Kill the 800: A Framework for AI Deal Screening
Of 1,000 deals, the job is to kill the 800 dead on arrival fast, learn from them, and find the 20 worth your bid. Here’s the screening workflow, and the metrics to watch instead of volume.

Key takeaways
- Most inbound teasers are a quick pass, but someone still spends the time, and the data usually disappears.
- Kill the dead on arrival fast, keep everything you learned from them, and spend your attention on the few that fit.
- Screen against your own criteria, not a generic model, with the reasoning shown.
- Anything you do more than twice a week should be an agent, not a chat session.
- Track IOI and LOI rate and time to decision, not the number of deals.
The number of good companies in your market is capped, so "10x deal flow" is the wrong goal. Of 1,000 deals, kill the 800 dead on arrival fast, keep their data so you're smarter about the next 200, and find the 20 worth your bid. Then measure IOI and LOI rate, not volume.
Deal flow isn’t the problem. Deal processing is.
Talk to enough deal teams and you hear the same shape of week. Mid-market firms see anywhere from five to fifty teasers and CIMs a week, and a handful reach investment committee. A big share are a five-minute no. Another chunk needs a banker call or some quick research before anyone can decide. And reading a CIM and writing a one-pager can still take an hour and a half per deal.
One phrase I’ve heard more than once sums it up: we don’t have a deal-flow problem, we have a deal-processing problem.
The funnel: 1,000 → 200 → 20
Here’s how I think about it. Picture a thousand deals.
Kill the 800 fast. Most are dead on arrival. Wrong size, wrong sector, wrong ownership. The goal is to get them out of the way in minutes, not hours.
Keep the data from all of them. This is where most firms lose. They delete passes instead of logging them, so they can’t say which bankers send what, how deal volume moves by sector, or what they’ve already seen. The 800 you passed on are the raw material for being smarter about the next 200.
Find the 20 that matter. Among the remaining deals are the ones that fit your firm uniquely, where you have an edge in judgment or operations. Those deserve your time, and sometimes a higher bid.
The screening workflow
Here’s what the workflow looks like when an agent does the first pass:
- The teaser or CIM arrives. You forward the email, or the agent watches the deal inbox.
- It’s read and scored against your criteria. Your buy box, written down once: size, ownership, sector, margin profile, customer concentration and whatever else matters to you. Every score shows its reasoning and cites the document.
- Dead on arrivals are logged and closed. They aren’t deleted. They’re recorded with the source, so your banker scorecard and deal-volume reporting stay complete.
- Fits get a short memo. A one-page pass/fail check against your criteria, then a two- or three-page initial review in your template for the ones worth discussing, with comparable deals from your own history.
- The deal team gets it Monday morning. You read the conclusion, not 84 pages.
A person stays in the loop for every decision that matters. The agent does the reading.
Make it an agent, not a chat session
A lot of people screen with ChatGPT or Claude today: paste in the CIM, ask for a summary, tweak the prompt, try again. It’s fine for one deal. It doesn’t work for a firm.
My rule: if you do something more than twice a week, it should be an agent, not you in a chat window. Agree on the perfect output once, the ideal screening memo for your firm, then codify it so it runs the same way every time from a forwarded email while you do other work. A recurring task with a clear trigger is the ideal case for an agent.
Measure what matters
If the goal isn’t volume, what do you measure?
- IOI and LOI rate. The share of screened deals that become real bids.
- Time to decision. How fast a deal gets a clear yes or no.
- Coverage. Which intermediaries you hear from, and which send deals that fit.
- Returns. Ultimately, better selection should show up in fund performance.
Frequently asked questions
How do you screen CIMs faster in private equity?
Write your investment criteria down once, have an agent score every teaser and CIM against them with the reasoning shown, close the clear passes in minutes, and write a short memo only for deals that fit.
Should you log the deals you pass on?
Yes. Passes are the data that tells you which bankers send deals that fit, how deal volume moves over time, and what you've already seen.
Can AI write a private equity screening memo?
Yes, in your firm's template, grounded in the CIM and cited to it, with a person reviewing before anything goes to the deal team.
What metrics should a deal team track instead of deal volume?
IOI and LOI rate, time to decision, intermediary coverage, and ultimately returns.