Ansarada pushes AI from documents to data with OS Bondi 3.0

Australian virtual data room (VDR) provider Ansarada is betting its next platform release, OS Bondi 3.0, on a single structural change: AI that analyses an entire deal room as one dataset rather than one document at a time. The company, whose platform has supported approximately 60,000 transactions, announced the expansion alongside new automation for due diligence question-and-answer workflows and a beta tool called Review with AiDA.

Managing Director Justin Smith framed the shift bluntly: "In OS Bondi 3.0, we're moving from documents to data, and that's a real big change here."

The stakes are material. Due diligence — the investigative phase of a merger or acquisition in which lawyers, advisers, investors and corporate teams review extensive documentation — is among the most labour-intensive and expensive parts of any transaction. Automating parts of it, without compromising confidentiality or auditability, has become a competitive frontier for deal-technology vendors as general-purpose large language models (AI systems trained on broad text corpora) raise client expectations for speed.

What does OS Bondi 3.0 actually change?

The release builds on AiDA, Ansarada's existing AI assistant, which already lets deal teams search documents, retrieve information and summarise key details inside their secure data room. Smith said demand for that capability is being driven by pace, not novelty.

"What we're really seeing in dealmaking is a rapid pace of change. Teams need to get the right information quickly without spending hours digging around in folders. AiDA is helping them ask questions of not only the documents but data in natural language that can get answers from the room in seconds."

The roadmap for OS Bondi 3.0 includes:

• Intelligent Q&A prioritisation and automated draft responses. The system will identify priority questions submitted by multiple parties, including potential buyers and investors, and prepare draft answers using relevant documents and other material held in the data room.
• Review with AiDA (Beta). A document analysis tool, currently in early beta and available to a limited audience, designed to examine information across multiple documents at once rather than file by file.
• My Documents. A planned personal file area where users can securely share and analyse documents before adding them to the indexed material used in due diligence.
• AiDA access packs. A subscription extension allowing additional authorised participants, including buyers, to purchase AI access, extending the capability to both sell side and buy side of a transaction, subject to permissions.

Smith was explicit that the earlier generation of the product stopped short of the new approach. "What has been available so far is document by document," he said. "This allows you to think about it as a data that you've uploaded into the data room."

Why automate due diligence Q&A?

The Q&A process sits at the centre of the workload problem. Smith described the volume in concrete terms: "The Q&A process is one of the most time-consuming parts of the due diligence process: thousands of questions coming in from external parties and their lawyers and their advisers. If you've ever been in that process and trying to run a business, it's an incredibly stressful time."

AiDA will identify similar questions submitted during due diligence and generate draft responses for human review. OS Bondi 3.0 extends this with prioritisation across questions from multiple parties.

The design keeps a human checkpoint. Reviewers remain responsible for assessing proposed answers, checking supporting evidence and approving information before it reaches external parties. Smith positioned the AI as a drafting engine, not a decision-maker.

"AiDA is doing a lot of the manual work; it's helping draft a lot of those answers, but the reviewer gets to review that," he said.

The intended payoff is time reallocated from searching to judgement. "What really matters in dealmaking, from a buyer's standpoint or corporate leader, is judgement. Getting them more time on making the right judgement, the right calls to move a deal forward, is what AiDA is allowing them to do."

How does cross-document review work?

Review with AiDA targets the manual burden of checking contractual material at scale. Rather than requiring users to open each file individually, the tool analyses the data room's contents as a collective dataset.

Use cases include:

• Reviewing multiple contracts simultaneously.
• Comparing clauses across documents.
• Examining addresses or remuneration details across a collection of files.
• Identifying inconsistencies, missing information and potential issues faster than manual review.

Smith said the capability matters most for discrepancy detection. "That will be a game changer for people looking at inconsistencies or looking at potential issues faster," he said.

For organisations running complex transactions, the tool is intended to surface patterns and relationships across material that would otherwise demand substantial manual cross-referencing.

How does Ansarada handle security and governance?

Ansarada's approach deliberately diverges from running general-purpose large language models outside a secure transaction environment. A virtual data room's value extends beyond storage when transactions involve sensitive commercial information and substantial capital: the platform maintains an auditable record of which documents were accessed, what information was shared and how participants interacted. That record matters when disputes arise over whether information was disclosed appropriately.

Ansarada is extending those protections to its AI tools. The technology operates within the company's own cloud environment and infrastructure, and it respects existing access permissions — differentiating between administrators, guests and buy-side participants. Customers control who can access AI functionality and can revoke access when required, within what the company calls a defensible by design system in which AI interactions remain subject to the governance requirements of the underlying platform.

The training posture is also differentiated. Smith said the models do not learn from the substantive content of customer documents.

"Our large language models aren't learning from the actual data that's in the documentation that's in the data room," he said. "It's learning from the critical clicks and the trends that happen, not by the actual data."

Will smaller businesses get access?

Over the next 12 months, Smith identified proactive assistance as the key development area. Ansarada's experience across approximately 60,000 transactions, he argued, gives the platform a base of common patterns, risks and stage-specific information needs.

"Proactive AiDA can help guide you along like an agent: these are the things that you should be doing at this stage in the deal. Here's what I've highlighted. Here's your risks. Here's your pieces of material information that you may be missing," Smith said.

Rather than waiting for user questions, a future AiDA could proactively highlight missing information, flag risks, suggest actions based on deal stage, identify documents a buyer might request, and show which investors or buyers are engaging most actively with the transaction so teams can prioritise attention.

Smith drew a firm boundary around advice. "We're not replacing advisers. We're not giving advice," he said. "We're just talking about the data room and where you are in your journey to help you do that, and highlighting insights that could help you get a better outcome."

The longer-term opportunity, he said, is giving smaller businesses and first-time sellers access to the same secure technology and transaction insights available to larger organisations.

"Every release we're doing is building AiDA more deeply into that operating system that is Ansarada's OS. It's becoming really part of their team," Smith said.

The benchmark for OS Bondi 3.0 will be whether it reduces the administrative burden of complex transactions without compromising the confidentiality, accountability and human judgement those deals require.