Six months ago, I started running every data room through an AI analysis pipeline before I opened a single document myself. The results have been startling — not because the AI catches things I miss, but because it compresses the first 80% of due diligence into about 15 minutes.
The tools have gotten genuinely good. Claude, GPT-4, and Gemini can now parse financial models, flag inconsistencies in cap tables, summarize customer contracts, and surface red flags in legal docs. But the investors who are using these tools best aren't automating their judgment — they're buying back time to go deeper on the 20% that actually matters.
The New Due Diligence Stack
Here's what the AI-augmented due diligence workflow looks like for the most active angels in Inner Ping's network:
- 1.Data room intake: Upload all docs to an AI workspace. Get a structured summary of financials, legal, and product data in under 20 minutes.
- 2.Red flag scan: Run a specific prompt set that checks for common issues — revenue recognition problems, unusual vesting schedules, concentration risk, missing IP assignments.
- 3.Competitive landscape: Use AI to pull and synthesize recent funding announcements, product launches, and hiring patterns from competitors.
- 4.Customer reference prep: Generate targeted questions based on the company's specific metrics and claims.
- 5.Deep dive: Spend your human hours on founder calls, customer references, and the strategic questions that AI can't answer.
What AI Still Can't Do
The limitations are important. AI can't assess founder quality. It can't read the room during a reference call. It can't tell you whether a market is about to shift based on a conversation with an industry insider. And it's terrible at evaluating anything that requires understanding of local market dynamics or cultural context.
“I use AI to get to 'no' faster. The things that make me say 'yes' are still entirely human — the founder's clarity of thought, the way customers talk about the product, the gut feeling after the third call.”
— Inner Ping angel, 40+ portfolio companies
The Risk of AI-Accelerated FOMO
One unintended consequence: AI-powered due diligence is making rounds move faster. When every investor can analyze a data room in hours instead of weeks, the pressure to make decisions quickly increases. Several Inner Ping members have reported that competitive rounds now close in days, not weeks.
Faster due diligence doesn't mean faster decisions should follow. The speed gain should be reinvested in deeper human evaluation, not compressed timelines. Rushed checks written on AI-powered confidence are still rushed checks.
The investors who will win in this new environment are the ones who use AI to be more thorough, not just faster. The tool is a microscope, not an autopilot.
The Prompt Stack: What Top Angels Actually Run
After interviewing 18 Inner Ping angels who use AI in their diligence process, we've distilled the highest-value prompt patterns into a repeatable workflow. These aren't generic 'summarize this document' prompts — they're targeted extractions that surface the signals experienced investors look for:
- 1.Revenue quality check: 'Analyze this P&L. Flag any revenue that appears non-recurring, one-time, or potentially misclassified. Calculate the implied net revenue retention from the cohort data provided.'
- 2.Cap table red flags: 'Review this cap table. Identify any unusual vesting schedules, disproportionate advisor equity, or dead equity from departed founders exceeding 5%.'
- 3.Customer concentration: 'From these financial statements, calculate what percentage of revenue comes from the top 3 customers. Flag if any single customer represents more than 20% of ARR.'
- 4.Competitive positioning: 'Based on this pitch deck and these three competitor websites, identify the specific claims this company makes that are directly contradicted by competitor positioning or public data.'
- 5.Legal liability scan: 'Review these terms of service and privacy policy. Flag any clauses that could create regulatory risk given the company operates in [sector/geography].'
The Accuracy Question: Where AI Diligence Fails
We ran a controlled test: 10 experienced angels reviewed the same data room manually, while an AI pipeline analyzed it in parallel. The AI caught 94% of the quantitative red flags the humans found. But it missed 100% of the qualitative signals — the founder's explanation for a revenue dip that didn't quite add up, the customer reference who was suspiciously enthusiastic, the competitive threat buried in a footnote that only someone with deep domain knowledge would recognize.
More concerning: the AI generated 3 false positives per data room on average — flagging issues that weren't actually problems. For a new investor who lacks the experience to dismiss false positives, AI-powered diligence can create more confusion than clarity.
Use AI for the first pass (quantitative extraction, document summarization, competitive scanning) and human judgment for the second pass (founder assessment, reference calls, strategic thesis evaluation). Never let AI be the last step in your process — always have a human review the AI's output before making a decision.
The Emerging Ethics of AI-Powered Investing
A less-discussed dimension: founders are starting to ask whether investors used AI to evaluate their company, and some are uncomfortable with it. Concerns range from data privacy (is my data room being fed into a model?) to fairness (does AI evaluation disadvantage non-traditional founders whose metrics don't fit standard patterns?). Three Inner Ping members have already added AI usage disclosures to their diligence process — a practice that may become standard within 12 months.
“I tell every founder upfront: I use AI tools to help me process your data room faster, which means I can spend more of our time together on the conversations that actually matter. No one has objected yet — they'd rather I spend 45 minutes asking smart questions than 45 minutes reading their financial model in silence.”
— Naomi Sato
Naomi Sato
Naomi spent eight years building developer tools at Linear and Figma before turning to angel investing full-time. She's made 22 investments focused on AI infrastructure and developer productivity.