answer is no. Few checks with a staffed team is rare, and it usually means the fund is later-stage than you are. The quadrant you are in should set what you send first and how long you wait before you stop waiting.
The open question is which parts of the day-to-day the AI is actually running. The source says the firm relies heavily on it for operations [5] and does not break out the tasks. Sourcing and memo drafting are one thing to a founder; portfolio reporting and follow-up ownership are another, and that is the difference between a reply in a day and a thread nobody has been assigned.
