Practice Areas
AI for M&A Lawyers: Diligence, Drafting and Deal Management
AI for M&A lawyers now compresses diligence by 15–75%, drafts disclosure schedules and runs closing checklists. A deal-phase map of agent tasks, gates and fees.
On this page
AI has changed three parts of an M&A mandate as of September 2026: diligence, where agents read data rooms and produce cited red-flag grids; drafting, where models conform schedules, definitions and ancillary documents to the negotiated text; and deal management, where checklist agents track conditions, signatures and deliverables. The negotiation of risk allocation has not changed. What has changed is that clients now ask what all this saved them.
A deal-phase map of agent tasks
The useful way to think about AI on a transaction is by phase and by gate: what an agent does, what it hands to a lawyer, and who signs off before anything leaves the building.
| Deal phase | Agent task | What it returns | Human gate |
|---|---|---|---|
| Pre-deal / outside-in | Public-record scan of the target, group chart from filings, litigation and regulatory hits | Briefing note with links | Partner reads before first client call |
| NDA and process letter | Redline of the buyer's NDA against the firm playbook | Tracked changes with rationale | Associate accepts or rejects each change |
| Data-room diligence | Tabular review of every contract against a question set: change of control, assignment, exclusivity, MFN, term, governing law | Grid with a source citation per cell, red-flag list | Reviewer opens every red flag and a sample of green cells |
| Diligence report | Draft report sections from the grid, in house style | Sectioned draft with open points | Senior associate rewrites conclusions; partner signs |
| SPA drafting | Conform precedent to term sheet; build definitions; cross-reference check | Marked-up draft, issues list | Lead lawyer owns every commercial clause |
| Disclosure schedules | Populate schedules from the diligence grid and the seller's responses | Schedule draft with source references | Seller's counsel confirms each disclosure |
| Signing | Signature pages, board consents, closing memorandum | Document set from a checklist | Review gate before any consent executes |
| Closing | Condition-precedent tracker, deliverables chase, funds-flow check | Live checklist, reminders | Closing lawyer confirms each CP satisfied |
| Post-closing | Integration obligations tracker, earn-out and escrow dates | Calendar with document links | Matter partner reviews quarterly |
The gates are not decoration. Harvey's own long-horizon benchmark found that on a strict all-criteria-pass metric even the best model completed only 13.3% of long legal tasks flawlessly as of 18 September 2026, with criteria-level pass rates around 83–88% (LAB leaderboard). Mostly right is what an agent gives you; the gate is where mostly becomes entirely.
Diligence: what compresses and what does not
The measured numbers come from firms that have run agents on live deals. GSK Stockmann reports 15–20% time savings on standard diligence and up to 75% on unstructured data rooms using Harvey (Harvey, 29 April 2026). PwC's Harvey partnership has generated more than 10,000 diligence red-flag reports. Addleshaw Goddard chose Legora after nearly a year of evaluation because its Tabular Review can analyse up to 100,000 documents in parallel (Addleshaw Goddard, March 2025). Thomson Reuters' rebuilt CoCounsel Legal offers Tabular Analysis at 10,000 documents against 100 questions in one pass (August 2026).
The figure most often repeated at conferences is the Legalweek 2026 claim (19 March 2026) that financial-contract diligence had been compressed from 15–20 associate hours to about two, and that "the junior associate is the new existential crisis". Treat it as reported rather than measured: no public primary source backs the specific hours. The sourced range is 15% to 75%, and the spread is the point. Standard leases and NDAs compress a lot; bespoke commercial agreements with side letters compress less; anything where the question is "is this a problem for the buyer" rather than "does this clause exist" compresses least.
Two tasks resist compression. The first is judgment about materiality, which is why the gate in the table above says the reviewer opens every red flag. The second is the data itself: scanned, mis-filed and duplicated documents. Run an OCR pass before the agent sees the room; Claude will say "illegible" rather than invent a figure, but only if it is looking at an image it cannot read rather than a text layer that says the wrong thing. The due diligence use-case page covers the question-set design that determines whether a diligence grid is useful at all.
Drafting the SPA and the disclosure schedules
On drafting, the independent evidence is sobering. In the Vals Legal AI Report of 27 February 2025, lawyers beat every tool on redlining, 79.7% against Harvey's 65.0%, while the tools won on document Q&A and extraction (Vals). Vendors say the gap has closed: Anthropic claims Claude Opus 5's first-turn NDA redlines scored "nearly double Opus 4.8" (24 July 2026), and GC AI's RedlineBench rose from 47.9 to 57.0 with Fable 5.1 (September 2026). Both are vendor numbers. The practical reading is that a model now produces a competent first markup of a standard document and a poor one of a heavily negotiated SPA, and that the lawyer must read every change either way.
Where models are already strong is the unglamorous drafting that consumes associate nights: conforming definitions after a negotiation round, checking cross-references, generating the ancillaries from an agreed SPA, and above all populating disclosure schedules from the diligence grid. Here is the prompt pattern for schedules:
Role: Corporate associate assisting seller's counsel. Work ONLY from the
attached files. Do not add any disclosure not supported by a source.
Attached:
A. Executed-form SPA, clause 9 (warranties) and Schedule 4 (disclosure)
B. Diligence grid (CSV): document ID | contract | counterparty | change of
control | assignment | exclusivity | term | governing law | red flag | note
C. Seller's responses to buyer's RFI, rounds 1–3
D. House-style disclosure schedule from a precedent deal (structure only)
Task:
1. For each warranty in A, list every item in B or C that would make the
warranty untrue or misleading if not disclosed. Cite document ID.
2. Draft a specific disclosure against each such warranty in the style of
D: one paragraph, the fact, the document, the amount or date if any.
3. Flag as [CONFIRM WITH SELLER] any disclosure that depends on a fact
stated in C but not evidenced in B.
4. Flag as [GENERAL DISCLOSURE?] anything that arguably belongs in the
general disclosures rather than a specific one.
5. Output: Warranty ref | Draft disclosure | Source IDs | Flag.
Do not draft general disclosures. Do not paraphrase the warranty text.
Every disclosure then goes past the seller's counsel, and the [CONFIRM WITH SELLER] flags become a call. That is the workflow the Claude for Legal corporate plugin implements with its "every cell cited" tabular diligence and disclosure-schedule skills; the open-source repository shows the gate design and is worth reading even if you never use the tool (github.com/anthropics/claude-for-legal). The contract drafting page goes deeper on first drafts and negotiated text.
Signing and closing: the checklist agent
Closing is project management, and project management is what agents do well when the inputs are structured. A closing agent needs three things: the conditions-precedent list from the SPA, the deliverables list from the closing memorandum, and access to the document store. From those it maintains a live table, chases named people for named documents on a schedule, and produces the signing and closing sets. This is a workflow, not an autonomous agent, and it should be built as one:
- Extract every condition, deliverable and signatory from the executed-form SPA and the closing memorandum into a table: item, responsible party, due date, source clause.
- Have the lead associate correct the table once. This is the only step that needs judgment.
- Connect the table to the document store through a governed connector (iManage and NetDocuments both ship MCP servers that run under the user's own credentials and respect ethical walls; see connecting Claude to your DMS).
- Schedule the agent to check daily which items now have a document attached, mark them satisfied, and draft chasers for the rest. Chasers are sent by a human.
- Put a hard review gate before any board consent, power of attorney or signature page is generated for execution. The Claude for Legal corporate plugin's review gate before board consents is the model.
- On closing day, the agent produces the closing index and the post-closing obligations calendar from the same table.
The A&O Shearman and Harvey agents launched on 6 April 2025 (antitrust filing analysis, cybersecurity compliance, fund formation, loan review) were pitched as doing "in minutes what previously took several hours" and are sold to clients on subscription with a revenue share (A&O Shearman). ContractMatrix, the same firm's drafting tool with Harvey and Microsoft, dates from 2023. The lesson for a deal team is that a repeatable deal workflow is a product, and the firms that built one first now sell it.
Data rooms, connectors and confidentiality
Anthropic's Claude for Legal (12 May 2026) ships connectors to Datasite, iManage, NetDocuments, Box and DocuSign among more than twenty; the corporate plugin includes a "dataroom-watcher" agent and detects Luminance or Kira output so it can be quality-checked rather than redone. The question this raises for a deal team is not capability but permissions. A data room is the seller's; the NDA and the process letter govern what leaves it. Before an agent reads a room, confirm three things: that your firm's platform terms exclude training on client data (Harvey, Legora, Team and Enterprise tiers of the frontier vendors and the API all do, as of September 2026; consumer tiers do not by default), that the connector respects the room's own access rights rather than your master login, and that the process letter does not prohibit automated extraction. The confidentiality checklist lists the current terms tool by tool.
The strongest models are also the least private. Claude Fable 5.1 requires 30-day retention and is not eligible for zero data retention without express authorisation; the Team and Enterprise chat interfaces are not ZDR-eligible either. For a public-company target with inside information in the room, that is a reason to run diligence on Opus 5 or Sonnet 5 under a ZDR arrangement, or inside Harvey or Legora, rather than on the newest model.
What clients now expect on fees
Deal clients have noticed. Deloitte Legal's survey of 121 senior legal leaders (April–May 2026) found 92% expect or are negotiating AI-related rate cuts, and 58% say firms "rarely or never" proactively discuss AI benefits (Deloitte). On 1 September 2026 the FT and New York Post reported that Citi now asks law firms bidding for its work to disclose their AI savings and that Morgan Stanley intends to put most outside work out to competitive bid with fixed fees by year-end (NY Post). Citi's own data, meanwhile, show demand for Big Law hours up 4.2% in the first half of 2026, which Bloomberg Law called "sort of a mystery" (Bloomberg Law).
The workable answer on a deal is to price the phases differently. Diligence on a defined data room at a fixed or capped fee, with the agent's grid as the deliverable and a stated sampling rate for human review; drafting and negotiation on hourly or blended rates; closing management as a fixed line. Texas Opinion 705 (February 2025) says you cannot charge for time "saved" by AI, Virginia LEO 1901 (November 2025) says a reasonable value-based fee need not be reduced because AI saved time, and California's 2026 guidance treats general AI subscriptions as overhead. The billing page reconciles those; the billable hour guide sets out who keeps the dividend. Module 4 of the AI Lab for Lawyers course builds a closing-checklist workflow of the kind described above, live and without code.
Where this leaves a deal team
Diligence has become a supervised extraction exercise with a human reading the exceptions; drafting has become a conforming-and-populating exercise with a lawyer owning the commercial text; closing has become a tracked table. The partner's job on a deal is unchanged, and the mid-level associate's has become more interesting. The junior's has changed most, which is why the junior lawyers page matters to every M&A group. What to do this month: pick one live deal, run the diligence grid on the data room in parallel with the conventional review, count the exceptions each method missed, and price the next mandate on what you learn. The corporate lawyers page picks up the post-closing governance work, and the Harvey and Legora reviews cover the two platforms most deal teams are choosing between.
Frequently asked questions
How much faster is AI due diligence in M&A?
It depends on the data room. GSK Stockmann reported 15–20% time savings on standard diligence and up to 75% on unstructured data rooms with Harvey in April 2026. PwC's Harvey partnership has produced more than 10,000 diligence red-flag reports. A widely repeated Legalweek 2026 figure put financial-contract diligence at about two hours instead of 15–20 associate hours, but that number has no public primary source.
Can AI draft a share purchase agreement?
It can produce a first draft from your precedent, a term sheet and a set of instructions, and it is good at conforming definitions, cross-references and disclosure schedules to the agreed text. It cannot decide the allocation of risk. In the last independent benchmark lawyers still outperformed every tool on redlining, so treat AI output as a prepared markup for a lawyer to own.
Which AI tools do M&A teams use in 2026?
Harvey (used by 84% of firms with 700-plus lawyers per the ILTA 2026 survey) and Legora (Tabular Review up to 100,000 documents) dominate at large firms, alongside Kira, Luminance and Thomson Reuters Tabular Analysis. Claude for Legal's corporate plugin adds tabular diligence, disclosure schedules and closing checklists with connectors to Datasite, iManage and NetDocuments.
Will clients expect a discount because we used AI on their deal?
Increasingly yes. Deloitte found 92% of in-house teams expect or are negotiating AI-related rate cuts, Citi asks bidding firms to disclose AI savings, and Morgan Stanley plans competitive fixed-fee bidding for most outside work by year-end 2026. Fixed or capped fees for diligence phases are becoming the norm on larger mandates.
Put this into practice
AI Lab for Lawyers: a live, hands-on course for practising lawyers
4 live sessions (8 hours), recordings included, taught by Dr. Niklas Schmidt of Wolf Theiss. Rated 4.7/5 from 381 ratings; alumni from Norton Rose Fulbright, Paul Hastings, Pillsbury and Taylor Wessing.
Related reading
Use Cases
AI Due Diligence: From Twenty Associate Hours to Two
AI due diligence now runs as a tabular agent over the data room: 10,000 documents against 100 questions, every cell cited. Numbers, a question matrix, the junior gap.
10 min read
Use Cases
AI Contract Drafting: From First Draft to Negotiated Text
AI contract drafting works from your precedents, not a blank prompt. A drafting prompt, a worked SaaS example under Austrian law and the risks a redline hides.
8 min read
Tools
Harvey AI in 2026: What It Does, What It Costs and Who Uses It
Harvey AI reviewed as of September 2026. The $15.5bn valuation, Harvey II, Tenet, 500-plus agents, who uses it, estimated seat prices and its own benchmark.
9 min read
Tools
Legora Review: The Agentic OS Challenging Harvey
Legora reviewed as of September 2026. The $5.6bn Stockholm challenger to Harvey: aOS, Tabular Review, Wexler litigation facts, pricing, clients to Salesforce.
7 min read
Guides
AI and the Billable Hour: Who Keeps the AI Dividend?
AI and the billable hour in 2026: associate rates at $798 and hours up 4.2% while Citi demands AI savings. What the ethics rules allow, three fee scenarios and a clear view.
9 min read
Practice Areas
AI for Corporate Lawyers: Governance, Board Papers and Company Secretarial Work
AI corporate lawyers use models for board minutes, resolutions, entity compliance and shareholder documents, and scheduled agents for housekeeping. Design and gates.
9 min read