Opinions expressed by Digital Journal contributors are their own.
Three AI capabilities have moved past the pitch deck and into active deal workflows. Here’s what automated redaction, deal analytics, and AI-powered translation actually do in practice and why governance is the question deal teams should be asking now.
The “AI-powered” label has spread across M&A technology marketing faster than most products have earned it. Pitch decks, product pages, and analyst briefings all use the phrase, often without drawing any distinction between what’s live and what’s in development. That gap is relevant for deal teams making platform decisions under real-time pressure.
Three capabilities have cleared that bar: automated redaction, deal analytics, and AI-powered translation. Each addresses a specific operational bottleneck in the due diligence process, and each has moved from feature list into active deal workflows. The mechanics of how these tools work and how they’re built determine whether they hold up in a live auction environment.
Datasite, whose platform processes 55,000+ deals annually for a client base of more than 626,000 users, has purpose-built these capabilities for M&A workflows specifically. That scale matters: AI tools exercised across billions of documents carry a different level of validation than general-purpose tools applied to deal processes.
When redaction has to be right the first time
In a virtual data room, redaction isn’t a formatting exercise. When a sell-side team needs to remove sensitive information from documents shared with a specific buyer group, that redaction has to be permanent for the viewer and reversible internally. Manual redaction across thousands of documents is slow and error-prone. AI-powered redaction handles the volume, but the accuracy bar is high. A missed redaction on a confidential document during a competitive auction can create significantly larger issues down the line.
The technical requirement is specific: redaction tools need to identify sensitive content at the document level without relying on keyword lists alone. To recognize patterns in legal language, financial data, and personally identifiable information, a system needs to be trained on documents that look like M&A materials. EY notes that AI can support sell-side teams by automatically organizing uploaded documents and proposing redactions where sensitive content is detected, though human review remains essential at every stage of a live transaction.
Reading buyer engagement in real time
Real-time tracking of buyer engagement has changed how sell-side teams manage processes by documenting each bidder’s reviews, the time they’ve spent, and the focus of their Q&A activity. This data serves as operational intelligence, influencing the allocation of attention among a buyer group and the timing of subsequent steps.
The practical value comes from the specificity of what’s being tracked. A buyer spending significant time on environmental compliance documents signals a different concern than one focused on customer contracts. When deal teams can see those patterns develop, they can shape the Q&A process and prepare targeted management responses without waiting for buyers to surface issues directly. Aggregated and anonymized engagement data across large deal volumes also gives experienced teams a reliable baseline for what typical buyer activity looks like at each stage of a competitive auction.
Clearing the language bottleneck in cross-border deals
Cross-border deals have always been slowed by the need to translate due diligence materials across languages. AI translation that supports multiple languages is removing that bottleneck. It supplements human review by handling the first pass at a speed that manual translation can’t match.
The operational shift is meaningful in auction timelines where even a few days of delay can affect buyer engagement. Datasite’s platform offers support in 20+ languages with 24/7/365 assistance, reducing the coordination overhead that cross-border deal teams have historically absorbed. Legal and financial terminology still requires expert review, but offloading the initial volume to AI translation allows that expert time to focus on nuance rather than throughput.
“There’s a gap between what vendors call ‘AI-powered’ and what actually works in a live deal environment,” said Matt Summers, Executive Vice President, Head of Product at Datasite. “It’s one thing for features to exist on a product page and another for them to actually perform well when you’re managing an auction with 30 buyers and a two-week timeline. That’s where accuracy in redaction, reliability in analytics, and speed in translation actually matter.”
Governance is where the next decision actually lives
What’s coming next is agentic AI, where systems can surface information and take actions within deal workflows. Automated responses to common Q&A questions, intelligent document routing, and predictive deal analytics are all in active development across the industry. Bain & Company’s 2025 Global M&A Report found that 21% of deal practitioners are already using generative AI in their processes, up from 16% the year prior, and the firm expects every step of the M&A process to be enabled by generative AI within five years.
For enterprise deal teams, the governance question is what determines which platforms earn procurement approval before any of those capabilities ship:
- How is the AI built?
- Is client data isolated from model training?
- Can AI features be disabled by request?
ISO 42001, the international standard for AI management systems, gives procurement teams an independent way to verify those answers rather than relying on vendor claims. Certified platforms with ISO/IEC 42001 and SOC 2 Type II attestation have created the necessary compliance systems for enterprise M&A. That certification is the evaluation that shapes long-term platform decisions, and it’s a conversation happening now, well before the next generation of AI capabilities reaches production.
FAQ
What is AI-powered redaction in M&A due diligence?
AI-powered redaction automatically identifies and masks sensitive content in documents shared through a virtual data room. In an M&A context, the redaction needs to be permanent from the buyer’s perspective while remaining reversible for the sell-side team. Purpose-built systems are trained to recognize the types of sensitive content common to deal documents, including personal data, confidential financial terms, and legally protected information, at a scale and speed that manual review can’t match.
How does deal analytics change the way sell-side teams manage a process?
Deal analytics gives sell-side advisors real-time visibility into how buyers are engaging with documents in the data room. Tracking which documents each party has reviewed, where they’ve spent the most time, and where Q&A questions are concentrated lets teams identify concerns early, prioritize management responses, and make more informed decisions about process timing without waiting for buyers to raise issues directly.
What is ISO 42001, and why does it matter for M&A technology platforms?
ISO 42001 is the international standard for AI management systems. It provides an independently verifiable framework for how organizations develop, deploy, and govern AI capabilities. For procurement teams evaluating M&A platforms, it’s one of the clearest signals that a vendor has built structured controls around how its AI uses client data, and it’s an increasingly relevant filter as agentic AI capabilities move closer to production.
What’s the difference between today’s AI features and agentic AI?
Current AI capabilities in M&A workflows, such as redaction, analytics, and translation, are primarily assistive. They process information and surface it for human review. Agentic AI takes the next step by performing actions within workflows without manual prompts, including drafting Q&A responses, routing documents, and generating deal insights autonomously. Those capabilities are in development, but the governance frameworks to deploy them responsibly in high-stakes transaction environments need to be established first.
