Why AI Should Start in Origination, Not Diligence
Most AI for private equity chases the deal team. I think that’s backwards. Origination has the most repetition and the most data, and it decides which deals you ever see.

Key takeaways
- Most AI-for-PE tools target the deal team. Origination is the better starting point: more repetition, more data, and it decides which deals you see.
- If you bet the models keep getting better, modeling and diligence are the first things they'll do end to end.
- What stays human is judging management teams, making decisions and owning relationships.
- Origination runs in three lanes: bankers, brokers and proprietary outreach. Each needs a different approach.
- The goal isn't more deals. It's seeing more of the market while keeping your rigor.
If the models keep improving, the technically complex work like modeling and deep research gets automated first, the same way coding did. What stays human is judgment and relationships, which is where origination lives. Origination also has the most repetition and the most data. Start there.
The bet I’m making about models
My starting assumption is simple: the models keep getting better. I’d bet on it. If that’s true, then the work that’s hardest for a person and most structured for a machine gets automated first. Building a model from a CIM, running deep research on a market, pulling comps. We’ve already watched this happen in software engineering, and deal work is next.
That’s why I think most AI in private equity is aimed at the wrong place. Vendors chase the deal team with diligence and modeling tools, which is exactly where the frontier models will be strongest on their own. It’s hard to build a lasting edge in the place everyone, including the labs, is racing toward.
What stays human
When the modeling is automated, the deal team’s job moves toward what doesn’t automate well: judging a management team, deciding what you actually believe, and owning the relationships that bring you the deal in the first place.
Origination sits right on top of that. It’s relationship-heavy, timing is everything, and it decides your opportunity set. You can’t win a deal you never saw.
Why origination is the right first move
Origination is where the repetition is. The same motions happen every week: reading teasers, logging deals, following up with bankers, building target lists, writing outreach. It has the most contacts and the most data of any part of the firm. Repetition plus data is exactly what agents are good at.
It also compounds. When origination runs well, the firm gets visibility across its whole market: who’s running which processes, which intermediaries send deals that fit, which owners have been warming up for two years. That visibility makes every later step smarter.
And it’s the most crowded private equity market there has ever been. Every edge in speed and in systematic coverage matters.
The three lanes of origination
Deal flow comes from three places, and each one wants a different kind of help.
Bankers. In the middle market, the large majority of deal flow is banked or intermediated. The game here is touch points: staying top of mind so you’re the first call when a mandate lands. AI helps with cadence, with remembering every conversation, and with signals like a banker changing firms or a deal closing in your sector. Most firms cover a hundred or more intermediaries and still can’t say which ones send deals that reach IOI.
Brokers. Brokers tend to send lists, and lots of them. This is volume, and the opportunity is screening: getting through the long tail fast without missing the one that fits.
Proprietary outreach. Direct outreach to founder-owned businesses is a sales motion with low hit rates. It’s often worth it, because there’s no auction. The work is building the target universe, finding the owner, checking your own history with them, and writing something personal. A lot of that is grunt work an agent can do, while the message stays human.
More deals is the wrong goal
I push back when people say AI should give you ten times the deal flow. The number of good companies that can trade in your market is capped. What you want is to see more of the market while keeping your rigor. You’re expanding your universe of luck, not your inbox.
So the metrics I care about aren’t volume. They’re the share of deals that reach IOI and LOI, how fast you get to a real answer, and ultimately returns.
Will AI replace associates?
I don’t think so, but the job changes. Less time reading the fortieth teaser of the week, more time on the five that matter. Less time typing call notes into the CRM, more time on the call. Associates who learn to run agents will cover far more ground than associates who don’t.
Frequently asked questions
Where should a private equity firm start with AI?
Origination: logging and screening inbound deals, tracking banker coverage and running proprietary outreach. It has the most repetition and data, and it decides which deals you see.
Can AI help with banker coverage in private equity?
Yes. Agents can track every touch point, rank intermediaries by the deals they send that fit, flag relationships going cold and catch moves and promotions.
Does AI replace private equity associates?
It changes their work. Agents take on repetitive origination and data work, so associates spend more time on judgment, relationships and the deals that matter.
Should AI increase the number of deals a firm sees?
Seeing more of the market helps, but the better goal is quality: more deals reaching IOI and LOI, faster decisions, and better returns.