Contracts reviewed in one pass
On a mid-market deal, the practical gain is full-population clause extraction and faster triage across the whole data room. Exact accuracy and timing depend on clause type, document quality, and human validation.
You've just been engaged to advise a Basel Mittelstand company on a €8M acquisition. The target has 140 contracts sitting in a shared folder. You have two weeks, two associates, and a fixed fee.
That is the reality for most M&A boutiques in the DACH region. The due diligence challenge isn't managing 3,000 contracts with an army of associates — it's doing a thorough job on 140 contracts with two people, on budget, before the seller loses patience. AI changes that calculation in very concrete ways.
What AI Can Actually Do for a 140-Contract Deal
Purpose-built legal AI tools can scan a full contract repository and extract specific clause types: change of control provisions, assignment restrictions, termination rights, IP ownership, non-compete obligations, data processing provisions. For a 140-contract folder, a well-configured review platform can cover the entire data room in hours rather than forcing lawyers into multi-day manual triage.
The practical capabilities that matter at SME scale:
Full-population review. Instead of sampling 20% of contracts and hoping you didn't miss the one agreement with a change of control blocker, you review everything. On a 140-contract deal, 100% coverage is now the baseline expectation.
Cross-document consistency checking. AI can identify contradictions across documents — an IP assignment in an employment contract that conflicts with a representations clause in the SPA, or a software license that limits the right to assign. These are precisely the issues that surface-level sampling misses.
Triage and prioritisation. The AI produces a red flag report that lets your two associates know which of those 140 contracts need same-day attention versus which are standard and low-risk. That prioritisation is where the time saving actually materialises.
The Real ROI at SME Scale
For a large firm, AI due diligence means cost savings measured in associate hours. For a boutique on a fixed fee, the ROI framing is different — and more compelling.
The benefit isn't saving CHF 50,000 on associate time. It's closing in ten days instead of twenty-one, which means your client gets to sign before the seller's other bidder arrives. It's identifying the change of control clause that requires a key supplier's consent on day two, not day twelve, giving the deal team time to negotiate a waiver. It's delivering a red flag report with full-population coverage that a competing firm couldn't match on the same timeline and fee.
General counsel such as DHL's Mark Smolik have publicly described AI-assisted due diligence as commercially credible when paired with legal review. The transferable lesson for smaller firms is not that every platform reaches one universal accuracy score. It is that broader document coverage and earlier issue-spotting are now achievable on matters where teams previously had to rely on selective sampling.
The Swiss Corporate Law Gap You Cannot Outsource to AI
Here is the issue no vendor will put in their marketing materials: most legal AI tools were trained predominantly on US and UK legal documents. This creates specific blind spots for Swiss M&A work.
When you run a Swiss deal through any of the leading platforms, the AI does not reliably flag:
- Swiss AG/GmbH structural issues — shareholder agreement mechanics under OR Art. 697 et seq., mandatory pre-emption rights under cantonal notarial requirements, or capital band provisions under revised Swiss company law
- Swiss mandatory law provisions — certain protections for employees and minority shareholders cannot be contractually waived and an AI trained on common law documents will not know they exist
- Swiss merger control — COMCO notification thresholds are different from EU and German rules; the AI may not flag a filing obligation that a Swiss lawyer would catch immediately
These are not edge cases. They are structural features of every Swiss acquisition. The workflow implication: your AI handles extraction and triage on the commercial contracts, and your Swiss-qualified lawyer handles a mandatory law checklist that the AI cannot run for you.
Build this into your engagement methodology as an explicit step, not as an afterthought disclaimer.
Professional Indemnity: Two Questions to Ask Now
AI-assisted due diligence creates two PI questions that most firms haven't resolved.
First, your engagement letter should explicitly state that AI tools are used in your due diligence process and that all AI outputs are reviewed and verified by qualified counsel before delivery. Most sophisticated clients expect this disclosure; some require it. The client who discovers later that you used AI tools without disclosure will have a harder time accepting your explanation of why the process was sound.
Second, call your PI insurer and ask directly whether your policy covers AI-assisted due diligence. Most policies pre-date the widespread use of legal AI and do not contain explicit language addressing it. Getting clarity now — rather than after a claim — is straightforward risk management. Some insurers are beginning to add endorsements; others will confirm that existing language covers the use case. You need to know which applies to your policy.
A 5-Step SME Due Diligence AI Workflow
Step 1 — Data room intake. Organise the contract folder into categories before running the AI. Separate employment agreements, customer contracts, supplier agreements, IP licences, real estate leases, and regulatory permits. AI tools perform better on well-organised inputs and the categorisation itself reveals gaps in the data room.
Step 2 — AI extraction run. Upload the full contract population to your chosen tool. Configure extraction for the key clause types relevant to the deal structure: change of control, assignment, termination, IP ownership, data processing, non-compete. Specialist review platforms support this workflow across different hosting models and pricing structures.
Step 3 — Swiss mandatory law checklist. Run a parallel manual review against a Swiss-specific checklist that the AI cannot replace: COMCO thresholds, mandatory employee protections, notarial requirements for any real estate or share transfers, cantonal-specific requirements. This is two to three hours of senior lawyer time — not a full parallel review.
Step 4 — Triage and deep review. Use the AI red flag report to prioritise. The five to ten contracts that the AI has flagged as transaction-critical get full associate review. The remaining low-risk agreements get a confirmation read for anything the AI may have missed on Swiss-specific points.
Step 5 — Verified output delivery. The client deliverable is built from human-verified findings. Document your methodology in the file: what tool was used, what it was configured to extract, which findings were verified against the underlying document. This protects you if the process is ever scrutinised and demonstrates to the client that AI was a starting point for analysis, not a substitute for it.
The Net-New-Work Case: Why AI Creates Revenue, Not Just Savings
Harvey's founder identifies a paradigm shift in how to measure AI ROI for M&A work. The traditional metric — "we reduced review time from 40 to 8 hours" — is compelling but limited. The more powerful metric is net-new-work: "we took on an 18-month contract review, completed it in weeks, and won the follow-on M&A work." For a 10-lawyer Basel firm, the strongest proof point is the mandate you can now accept that you would previously have declined because the data room was too large for your team.
Public case studies from large legal departments and firms point in the same direction: AI-assisted due diligence is attractive because it improves coverage and speeds up prioritisation. For a regional or mid-market team, the useful takeaway is not a headline percentage. It is that full-population review becomes commercially viable on matters that were previously staffed through sampling and time pressure.
Information Architecture Before AI
Before licensing any AI tool, audit your document management system. Harvey's data shows that decades of poor DMS hygiene — wrong matter codes, "other" as the default category, inconsistent profiling — actively block AI value extraction. The prerequisite is what some practitioners call "IIA before AI" (Information Architecture before AI). A well-organised data room does not just help the AI perform better. It is the foundation without which the AI cannot perform at all.
AI Due Diligence Checklist for M&A
0/0Planning to deploy AI in your next transaction? Get in touch to discuss how to build AI use into your due diligence methodology with appropriate safeguards.