What problem does it solve?
Deal teams spend much of their time on work that comes before judgment. On the sourcing side, private equity firms, corporate development teams and bankers track large numbers of companies to find the few that fit a thesis, often with analysts who assemble lists and one pagers by hand. Testing whether one business fits the thesis can take an analyst 20 to 25 hours, according to a startup founder quoted in EQT's ThinQ publication. Good targets are missed because nobody noticed the signal in time, and the same research is redone for every new mandate because lessons from past deals sit in individual inboxes.
Once a deal is live, the data room opens and the clock starts. Associates, lawyers and advisers read contracts, financial statements, customer agreements and corporate records to find change of control clauses, unusual liabilities, customer concentration and missing documents, then write it up. Sellers must redact personal data before bidders see it. All of this is repetitive, high stakes and time boxed, and fatigue raises the risk that material issues slip through.
- Datasite's chief product officer says sellers might have up to 100,000 documents associated with a deal.Datasite automates M&A and speeds redaction by 80%, saving customers valuable time with Azure Cognitive Services (2021)
- Testing whether a business fits a private equity firm's thesis and has value creation potential can take an analyst 20 to 25 hours, according to Clarum cofounder Anton Otaner.AI Promises to Make Private Equity Faster as Competition Heats Up (2025)
How does it work?
- Screen the market against the thesis. The assistant maps companies from licensed data, filings, news and the firm's own CRM, finds similar companies and ranks them against the investment thesis and signals such as growth, hiring or ownership changes, with the reasons.
- Build the company profile. For a shortlisted target it drafts a profile: business model, financials, competitors, ownership, management and news, with a source for every figure, and adds what the firm learned from comparable past deals.
- Read the data room. When diligence starts it classifies the documents, extracts key terms (change of control, exclusivity, termination, liabilities, key customers and suppliers) into a structured table and flags what is missing against the diligence request list.
- Flag risks and redact. It highlights clauses and figures that deviate from the norm, and on the sell side finds personal and sensitive data for batch redaction before bidders get access.
- Answer questions with citations. Deal team members ask questions across the whole data room and get answers that link to the exact page, so every statement can be checked.
- Draft the memo. It drafts the diligence findings and the investment committee memo from the firm's template; the deal team verifies, completes and owns every conclusion.
- Audience
- Employee facing
- Autonomy
- Copilot
- Adoption
- Early adopters
- Channels
- Internal tools, API and system to system
What is it worth?
Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.
| KPI | Median | Reported range | Data points | Claimed by |
|---|---|---|---|---|
| Users served | Not pooled | at least 6000 | 1 | 1 vendor |
| Productivity gain | Too few to pool | Not pooled: up to 80% | 0plus 1 up to | 1 vendor |
Value drivers: Speed and cycle time, Employee productivity, Risk and loss reduction.
Indicative value
A mid market private equity firm that takes 15 companies a year into full due diligence
USD 135,000 to USD 1.1 million
Deal team time released in due diligence per year
How this is calculated
Formula: deals * hoursPerDeal * shareSaved * hourlyCost. The low scenario uses every low input, the high scenario every high input.
| Input | Low | High | Basis |
|---|---|---|---|
| Companies taken into full due diligence per year deals, deals per year | 15 | 15 | The reference firm. |
| Internal hours per deal on document review, extraction and memo drafting hoursPerDeal, hours per deal | 300 | 600 | Editorial assumption for the deal team's own time, excluding external advisers. Replace with your own time records. |
| Share of those hours the assistant saves shareSaved, fraction of hours | 0.2 | 0.4 | Conservative against the evidence on this page (Microsoft reports that Datasite's Redaction AI cuts redaction times by up to 80 percent), because redaction is the most mechanical step and review and judgment stay with the team. |
| Blended cost per deal team hour hourlyCost, USD per hour | 150 | 300 | Editorial assumption for a blended associate and principal cost. Replace with your own. |
What it leaves out: Internal time only. It leaves out savings on external advisers, the value of deals found earlier or not missed in sourcing, the effect of issues caught or missed on deal value, and the cost of the assistant and data licences.
Market estimates (analyst estimates, not deployments)
- Allvue research cited by EQT found that 82 percent of private equity and venture capital firms reported using AI in some capacity by the end of 2024, up from 47 percent a year earlier. Why Private Capital Needs to Embrace Artificial Intelligence (2025)
Who already uses it?
4 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.
Freshfields
United Kingdom · Professional services · 2025
Freshfields, a global law firm, built Dynamic Due Diligence (D3), a proprietary tool designed to enhance legal reviews and due diligence, and in 2025 announced that Google's Gemini models would power it. A year into the collaboration the firm reported that D3 is one of several bespoke Freshfields Lab platforms now running on Gemini, and a partner who co leads Freshfields Lab said teams and clients use Gemini daily across those platforms. The firm has not published a separate outcome for D3.
No outcome disclosed.
EQT
Sweden · Wealth and asset management · 2016
EQT, a global private markets investor, has run Motherbrain, its in house data and AI team and platform, since 2016. EQT says it uses Motherbrain across the firm to source deals and help investment teams make better informed decisions, for example by measuring the similarity between companies for competitor mapping, and a partner describes tools that rank potential targets by attractiveness across a range of criteria. The platform also captures lessons from past deals and tracks the deal pipeline, and EQT stresses that AI supports rather than replaces the dealmakers' judgment. No outcome figures have been published.
No outcome disclosed.
Rogo
United States · Technology and software · 2025
Rogo, a New York AI company serving investment banks and private equity firms, combines a firm's own memos, research and files with external sources such as SEC filings, PitchBook, S&P Global, FactSet and Preqin, and automates workflows such as company profiles, competitive benchmarking, slide decks and investment memo drafts. Google Cloud reports that moving to Gemini 2.5 Flash cut hallucination rates in Rogo's evaluation, and counts thousands of bankers and analysts on the platform.
- Users served: at least 6000, investment bankers and analysts on the platform
"Builds trust in the Rogo AI platform among 6,000+ investment bankers and analysts"
Claimed by: vendor
Datasite
United States · Technology and software · 2021
Datasite, a virtual data room provider for mergers and acquisitions, added Redaction AI to its Datasite Diligence application. It uses named entity recognition to find personal and sensitive data across the documents a seller prepares for due diligence, so bankers and lawyers on the sell side can redact in batches instead of one document at a time; machine translation helps them file documents in other languages. Microsoft reports that redaction time falls sharply.
- Productivity gain: up to 80%, redaction step only (share of time and resources saved on redacting data room documents before due diligence), not the end to end deal cycle
"Now, the lawyers and investment bankers who coordinate sales can reduce redaction times by up to 80 percent, helping to move deals forward faster and support successful outcomes."
Claimed by: vendor
How do you implement it?
A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.
Data you need
- A written investment thesis and screening criteria per strategy or mandate
- Licensed company, financial and transaction data, plus the firm's CRM and past deal records
- A diligence request list and a key terms checklist per deal type
- Memo and findings templates approved by the investment committee
- Confidentiality and information barrier rules per deal
Systems to integrate
- Market data and company databases (for example PitchBook, S&P Global, FactSet, Preqin)
- CRM or deal pipeline system
- Virtual data room and document management
- Document storage such as SharePoint for memos and past deal files
- Collaboration tools where the deal team works
Complexity: High
Summarizing one document is easy. The hard parts are licensed market data and entity matching for sourcing, secure access to data rooms under strict confidentiality, reliable extraction from scanned and inconsistent documents, citations for every statement, and a review process that deal teams, lawyers and investment committees accept.
- 1
Start with one deal type and one step
Pick the most repetitive step for your team, such as first pass contract review for one deal type or company profiles for one strategy, and measure it before widening the scope.
- 2
Write the checklist before the prompt
Turn the diligence request list and key terms checklist into explicit fields with definitions, so the extraction is complete and comparable across deals.
- 3
Make citations mandatory
Every extracted term, figure and memo statement links to the page it came from, and the assistant says when something is not in the data room rather than guessing.
- 4
Test on closed deals
Run the assistant on data rooms and outcomes from past deals and compare its findings with what the team and advisers found, including the issues that mattered most.
- 5
Protect the deal
Keep each deal's documents in a separate, access controlled space, enforce information barriers, and make sure no deal data trains or reaches an unapproved model.
- 6
Define who signs what
Agree with the deal team, counsel and the investment committee which outputs are drafts, who reviews them and how reviewed findings are marked in the memo.
Guardrails
- Every statement in a profile, table or memo cites its source page, with a refusal when the source is missing
- Deal data isolated per deal and user, with information barriers and no training on client data
- Extraction checked against a fixed checklist, with confidence flags on uncertain fields
- Redaction reviewed by a person before any document is released to bidders
- Screening criteria documented, so targets are not excluded for reasons nobody can explain
KPIs to instrument
- Hours per deal on document review and memo drafting, before and after
- Recall of material issues on a test set of closed deals, compared with the team's own review
- Share of extracted terms corrected by reviewers
- Time from data room opening to first findings
- Share of sourced targets that reach a first meeting or a term sheet
Human in the loop
The deal team owns sourcing decisions, the diligence findings and the memo. Associates check extracted terms against the source, counsel reviews legal findings and redactions, and the investment committee decides on an explicitly human reviewed memo. A sample of AI findings is compared with adviser reports after each deal to calibrate trust.
Common failure modes
- A fluent memo with a wrong number
- A figure is misread from a scanned document or taken from the wrong period, and it survives into the investment committee memo. Require citations and check every material figure against the source.
- Silence taken for comfort
- The assistant finds nothing on an issue because the document is missing or unreadable, and the team reads that as no issue. Report gaps against the request list explicitly.
- Confidential deal data leaking
- Documents from one deal reach another team, another deal or an external model. Isolate data per deal and control which models and tools may see it.
- Sourcing that only finds the obvious
- Rankings built on the same data every competitor licenses surface the same companies. Combine proprietary signals and past deal knowledge, and review targets the model ranked low.
What are the risks and rules?
EU AI Act
Depends on design
Decision support for professional investors and advisers about companies is not a use listed in Annex III and is not a practice prohibited by Article 5. The users are deal professionals who know they are working with an AI tool, and no consumer interacts with it, so the Article 50(1) duty to disclose an AI interaction has little practical effect. Article 50(2) is different: a firm that builds the assistant itself, including on a platform such as Blits.ai and putting it into service under its own name, is the provider of that system and must mark generated text in a machine readable format, unless the system only performs an assistive function for standard editing or does not substantially alter the input data or its semantics, which may cover extraction and redaction. A firm that instead licenses a vendor product, such as Datasite or Rogo, should confirm that the vendor meets this duty. Obligations are otherwise general: AI literacy for the deal team under Article 4 and, where personal data in the data room is processed, the GDPR.
Rules that apply
Guidance
- Risk Outlook report: The use of artificial intelligence in the legal market (Solicitors Regulation Authority, Europe). The regulator of solicitors in England and Wales on the risks of AI in legal work, including accuracy, confidentiality and supervision, relevant when AI supports legal due diligence.
- Guidance on AI and data protection (Information Commissioner's Office, Europe). How data protection law applies to AI systems, relevant to the employee and customer personal data found in data rooms.
Controls to put in place
- Inventory entry for the assistant with an owner, approved data sources and approved models
- Access control and information barriers per deal, with logging of every query and document read
- Insider list and inside information handling for deals involving listed companies
- Human review and sign off recorded for every finding that enters the investment committee memo
- Periodic accuracy testing on closed deals and after every model change
Frequently asked questions
- Can AI do due diligence on its own?
- No. It can read, extract, compare and draft, and Datasite's chief product officer says its AI features can potentially compress weeks of work into days, but the findings and the decision stay with people. An article in EQT's ThinQ publication reports a consensus that screening and early diligence are ripe for automation, while confirmatory checks and negotiation are not.
- How much time does AI save in M&A due diligence?
- It depends on the step. Microsoft's case study on Datasite reports that Redaction AI lets sell side bankers and lawyers reduce redaction times by up to 80 percent, a figure Datasite bases on what customers tell it. Review and memo drafting are likely to save less because every finding needs checking, so measure hours per deal before and after.
- How do private equity firms use AI for deal sourcing?
- EQT, for example, says it uses its Motherbrain platform to source deals, including a model that measures how similar companies are for tasks such as competitor mapping, and an EQT partner describes tools that rank potential targets by attractiveness across a range of criteria. EQT stresses that AI supports, rather than replaces, human decision making.
- Is it safe to put data room documents into an AI tool?
- Only with controls: data isolated per deal, information barriers, no training on client data, logging of every access and approved model providers in approved regions. Freshfields, for example, runs its Dynamic Due Diligence tool on Google's Gemini models as part of a strategic collaboration with Google Cloud, and says its people use Gemini, NotebookLM Enterprise and Google Workspace daily with strong governance.
How to cite this page
Blits.ai AI Use Case Library, "AI assistant for deal sourcing and M&A due diligence", last verified 27 September 2026, https://www.blits.ai/ai-use-cases/deal-sourcing-and-due-diligence-assistant. Licensed under CC BY 4.0. Method: how we verify use cases.
Changelog
- 27 September 2026: First published