Rusty Wiley, CEO and President of Datasite – Interview Series

Rusty Wiley, CEO and President of Datasite, has led the company since 2014, overseeing its transformation from Merrill Corporation, a business focused largely on financial communications and print-related services, into a global software-as-a-service (SaaS) company focused on mergers and acquisitions (M&A) and strategic transactions. Prior to joining Datasite, Wiley spent more than seven years at IBM, where he served as General Manager of Banking and Financial Markets and as a Managing Partner, building extensive experience across financial services, enterprise technology, and business transformation.

Datasite is a global technology platform designed to support dealmakers across the M&A lifecycle, including deal sourcing, preparation, due diligence, financing, restructuring, and post-transaction processes. Its platform combines secure virtual data rooms with purpose-built artificial intelligence capabilities including automated redaction, semantic search, document summarization, translation, AI-assisted Q&A, and deal origination intelligence. Datasite says its technology supports more than 626,000 dealmakers and over 16,000 transactions annually, providing investment banks, private equity firms, corporations, and law firms with a secure environment for managing complex transactions and sensitive information.

You joined Datasite in 2014 after leading banking and financial markets at IBM and subsequently helped transform Datasite from a financial communications and print business into a global mergers and acquisitions software platform. How did that experience shape your conviction that the next phase of dealmaking will be built around AI agents rather than traditional virtual data rooms?

When I came to Datasite in 2014, the virtual data room was at the center of financial transactions. It had already transformed dealmaking by moving sensitive information out of physical rooms and into a secure, searchable digital environment. What became clear over the next decade was that the data room was also creating the foundation for something much bigger.

As our scale grew, so too did the opportunity. Datasite now supports approximately 55,000 transactions a year, giving us a unique view of how deal work happens. At that volume, we see more than documents. We see workflows, patterns and the thousands of repetitive tasks that surround every transaction. That creates an enormous opportunity to apply AI and automation to help teams move from information to action much faster. In fact, we are already seeing this happen. Recent Datasite research, The New Deal Team, shows that firms that ignore AI will struggle to compete within five years, and most dealmakers say relying only on human decision-making is no longer defensible in complex deals.

That’s why the next phase is about building intelligence and automation on top of that trusted infrastructure. AI can analyze information, surface risks, and opportunities, connect institutional knowledge with live transaction data, and increasingly execute multi-step workflows that today consume significant amounts of a deal team’s time. Agents can continuously monitor information, perform analysis, and move work forward while people remain focused on judgment, negotiation and the decisions that determine outcomes.

Given this, data rooms are the foundation for the next generation of dealmaking. The documents, permissions, security, and collaboration layer remain critical. AI makes that foundation far more valuable because it can understand the context around the information and act on it. We’re already seeing that model extend beyond the data room, with Datasite content flowing securely into AI-powered workflows while existing permissions remain intact.

You have described the future of Datasite as an AI-native operating system for mergers and acquisitions. What distinguishes an operating system capable of executing deal workflows from a virtual data room that simply adds document search, summarization, or a conversational assistant?

Search and summarization are genuinely useful, but they operate on one document at a time, and they only respond to a question somebody thought to ask, which is a constraint when you consider what a transaction looks like from the inside. Any given deal consists of hundreds of interconnected tasks running at once, including diligence requests, management presentations, legal reviews, financing workstreams and approvals; all of which involve different people working under different permissions against different deadlines. Experienced bankers are reconstructing the deal status from email threads and a tracker somebody updated on Tuesday, and by Thursday the data is already outdated.

What an operating system does is hold that state so that it updates connected data in real time, knows which workstream is behind, which request has been sitting unanswered for six days, and what information contradicts what was filed last week. Once the system genuinely understands that it can prepare work before anybody asks for it, flag where a process has stalled and coordinate the next step between parties who may not be communicating directly at that moment.

This is why the conversation has moved so quickly from whether AI belongs in M&A to how it should be integrated responsibly. In our research, 96% of dealmakers are already using or exploring AI in sourcing and screening, and due diligence has become one of the clearest use cases, with half of dealmakers using AI regularly or already embedding it into the process. The point is that the industry is past the starting line, and the advantage will go to firms that can turn AI from a tool into a governed operating layer.

Why will proprietary transaction data be more important to the performance of an M&A agent than access to the most powerful general-purpose large language model? What signals allow an agent to understand how a real transaction is progressing rather than merely interpret individual documents?

Foundation models are extraordinary and they keep improving, but they’re also available to all companies at the same time. Whatever advantage a firm gets from having the best model today has a short shelf life when the competitor down the street gets access simultaneously, or a new model is announced weeks later.

Proprietary context doesn’t become widely available in the same way, and that’s where I think the durable difference and competitive advantage sit. Underneath the contracts and financial statements sit thousands of interactions between buyers, sellers, advisors and management teams and the sequence of those interactions carries real meaning. The order in which documents get uploaded tells you something about how prepared a seller actually is, just as a diligence question that comes back sharper on its third pass tells you where a buyer’s real concern lies, and a workstream where response times have quietly doubled tells you something has gone wrong well before anyone has said so.

That’s the intelligence an agent needs if it’s going to be genuinely useful throughout the M&A process. Reading a document well is one capability, and it’s increasingly a commodity. Recognizing that a deal is drifting three weeks before anybody says it aloud is something else entirely, and it comes from having observed tens of thousands of real transactions rather than from training on public information.

An agent could potentially assemble target profiles, review thousands of documents, identify conflicting disclosures and coordinate diligence workflows. Which of these activities can agents reliably perform today, and which should continue to require explicit human approval?

For many firms, agentic AI is already organizing enormous volumes of information, comparing documents for inconsistencies, surfacing potential risk, drafting diligence summaries, and keeping routine workflows moving, all faster than a team could manage manually. Dealmakers are becoming comfortable with that because these are precisely the areas where AI can remove friction without taking judgment away from people. In our research, 46% of dealmakers said AI already leads on identifying targets and drawing up long and short lists, while 45% said the same about identifying red flags during diligence. More importantly, dealmakers are beginning to see commercial value, not just efficiency. About a quarter of respondents in the Datasite research said AI helped them complete a deal they otherwise would have missed, and two-thirds see AI across the lifecycle as a way to de-risk a transaction.

The picture changes noticeably as you move toward the close. Only 22% of dealmakers say they would follow an AI-generated recommendation on whether to proceed to signing, while 45% believe signing should remain an entirely human decision. Instinct is sound. Every acquisition comes down to confidence. Do you trust this management team enough to back them with capital? Are you comfortable with the risks diligence could not fully resolve? Those questions require judgment, and more importantly, somebody willing to be accountable for the answer. AI should help you reach a better decision faster, but it should not be the decision-maker. That’s why tools grounded in governed AI built for the deal environment that include permissioned content, source-backed answers, workflow context, and human approval where judgment and accountability matter most, are so important.

How do you defend an agentic deal room against risks such as prompt injection, unauthorized data extraction, compromised documents, or actions taken beyond a user’s authority?

The mistake is treating security as a layer added after AI is deployed. In M&A, security and governance must be embedded from the start because the underlying information is highly sensitive and often market moving.

Every transaction contains material that may never become public, from financial performance and intellectual property to customer relationships and strategic plans. Any agent operating in that environment must follow the same governance framework that protects the deal itself. It should never access information beyond a user’s permissions, which means identity and entitlements must be enforced at the data layer, not left to a prompt. Every action should be logged, every recommendation should trace back to a source document, and every automated workflow should inherit the authority of the person who initiated it without exceeding it. That’s why we designed our MCP server security-first. The market is moving in the same direction. In our research, accuracy and security were the AI attributes dealmakers valued most, with security ranking first among respondents in EMEA and APAC. In M&A, trust is not separate from the product experience. It’s what makes the product usable.

Hallucinations are especially dangerous when an AI system is analyzing contracts, financial statements, and regulatory disclosures. How should an M&A agent communicate uncertainty, provide evidence for its conclusions, and recognize when an issue must be escalated to a banker, lawyer, or subject-matter expert?

In M&A, if two documents contradict each other, that contradiction is the finding. If the answer to a question isn’t available based on the documents and data in the room, that gap in information is a finding, and frequently a more consequential one than the answer would have been.

Every AI system used in M&A should be designed to be solely evidence-based, as every conclusion should link back to the underlying document, conflicting information should be clearly surfaced rather than quietly reconciled, and low confidence should be expressed transparently in the output. The industry is already building that discipline, which I find encouraging.

Evidence is where the conversation must start. Accuracy was the single most important attribute dealmakers named in our research, and the most common way firms are building trust is still human review of AI-generated outputs. A confident but unsupported answer is dangerous in a transaction. So is an answer that looks complete but quietly skips over uncertainty. What I find equally important is that a quarter of dealmakers expect poor use of AI to destroy several high-value deals over the next five years. That is a reminder that the risk is not just failing to use AI. It’s using AI without the right controls, evidence, and escalation points. My own way of thinking about it is that AI gives deal professionals a remarkably thorough first pass by doing the research and organizing the evidence, but what that evidence means, and what should happen next, should remain a human judgment call.

You have suggested that smaller, AI-enabled deal teams could compete with much larger traditional teams. How could this change the economics and staffing of transactions, and what happens to the junior roles through which bankers and lawyers traditionally learn the fundamentals of dealmaking?

Every generation of technology changes the composition of work rather than removing the need for people, and I think that is exactly what is happening in M&A. Junior bankers, lawyers and analysts have historically spent thousands of hours reviewing documents, updating trackers, organizing diligence requests and assembling materials. AI should alleviate the 2 am work cycles because it can take on a lot of that heavy lifting. The productivity upside is real. But the more important question is what will firms do with the time they get back.

Junior professionals still need to develop judgment, sit in negotiations, and learn how experienced dealmakers think. If AI absorbs the administrative information-gathering, a second-year associate can participate in substantive work and sit in on the meaningful discussions that result in business-critical decisions far earlier than my generation did. The firms that handle this well will reinvest the time they save in their people rather than simply taking it as margin.

What benchmarks should the industry use to evaluate an M&A agent? Should performance be measured through document-review accuracy, diligence issues discovered, false-positive rates, workflow completion, reductions in deal time or the eventual quality of transaction outcomes?

The operational metrics matter, and we track all of them to ensure any tool we deploy is effective and accurate. However, those measure whether the AI performed its task correctly. Customers want to know whether the deal itself was executed better for having used it.

The questions and benchmarks I find more useful are things like whether the system surfaced a risk that would otherwise have been missed, whether the team was able to evaluate more opportunities than they could have before, whether diligence got shorter without getting thinner, and whether senior people spent more of their hours negotiating and fewer searching for information. We have some evidence on that front already. Oue research shows that 24% of dealmakers say AI has helped them complete a deal they would otherwise have missed, and 66% say using AI across the lifecycle is a crucial way to de-risk a transaction.

Datasite’s H1 2026 report shows global deal kickoffs increasing 31% year over year, led by 52% growth in the Americas, compared with 13% in Europe, the Middle East and Africa and 4% in Asia-Pacific. What is driving this regional divergence, and could access to agentic technology widen or narrow the productivity gap between markets?

The data reflects what a lot of dealmakers have been feeling over the past several months, in that confidence came back more quickly in the Americas, supported by financing conditions and a greater willingness to bring assets to market. Interestingly, APAC posted the slowest kickoff growth at 4%, and yet APAC dealmakers are the most aggressive AI adopters by a clear margin based on our research. They lead every region in AI use from sourcing and screening to due diligence, where 56% use it regularly or have it fully embedded compared with 49% in EMEA and 47% in the Americas. They’re also the most comfortable handing meaningful decision authority to AI. My read is that constraint is driving adoption because where deal flow is harder to come by and teams are running leaner, the incentive to get more out of every professional on the desk becomes stronger.

Of course, technology can’t create transactions. It won’t move interest rates or shorten a regulatory review. But what it can do is raise the productivity of every team by executing a deal, and that capability scales far more quickly than headcount, which is why over time I’d expect AI to narrow the productivity gap between markets rather than widen it.

Datasite’s Insight report also indicates that teams are launching transactions more quickly in several markets, while diligence remains lengthy, including a median of 238 days in Asia-Pacific. Which remaining bottlenecks are best suited to AI agents, and which are fundamentally tied to negotiations, regulation, financing conditions, or human decision-making?

Launching a deal has become significantly faster because firms have better data, more automated processes, and more confidence identifying opportunities early. Diligence is different because it’s where complexity accumulates, and complexity doesn’t compress the way volume does. Diligence still runs a median of 238 days in the APAC region, and in my experience, little of that time is used for analysis. Most of it is coordination like reconciling documents against changing requests, multiple advisors, and shifting priorities.

However, diligence is precisely the area that agents are optimized to support. They can monitor progress across workstreams, identify what’s missing, surface inconsistencies between documents and prepare the follow-up materials before the team even asks. That’s why our respondents rank due diligence as the stage where AI delivers the best return on investment and only 4% of them aren’t using it there at all.

AI, of course, has meaningful limitations. People still need to build trust, weigh the risks and rewards, and decide whether the deal is worth the price. The strongest finding in our research, at least for me, was that dealmakers don’t see AI replacing judgment as much as making judgment more quickly and better informed. While more of the process and administrative work moves to technology, the conversations that only people can have become the most valuable hours in the deal lifecycle.

Thank you for the great interview, readers who wish to learn more should visit Datasite.