TL;DR: AI automated document-heavy due diligence in 2024. PE firms process CIMs in 2 to 3 minutes instead of 40, cutting review times by 70% to 93%. Nearly two-thirds of private equity general partners run GenAI pilots. When every firm has the same AI analyzing the same documents, alpha stops coming from speed. Primary intelligence is what's left: customer conversations, expert interviews, channel checks AI won't replicate. The expert network industry generated $2.5 billion globally in 2024, up from $1.5 billion in 2020. The gap between what documents say and what people know is where returns get made.

What you need to know:

AI cuts due diligence document review by 70% to 93%, freeing up 14+ hours per analyst per month.

Expert network spending hit $2.5 billion in 2024 because documents don't tell you what customers think.

When everyone has the same AI processing the same documents, differentiation comes from proprietary primary intelligence.

The ROI on AI isn't the time saved. It's what you do with freed-up analyst time.

Where Alpha Comes From When AI Automates Documents

I've watched this shift from both sides. On the buyside at Goldman and Citadel, I saw how much time got burned on document work. Now, at Woozle, I see how firms respond once AI handles the grind.

Private equity firms process CIMs in 2 to 3 minutes instead of 40. AI-powered document review cuts manual effort by 70% to 80%. Analysts who spent 90% of their time crunching numbers now spend 90% on strategic judgment.

Speed is necessary. Speed is no longer differentiated. When document work becomes commodity infrastructure, competitive advantage shifts to intelligence AI won't access.

The pattern I'm seeing: firms automate CIM processing and contract review, analysts get 14 hours back per month, and they go right back to reading more documents faster. The documents got faster to read. They didn't get smarter.

Bottom line: Automating document work creates advantage only if you deploy freed-up time toward proprietary intelligence. Otherwise, you're making commodity research faster.

Why Document Automation Creates a Commoditization Problem

Nearly two-thirds of private equity general partners run GenAI pilots as of September 2024. Over 40% use it in business processes. The adoption curve is steep and fast.

The irony is brutal. All the freed-up analyst time still depends on the same commoditized information everyone else has.

When too many firms access the same datasets, the competitive edge erodes. Quant firms have similar data vendors, infrastructure, and machine learning models. Differentiation gets harder, not easier.

Renaissance Technologies capped its Medallion fund after discovering increased scale led to deteriorating returns. When everyone analyzes the same documents with the same AI, alpha doesn't multiply. It divides.

I've seen this pattern repeat across funds. Analyst time gets freed up by AI. They spend it reading more of the same public documents. The research stack gets faster. The information asymmetry doesn't improve.

Bottom line: Automating document work creates advantage only if you deploy freed-up time toward proprietary intelligence. Otherwise, you're making commodity research faster.

What AI Processes vs. What It Creates

Man Group's executives wrote about their AlphaGPT system: "AlphaGPT doesn't replace human judgment but amplifies it."

Even the most sophisticated AI tools at one of the world's largest quant funds still depend on humans for the insight moving conviction. The AI handles data processing. Humans provide strategic direction, market context, and final decision-making.

The line is simple. AI processes information. Primary research creates it.

Three Things Staying Fundamentally Human

Getting people to say what they think. High-stakes experts and operators don't talk the way documents read. Getting beyond PR-safe answers requires reading the room, adjusting tone, pushing gently on contradictions, and knowing when silence produces the real answer instead of the rehearsed one.

Interpreting messy, conflicting reality in context. Real operators disagree, misremember, contradict each other, and sometimes lie. Turning those conversations into an investment view means judging who to weight more, which anecdote is an outlier, and when a contradiction is the interesting signal.

1. Deciding what is investment-grade. The toughest part of primary research isn't asking or transcribing. It's deciding, with a straight face, "This is solid enough to change a position." You need domain expertise, ethical judgment, and a feel for how today's datapoints sit inside the broader market and cycle.

   I've run thousands of expert interviews. The skill isn't getting someone on the phone. The skill is getting them to stop giving you the talking points and start telling you what's broken, what's working, and where the real friction lives.

   Bottom line: AI automates pattern recognition on existing data. Primary research generates proprietary data through persuasion, interrogation, and judgment. The second is defensible. The first is not.

Why Expert Networks Generated $2.5 Billion in 2024

In 2024, expert network companies generated an estimated $2.5 billion in revenues globally, up from $1.5 billion in 2020. The investment community is the largest consumer of expert network services.

The reason is straightforward. After you've read every document AI processes, you still don't know what customers think, whether the product roadmap works, or if the competitive threat is real.

Hedge funds are building proprietary LLMs trained on financial data, earnings transcripts, and proprietary datasets. Here's the strategic question nobody's asking: if you train an LLM on documents everyone else accesses, you've automated commodity research at massive expense.

The only proprietary dataset nobody replicates is primary intelligence from customers, competitors, and channel partners who won't talk to the next firm.

I've watched firms spend millions building internal LLMs trained on public documents. The model gets faster. The edge doesn't appear. Because everyone downstream is analyzing the same information, and AI doesn't change the information itself.

Bottom line: Documents are necessary but not sufficient. Expert networks exist because the gap between what's written and that people know is where conviction gets built or broken.

Where the Real Edge Lives Now

Bridgewater launched a $2 billion fund in 2024 run by machine learning. CEO Nir Bar Dea said the strategy produces "a unique alpha uncorrelated to what our humans do."

The key word is uncorrelated. Machine learning on public documents generates returns different from, but not necessarily better than, human judgment informed by primary intelligence.

Both matter. Only one is defensible when competitors deploy the same AI infrastructure.

AI infrastructure is necessary but not sufficient. The firms winning with AI integrate it with decision-making processes still depending on knowing what the market doesn't.

Due diligence speed gains are real. AI tools reduce CIM processing times by up to 93%, and overall due diligence timelines by 2 to 3 weeks. Speed is valuable. Speed is not differentiated.

When everyone reads documents in minutes instead of days, competitive advantage shifts entirely to what you know beyond the documents.

I've seen this play out repeatedly. Two firms analyze the same target with the same AI tools. One wins because they talked to 12 customers and three competitors before the bid. The other loses because they read faster but knew less.

Bottom line: AI creates speed. Primary research creates information asymmetry. The second is where alpha lives when the first becomes infrastructure.

What This Means for Investors

AI will eat anything resembling document review, pattern recognition, or mechanical analysis of existing data. It won't replace the human work of persuading the right people to share what they know, interrogating responses in real time, and exercising judgment about what is true, what matters, and what should move capital.

Leading private equity teams are making this shift. Bain found top firms are developing scorecard-based protocols to assess generative AI threats and opportunities in every diligence. The aim is to make it as routine as legal or commercial diligence.

Here's the blind spot. Firms use AI to evaluate AI readiness in targets, and they're still relying on documents to do it.

The investors who win aren't the ones with the best AI. They're the ones using AI to free up analysts to do the primary research AI won't replicate.

You automate the document work. You don't automate the conversation changing how you think about a position. The gap is where alpha lives now.

I've built Woozle around this premise: the ROI on AI-powered due diligence isn't the time saved. It's what you do with the time once desktop work is automated. We sell finished intelligence, not access, because the opportunity cost of analyst time is the highest cost in the research stack.

Bottom line: The ROI on AI-powered due diligence isn't the time saved. It's what you do with the time once desktop work is automated.

Frequently Asked Questions

What types of due diligence work has AI already automated in 2024-2025?
AI automates contract review, financial statement analysis, CIM processing, legal document parsing, and pattern recognition across large datasets. Private equity firms process CIMs in 2 to 3 minutes instead of 40. AI-powered document review cuts manual effort by 70% to 80%.

If everyone uses the same AI tools, how do investors differentiate?
Differentiation shifts from document processing speed to proprietary intelligence. When AI analyzes the same public documents for every firm, alpha comes from primary research: customer interviews, expert networks, and channel checks producing information competitors don't have.

Why do hedge funds still spend $2.5 billion on expert networks if AI processes documents so well?
Documents tell you what happened or what management says. Expert networks tell you what customers think, whether product roadmaps will work, and if competitive threats are real. AI processes existing information. Primary research creates new information. The expert network industry grew from $1.5 billion in 2020 to $2.5 billion in 2024.

What parts of due diligence will AI never automate?
AI won't replace getting people to say what they think beyond PR-safe answers, interpreting messy and conflicting reality in context, and deciding what intelligence is solid enough to change a position. These require persuasion, real-time interrogation, and judgment.

How should firms allocate analyst time freed up by AI?
Freed-up time should go toward primary research AI won't replicate: structured expert interviews, customer conversations, channel checks, and building proprietary intelligence. The ROI on AI isn't the time saved. It's what you do with the time once desktop work is automated.

Are firms building proprietary LLMs gaining an advantage?
It depends on the training data. If you train an LLM on public documents everyone else accesses, you've automated commodity research at high cost. Proprietary advantage comes from training on unique datasets: primary intelligence from customers, competitors, and partners who won't talk to the next firm.

What is the biggest blind spot for firms adopting AI in due diligence?
Firms use AI to evaluate AI readiness in targets, and they're still relying on documents to do it. The blind spot is assuming faster document processing creates differentiation. Differentiation comes from knowing what isn't in the documents.

How much are private equity firms investing in AI in 2024-2025?
Nearly two-thirds of private equity general partners run GenAI pilots as of September 2024, and over 40% use it in business processes. Private investment in generative AI reached $33.9 billion in 2024, up 18.7% from 2023. GenAI funding exceeded $56 billion in 2024, nearly double the $29 billion in 2023.

Key Takeaways

AI automated document-heavy due diligence work, cutting review times by 70% to 93%. When every firm has the same tools analyzing the same documents, speed no longer creates alpha.

Competitive advantage shifts to primary intelligence: customer conversations, expert interviews, and channel checks producing proprietary information AI won't access or replicate.

The expert network industry grew from $1.5 billion in 2020 to $2.5 billion in 2024. The gap between what documents say and what people know is where differentiation lives. AI processes existing information. Primary research creates it.

Nearly two-thirds of PE general partners run GenAI pilots as of 2024, but the ROI on AI-powered due diligence isn't the time saved. It's what you do with freed-up analyst time once desktop work is automated.

Are you deploying freed-up analyst time toward primary research, or are you processing commodity documents faster?