How AI can Improve Due Diligence Processes

Having spent the last 18 months researching AI in professional services and M&A transactions, I finally decided to put some of that learning into practice. At the beginning of this year, I embarked on a personal project to build an AI-driven Operational Due Diligence (ODD) and Technology Due Diligence (TDD) intelligence platform. The goal was simple, test whether AI could fundamentally transform how operational and technology assessments are conducted in M&A. I set out with three clear objectives: 1. Automate as much of the diligence process as possible (e.g. data ingestion, analysis, and report generation). 2. Accelerate insight generation while improving coverage of both deal risks and value creation opportunities. 3. Establish a foundation for institutional knowledge, benchmarking, and repeatable diligence playbooks across transactions. I had a couple of points in my favour. I know ODD and TDD processes extremely well, having advised on over 800 transactions and I also have a background in software development from earlier in my career. What genuinely surprised me throughout this project was how quickly a working platform came together. Leveraging AI coding agents, open-source tools, and some commercial infrastructure, I was able to build in a matter of weeks what would traditionally take a team months or an individual years. The result was an AI-powered diligence co-pilot: a platform capable of rapidly processing large volumes of unstructured data room materials, extracting structured insights, and generating presentation-ready outputs with full traceability back to source evidence. More important than the platform itself were the insights gained through building it. It reinforced the critical role of institutional knowledge in developing effective AI systems, this is not just a technology problem, but a domain expertise multiplier. It also gave me a far deeper appreciation of what GenAI can truly achieve in an M&A context. In many cases, the level of insight generated went beyond traditional diligence. For example, the ability to produce multi-scenario separation models with fully built cost-to-achieve schedules was genuinely striking. It offered a glimpse into what the future of M&A advisory could look like. Before retiring the platform, I documented a very high-level overview for anyone interested in understanding what is now possible. As always, feedback and comments are welcome. #AI #GenAI #MergersAndAcquisitions #PrivateEquity #DueDiligence

Adverse media is a major KYB bottleneck but banks are wary of automation. Our new Taktile Labs + Parallel Web Systems benchmark shows AI agents can now exceed human performance on this task. We partnered with Parallel to test an adverse media agent built on top of 7 frontier AI models with 47 real businesses. tldr: High performance of AI in this task signals the level of automation we are likely to see in core compliance work. our top findings: 1. AI agents produce better quality evidence than humans. This is based on a 20-point rubric covering evidence quality, ability to identify the correct entity, and risk assessment of findings. Human analysts average 13.50 while AI agents score 14.60. 2. A hybrid agent x human setup reduces workload by 93% Full automation is not necessary for high impact. A hybrid model where an agent screens first and a human reviews uncertain cases reduces analyst workload by 93%. 3. Even the “cheaper” models are still very good at fulfilling the task. Agent configurations using some of the less expensive models can still match human performance. Costs will continue to go down as new models are released and accuracy will keep getting better. 4. How you configure the agent matters more than the model you choose. The investigation strategy and prompt you put behind the agent has a much bigger impact on results than the model itself. 5. Agents do not hallucinate in this task but they can “over-flag”. This is huge. When we see false positives from agents, they are not fabricated sources (0% hallucination rate in this dataset). However the agent will sometimes flag sources too eagerly. Huge thanks to David Ahn and Maximilian Eber for leading the research and Parag Agrawal and Sahith J. for your partnership on this. See the full study via comments.

AI Is Changing Due Diligence Faster Than Most Dealmakers Realize For most of my career, diligence was a function of hours. You staffed a room, you read every contract, and you hoped the issue that killed the deal surfaced before signing rather than after. That model is quietly collapsing. In recent deals, I’ve watched AI tools surface change-of-control provisions across thousands of contracts in an hour, work that once consumed a team for a week. They flag the anomalous indemnity, the off-market earn-out, the buried regulatory exposure. Not perfectly. But fast, and improving faster than most boards appreciate. Here’s what I tell clients: the technology isn’t the disruption. The expectations are. When the other side can review everything, “we didn’t have time to look” stops being a defense. The premium shifts away from coverage and toward judgment, knowing which flagged clause actually matters, what a finding means for price, where the real risk lives versus where the machine is merely nervous. The lawyers and bankers who thrive in the next decade won’t be the ones who resist these tools. They’ll be the ones who let the machine do the reading so they can do the thinking. Diligence was always about finding the truth before you committed capital. AI just moved the deadline up. #MergersAndAcquisitions #DealMaking #ArtificialIntelligence #CorporateGovernance #PrivateEquity #Leadership

Just had a call with a customer whose risk analysts are firmly stuck in the pre-AI world. Here's what they're doing manually that will be automated: (Disclaimer: I don't blame this team at all, and there are so many like them making the jump to AI workflows. We're here to help!) Their current (manual) merchant verification process: 1. Manually searching business names across multiple sources 🔍 2. Cross-checking Secretary of State registrations 📑 3. Comparing website domain creation dates with "in business since" claims 📅 4. Reviewing Google/Yelp business status and ratings ⭐ 5. Scanning for adverse media mentions 📰 6. Checking physical location via Google Maps 🏢 7. Verifying social media presence (Instagram/YouTube) 📱 8. Looking for suspicious website elements (stock images, template text) 🚩 9. Verifying the payout bank account with voided checks 🏦 10. Calculating potential credit exposure for risk assessment 💰 Every analyst does this, and I don't blame them. The problem is, it's time-consuming, inconsistent across analysts and teams, and doesn't scale 👎 What excites me is how AI agents 🧠 can transform this workflow: - Automated data collection: Connect to multiple sources simultaneously to gather all relevant data in seconds ⚡ - Pattern recognition: Flag discrepancies that matter (like a business claiming 20 years of history with a 6-month-old domain) 🧩 - Contextual intelligence: Understand industry norms (like towing companies typically having lower ratings) 🔄 - Risk summarization: Provide the "net net" with key findings and specific risk factors, not raw data dumps 📊 - Guided recommendations: "Pause payouts," "Request additional documentation," or "Approve with monitoring" based on risk patterns and the company's risk appetite 📋 - Continuous learning: Improve detection by incorporating feedback from confirmed fraud cases 📈 Transitions like this are difficult once. The outcome is a senior analyst team member for everyone on your risk team that never gets tired, and always delivers insights. Leaving the manual processes behind forever. Trust me, it's worth it 🚀

AI in PE’s investing workflow today is a "v0 generator." It gets you to a strong first cut, fast. But Anita Kavalan and I believe the real edge lies elsewhere. We spoke to several global investors, and did a lot of research, and here’s where AI in investing shows up beyond the obvious (drafting emails, memos): 🏭 Market scans & benchmarks (revenue, EBITDA, fundraises): Days of reports and expert calls shift to minutes with deep research tools for a solid v0 📞 Expert calls: We’ve moved from frantic note-taking and manual synthesis across 30-40 calls, to auto-transcription (Granola) & LLM-powered pattern detection 📄 Complex docs (LDDs, CIMs): LLMs help with faster synthesis & clearer signals on where to focus 📈 Charts and images: LLMs convert them into usable tables sans manual cleanup ❓Internal data: Querying past deals, IRRs, MOICs gets easier - no need to dig through old models/chase teams 📊 Financial models: AI got us ~10% there in early 2025 to 70-80% today for simpler structures and credible v0s even for LBOs But here’s how firms can really make the most of AI: 1️⃣ Clear AI strategy - cause LPs are asking harder questions: ~47% of LPs are now monitoring how GPs adopt AI, and it's showing up in fundraising diligence. GPs are asked to cite deals they passed on where AI made the moat look fragile. Anyone can list tools they’ve bought. But if firms can’t answer how AI risks impact deals, that’s the gap before the next fund. 2️⃣ The real moat is decades old: Most firms using AI on internal data have a chatbot over a recent slice. The real edge is a model fine-tuned on IC memos, deal post-mortems, and portfolio performance over 20–30 years.  3️⃣ Beyond retrieval: A simple chatbot is a librarian. It’s key to go beyond "what's the underwritten IRR" but "this deal looks like three we passed on in 2014, here's why." The real copilot connects portfolio datasets, flags contradictions between data and management commentary, and helps form a view. That’s a different product entirely. 4️⃣ Getting AI to be the toughest Partner on the IC: Deal teams should feed deal docs into an AI sandbox, simulate the most contrarian IC voice, and generate counterarguments. If a deal doesn’t survive, it needs a rethink. If it does, conviction is stronger. AI is already being used across PE workflows to generate fast v0s in research, diligence, data, and modelling. But the real opportunity is to treat it not as automation, but as a strategic, #agentic layer that reshapes how conviction is built and investment decisions are made. Next up: AI in value creation and the eval problem. #StanfordGSB #PrivateEquity #AIinFinance #GenerativeAI #FutureOfFinance #AIinInvesting #DueDiligence #InvestmentStrategy #AlternativeInvestments #FutureOfWork #ArtificialIntelligence

A red flag without a next step is just décor. In most diligence processes, finding a red flag is the start of a chaotic scramble. It lands on a messy list of "things to check," leading to unfocused management meetings and expensive, open-ended workstreams for advisors. It's a recipe for burning weeks and losing focus on what truly matters. The best deal teams use AI to build a closed-loop system. It doesn’t just find anomalies; it forces a structured workflow from detection to decision. The process is simple and disciplined: Step 1: Detect the Anomaly. AI automatically scans all materials and flags a material inconsistency. (e.g., Gross margin is 55% in the CIM but 51% in the VDR). Step 2: Bind it to a Question. The system forces the team to immediately attach a concrete, specific question for management. (e.g., "Please explain the 400bps margin discrepancy for Q2."). Step 3: Scope the Work. The question is then linked to a time-boxed diligence task with a clear owner. (e.g., "Assign scope for review of customer profitability data to validate."). This transforms diligence from a chaotic treasure hunt into a hypothesis-driven process. The goal isn't to admire interesting problems; it's to efficiently de-risk the ones that can actually move your bid price. The result is a radically compressed timeline. Weeks of aimless investigation become days of focused inquiry. Management meetings get shorter and more productive, and your team builds conviction faster by systematically neutralizing the biggest risks. With AI, we stop admiring problems. The fastest path to conviction is a direct line from every red flag to a decision.

The compliance landscape is shifting from manual processes to intelligent automation. While traditional rule-based systems have improved efficiency, they're fundamentally limited by their inability to adapt to regulatory changes or interpret the significance of collected data. Our latest research examines how AI-powered evidence collection transcends these limitations. Unlike automated tools that simply gather predetermined artifacts, AI systems can: • Interpret compliance context – Understanding whether collected evidence actually demonstrates control effectiveness • Adapt to regulatory evolution – Automatically adjusting collection criteria as frameworks evolve • Identify risk patterns – Spotting anomalies and control gaps before they surface in audits • Provide continuous validation – Moving beyond sampling to real-time control effectiveness monitoring The data is compelling: Organizations implementing AI-driven compliance systems report significant reductions in audit preparation time while improving evidence quality and coverage. This isn't about replacing human judgment—it's about amplifying expertise with systems that can process, analyze, and contextualize compliance data at scale. Key insight from our analysis: The most successful implementations treat AI adoption as organizational transformation, not just technology deployment. Teams that invest in proper training and domain-specific configuration see measurable improvements in compliance posture within the first quarter. For compliance leaders evaluating this technology, the question isn't whether AI will reshape evidence management—it's how quickly your organization can implement these capabilities strategically. Read the full analysis: https://lnkd.in/e3YdB5YW #Compliance #RiskManagement #AI #GRC #AuditPreparation

Every LMM Acquisition Now Needs an AI Due Diligence Workstream Everyone has added technology diligence. Almost no one has added AI diligence. Those are no longer the same thing. Traditional technology diligence asks whether the systems work. AI diligence asks whether the company can still produce a number anyone is willing to rely on. Before closing, I would ask five questions. • What AI is being used today—even if no one formally approved it? • Which financial processes depend on AI-generated outputs? • Who owns governance for those systems? • Can every AI-assisted number be traced back to its source data? • What happens if every AI tool is disabled tomorrow morning? That last question surprises management teams. It shouldn’t. If turning AI off means payroll stops, forecasting breaks, revenue recognition becomes manual, or customer support collapses, AI is no longer an enhancement. It is critical infrastructure. And critical infrastructure deserves the same diligence as ERP, cybersecurity, or debt. Private equity spent decades perfecting financial, tax, legal, operational, and commercial diligence. AI deserves its own workstream. Because the next quality of earnings question won’t be: “How did you calculate EBITDA?” It will be: “Show me how AI participated in producing it.” The firms that can answer that question confidently will close faster, integrate faster, and command better valuations. The rest will spend diligence explaining what they don’t know.

Just built an IT Audit AI agent use case — and it’s only the beginning. Having shared posts over the past few weeks about the disruption Agentic AI is bringing to technology auditing, I thought… why not get stuck in myself? What many people don’t know: I started out as a coder. My early career included hacking Linux as a teenager — and now, years later, I’ve rekindled that curiosity. As a leader, I believe it’s important to understand things hands-on — especially when advising clients or coaching teams. I recently highlighted the emerging skills needed in risk and audit, and decided to upgrade myself. This week, I built a document summarisation agent using Python and OpenAI. I know there’s tons out there, but the point here is I put this together myself and any IT Internal Audit/ Compliance team can do this too! It scans long PDF reports, extracts key insights, and generates audit-relevant summaries — within seconds. Think about the time saved on policies, SOC reports, or control narratives. But more than that: Agentic AI isn’t just a trend. It’s a smart, practical partner for IT risk professionals. I’ve written a whitepaper on “Agentic AI in IT Risk & Controls” exploring how goal-driven agents can enable: • Continuous control monitoring • Automated evidence collection • Change risk evaluation in DevOps • Vendor risk review automation • Narrative generation for audit reports Client example: One of our consumer & retail clients manages over 20 third-party systems. To improve access reviews and change assurance, we deployed an agent that connects to ServiceNow and IAM logs, flags risky changes, and drafts evidence summaries. It’s reducing audit prep time and enabling real-time insight. At EY, our audit and technology risk teams are diving deep into how Agentic AI is disrupting business processes. We’re mapping out new risk models and defining what future-ready auditors will need in terms of skills and mindset. We’ll be sharing our views on this changing compliance landscape soon. Big thanks to Piers C. Maree-Louise Kernick Gareth James Abhishek Mohal Jason Walters Danila Solovyev Kevin Duthie — and everyone else helping shape the future of digital trust. If you’re in risk, audit, or compliance — or just curious about how AI is reshaping our field — let’s connect. The journey has only just begun. #ITRisk #AI #AgenticAI #Python #AuditInnovation #OpenAI #RiskTransformation #DigitalTrust #Leadership

“How can Business Analysts actually use AI in day-to-day project work?” This is one of the most common questions I get from aspiring and even experienced BAs. Let me share a few practical examples from a KYC (Know Your Customer) project: 🔹 Requirement Workshops In KYC, stakeholders often throw around terms like “CDD” or “Enhanced Due Diligence.” AI can help prepare a draft glossary and potential discussion questions so you enter the workshop with clarity and confidence. 🔹 User Stories & Acceptance Criteria A raw need like “We must capture beneficial ownership details for corporate clients” can be structured by AI into: As a Compliance Officer, I want the system to capture beneficial ownership details, so that regulatory checks can be completed. AI can also suggest acceptance criteria such as validation rules, mandatory fields, and exception handling. 🔹 Process Mapping From interview notes with Operations, AI can generate a draft KYC process flow — e.g., customer onboarding → document upload → screening → risk scoring → approval. You still refine it, but it saves you hours of manual effort. 🔹 Impact Analysis When regulations change (say FATF or local AML rules), AI can help brainstorm downstream impacts: 👉 Updates to the screening engine 👉 Changes to data shared with reporting regulators 👉 Impact on customer onboarding SLAs 🔹 UAT Test Cases From requirements like “System should flag PEPs (Politically Exposed Persons)”, AI can generate test scenarios: 👉 Positive case: Customer is a PEP → flag raised 👉 Negative case: Customer not a PEP → no flag 👉 Edge case: Missing PEP database connection 🔹 Documentation & Communication KYC projects involve Compliance, Risk, IT, and Business teams. AI can help you tailor the same update into: 👉 A regulatory summary for Compliance 👉 A technical impact note for IT 👉 A business-friendly update for Operations 👉 The takeaway: AI doesn’t replace a BA in KYC projects—it acts like a co-pilot. It handles the heavy lifting of structuring, summarizing, and brainstorming so you can focus on what matters most: stakeholder alignment and delivering value. Create AI agents for BA: https://lnkd.in/eb8t2Ye3 AI Playbook for BA: https://lnkd.in/e49Hp5t6 https://lnkd.in/ekUqn2Vx BA Helpline