40 Questions Private Equity Analysts Should Ask Before Trusting AI Deal Screening
AI deal screening tools filter companies at machine speed, but they filter by pattern, not by the irregular characteristics that experienced PE investors recognise as hidden value. Your judgement remains your competitive edge only if you know exactly what questions to ask before you trust what the algorithm shows you.
These are suggestions. Use the ones that fit your situation.
1PitchBook flagged this company as a low match because it has a non-traditional management team. Did the AI weight founder-led versus professional CEO as a deal killer, or did I build that bias into the screening parameters myself?
2The AI screening tool excluded this deal based on revenue multiple thresholds. What is the actual revenue multiple compared to similar deals I have closed, and does the AI know about the recent customer concentration shift that makes this company actually cheaper?
3ChatGPT summarised this sector in three sentences. What did it leave out about the regulatory changes coming in six months that every operator in the space is watching?
4Preqin AI ranked this deal source low because they have produced few exits in my sector. How many deals have they sourced that I rejected in the past three years, and were those rejections based on price or on my lack of conviction?
5The AI scored this founder as lower risk because she has previous exit experience. What does the AI not know about how she handles operational pressure or works with co-founders?
6This company scored high on AI financial health metrics but low on my intuitive assessment during calls. What financial signals did the AI pick up that I missed, and what signals am I picking up that the AI cannot measure?
7The screening tool filtered out companies with revenue below ten million pounds. How many acquisition targets at five to eight million pounds in this space have I seen acquired by competitors in the past two years?
8Microsoft Copilot generated a sector summary for my investment thesis. Which parts came from consensus reports that all funds read, and which parts represent original observation of actual market behaviour?
9The AI sourcing recommendation is sound by its metrics. What would change my conviction about this deal if I spent three days talking to customers instead of relying on the recommendation?
10Kensho flagged that this company's growth rate is an outlier in its sector. Is it an outlier because it is genuinely better, or because it operates under a different business model that the AI treats as anomalous?
Due Diligence and Investment Memoranda
11ChatGPT produced a balanced risk analysis for this deal memo. Which risks did it flag because they appear in template memoranda across all deals, and which risks did I actually discover by talking to the management team?
12The AI due diligence tool highlighted three customer concentration risks. How severe is each one given what I learned in customer interviews, and did the AI miss the offsetting factor that the largest customer is contractually bound through 2027?
13PitchBook AI suggested comparable transactions. Do these comparables reflect the actual market price paid by informed buyers, or do they reflect what appears in public filings and headlines?
14The AI generated a valuation range for my memo. Which part of that range reflects my own view after working through unit economics, and which part is the algorithm interpolating from precedent deals?
15Copilot drafted market size analysis for my thesis. Did it source bottom-up data from industry associations and field work, or did it aggregate published market research reports that may be outdated?
16The AI flagged management experience gaps in the team. Which gaps are actually material to our investment thesis, and which ones am I worrying about because the AI highlighted them?
17ChatGPT produced a competitive landscape section that looks professional. Did it identify the three private competitors that matter most, or only the public companies and well-known private ones that dominate search results?
18The AI due diligence summary says the technology is proprietary and defensible. Did it actually assess the defensibility by understanding the underlying technical differentiation, or did it extract that language from the company's marketing materials?
19Kensho analysed the regulatory environment for this sector. Which upcoming regulatory changes did it flag from official sources, and which ones did it miss because they exist in trade publications and private analyst conversations?
20The AI memo recommends a three-year hold period based on comparable exit timelines. What is my actual plan for value creation over those three years, independent of what precedent deals showed?
Portfolio Monitoring and Decision Making
21The portfolio dashboard shows EBITDA margin compression at one portfolio company. Did the AI identify a real operational problem, or did it flag normal seasonal variation that I would catch if I looked at the monthly trend rather than the quarterly snapshot?
22Preqin AI aggregated metrics across all holdings and highlighted this company as underperforming peers. Against which peers did it compare, and are those the right comparables for my value creation plan?
23The AI monitoring system flagged cash burn at one company as concerning. What is the actual trajectory of the cash burn, and does it align with the growth investment plan the board approved six months ago?
24ChatGPT produced a market update suggesting headwinds in this portfolio company's sector. Which part of this analysis came from earnings calls I have already heard, and which part represents new market intelligence?
25The AI dashboard shows customer acquisition cost rising at this holding. What does the management team attribute this to, and does the underlying cause affect our value creation timeline?
26Microsoft Copilot suggested a strategic pivot based on sector trends. Is this suggestion based on what other companies in the space are doing, or on what our actual customer conversations are telling us about demand?
27The monitoring system flagged key person risk at this portfolio company. Have I assessed whether this risk actually affects our exit timeline and value creation plan, or am I treating the flag as a problem simply because the AI raised it?
28Kensho analysis shows competitor activity increasing in this space. Which competitors matter to our actual strategy, and which ones is the AI treating as relevant because they operate in the same SIC code?
29The AI recommended accelerating the exit timeline for one holding based on market multiples. What is the company's actual operational readiness for exit, independent of what public markets are paying?
30The portfolio dashboard displays standardised metrics across all twelve holdings. For which holdings does this standardisation obscure the specific value creation levers that actually matter?
Investment Conviction and Cognitive Independence
31When I read the AI analysis, can I identify the specific data point or customer insight that changed my mind, or do I feel convinced simply because the analysis was thorough?
32The AI tool produced a risk summary that looks balanced. Am I more confident in this deal because the analysis acknowledged both upside and downside, or am I confusing thoroughness with wisdom?
33My investment memo now includes sector analysis from ChatGPT. If I removed that section entirely, would my actual investment thesis change?
34Three analysts at my fund are all using the same AI tools for screening and due diligence. How would my investment perspective differ if I conducted the entire process without those tools?
35The AI screening tool recommended this deal as a strong fit. Before the recommendation appeared, did I believe this deal was worth pursuing, or did the AI recommendation create my conviction?
36I have reviewed this company using PitchBook AI, Preqin data, and ChatGPT analysis. Which part of my investment case rests on original observation, and which part comes from the AI summaries?
37The AI due diligence raised this specific risk. If the risk turned out to be irrelevant, would I have discovered that through my own questioning, or was I relying on the algorithm to identify what matters?
38My conviction level on this investment is high. How much of that conviction comes from the quality of the underlying business, and how much comes from the quality of the analysis the AI produced?
39The AI tool provided a balanced view of this deal that raised my confidence. What would restore my intellectual independence: spending more time challenging the analysis, or spending less time reading analysis and more time with the actual business?
40I rely on AI screening to eliminate poor deals and find good ones. If I removed the AI tool entirely, which deals would I source differently, and would my returns actually improve?
How to use these questions
Before trusting a PitchBook or Preqin AI screening score, spend one morning manually evaluating five deals the AI ranked low. You will discover what characteristics the algorithm treats as irrelevant but you recognise as valuable.
When ChatGPT or Copilot produces sector analysis for your memo, compare it against one recent investor call transcript from an operator in that space. You will see immediately where the AI analysis is generic consensus versus specific market intelligence.
Portfolio monitoring dashboards show you aggregated metrics. Once per month, call one portfolio company and ask the CFO what changed this quarter. Compare that conversation against what the AI dashboard flagged. You will train yourself to read behind the metrics.
After you write an investment memo, remove all the AI-generated sections and reread it. If your thesis still makes sense and still convinces you, the AI was useful. If the memo feels thinner, the AI was doing thinking you abdicated.
When multiple analysts at your fund reach the same conclusion about a deal, pause and ask whether that convergence reflects genuine market insight or whether everyone is starting from the same AI recommendation. Original dissent is valuable.