AI is winning the data room but the relationship is still yours

Ask where it hasn't, and a common theme emerges. Building the trust that gets a deal signed. Reading the room when a vendor wobbles on price.

Over recent weeks we've been talking with customers about how AI is changing their work. For many, AI is reshaping the mechanics of dealmaking, not its human core. But many also flagged caveats that can't be skipped, and commercial consequences firms should be ready for.

Where AI has earned its place

Unsurprisingly, many of the customers we spoke to point to diligence as the place where the gains of AI in dealmaking are real, measurable and already priced in. The work suits the technology: high volume, structured, pattern-driven and unforgiving of fatigue.

The wider market data points the same way. Bain & Company's Global M&A Report 2026 found 45% of M&A executives used AI tools in 2025, more than double the year before, and around a third are now using it systematically or redesigning processes around it. McKinsey research puts the impact at deal timelines 10% to 30% shorter and costs roughly 20% lower.

In the data room, AI can classify documents, flag what's missing and surface the clauses that move value. Change-of-control triggers, unusual indemnities, inconsistencies between the financials and the contracts behind them. Some describe tasks that once took a team days now taking hours.

The benefit cuts both ways. Buy-side teams reach the red flags faster. Sell-side teams can see their data room the way a bidder will, before a bidder ever logs in. Neither is a novelty any more – many dealmakers see both fast becoming the baseline.

Where the human edge holds

From our conversations, there was broad agreement on what AI hasn't touched. The deals that matter still start with a conversation, and origination remains relational.

AI can map a sector, screen targets and build a longlist in an afternoon. What it can't do is know that a founder is quietly ready to sell, or that a board is split on a merger it publicly supports. That knowledge comes from years of trust, built one meeting at a time.

The same goes for the moments that decide outcomes: structuring around a vendor's real priorities, holding competitive tension when bidders waver, telling a client something they'd rather not hear. These are judgement calls, and several customers told us their own clients want a person accountable for them. An M&A partner we spoke with says AI can assist with valuation, negotiation and stakeholder management, "but you still require extremely experienced senior human oversight." Bain's research on generative AI in M&A reaches a similar conclusion: the technology supports skilled practitioners, but it doesn't take their seat.

The caveats nobody can skip

Where customers were enthusiastic about AI in diligence, that enthusiasm often came with conditions. Three came up repeatedly.

Liability doesn't transfer to the model

Some AI models are still known to hallucinate, and in a transaction a confident error can be worse than an obvious one. If a model misreads a material contract and the deal prices on that mistake, the question is simple: who carries the loss? Customers who raised this were clear that accountability stays with the firm that signed off. A head of M&A who builds models with AI is blunt about where it fails: "It'll do 95% of things better than any human could, but there will be 5% of things that inexplicably, it'll just mess up," he says. "You cannot pass that buck on." The risk isn't theoretical. In 2025, Deloitte agreed to partially refund the Australian Government for a A$440,000 report containing apparent AI-generated errors, including a fabricated quote from a Federal Court judgment.

Human verification is mandatory, not optional

Some customers rolling out AI across legal and advisory work tell us they're building review into the process from day one. One leading law firm trains every lawyer on its AI tools and their limitations. As a partner there puts it: "We need to be using it in a responsible way that enables us to continue to provide the robust, defensible and trusted advice that our clients expect." AI drafts, flags and summarises. A qualified person checks, decides and owns the output. Firms that treat verification as a nice-to-have are taking on risk they can't see. They're not alone in their caution: half the professionals in Thomson Reuters' Future of Professionals Report 2025 named the accuracy of AI-powered tools as a concern.

Client data will always be a pressure point

Clients want to know which tools touch their confidential information, where it's stored and whether it trains anyone's model. A managing director at a corporate finance boutique says client NDAs increasingly rule out training on data room content, and expects clauses like these to be standard by 2027. In-house teams share the worry: in an Association of Corporate Counsel global survey , in-house counsel ranked generative AI tools as the technology most likely to compromise legal privilege. AI governance is fast becoming a reason to win, or lose, a mandate.

Clients pay for insight, not output

One of the sharpest points some customers made is about value. When AI can produce the diligence summary, the summary stops being the thing clients value. They know it took hours, not weeks. What they'll pay for is what comes next. As one valuations partner puts it: "AI is very good at saying more with more. The real value is increasingly in saying more with less."

What does that red flag mean for price? Should it change the structure, the warranty package, the decision to proceed at all? Which of forty findings actually matters? That interpretation, they told us, is where advisors now earn their fee.

The memo is table stakes. The judgement is the product.

The commercial squeeze is already here

Some customers also see a price tag attached to that shift. Clients who know the work is faster may expect to pay less for it, and they're starting to expect it sooner.

The numbers explain why. Professionals surveyed for Thomson Reuters' Future of Professionals Report 2025 expect AI to save them five hours a week within a year, worth around US$19,000 per person annually. In a business that bills by the hour, that time has to go somewhere.

Fees for process-heavy work are under pressure, and models built on billable hours are the most exposed. So is the traditional leverage pyramid, where junior hours on document review subsidise senior time on strategy. Turnaround expectations are resetting too. If the analysis takes an afternoon, no client will wait a fortnight for the report.

Those who raised it don't describe this as comfortable, and they don't expect it to hit every firm equally. Those that keep pricing on effort will be undercut. Those that price on outcomes, and can show the insight behind them, have a margin worth protecting.

What this means for dealmakers

Drawing on these conversations, the shift asks different things of each side of the table.

For advisors and professional services firms

• Reinvest the hours AI saves in client-facing judgement, not in more output.
• Give junior teams AI for the first pass, so their time goes on the answers clients notice rather than sifting folders.
• Find the gaps before the mandate. A sharper read of a prospect's documents changes the pitch conversation.
• Build verification into every workflow, and be able to show clients how it works.
• Have an answer ready on AI governance before a client asks for one.
• Rethink pricing before clients rethink it for you.

For CEOs, CFOs and corporate development teams

• Expect faster diligence, and ask your advisors for it.
• Get your own house in order. A buyer's AI will find the gaps in your data room, so find them first, while fixing them can still protect value.
• Keep control of every deal room, document, user and question in one place, so the process doesn't depend on who's in the office.
• Scrutinise how your advisors use AI on your confidential information, and write it into the engagement.

The machine reads. The dealmaker decides

If there's one thread running through these conversations, it's this: dealmaking in the AI era rewards a clear division of labour. Let the technology do the reading, the sorting and the first pass. Keep people on the reasoning, the relationships and the risk.

It's also the division of labour Ansarada is designed around: secure dealmaking infrastructure with AI inside the data room, giving deal teams clarity across every document and control over every call that matters.

That design answers the caveats our customers raised. Ask AiDA works inside Ansarada's ISO 27001-certified data room, so confidential material never has to go into an external AI tool. Room content isn't used to train AI models , answers follow existing room permissions, and responses come with references back to the source documents, so the human check stays quick. AI takes the first pass. Your team keeps the judgement.