Due diligence 2.0: how AI agents transform data room screening

During a company acquisition or merger, due diligence is a crucial investigation for every party involved. AI agents transform this process by rapidly scanning large data rooms for hidden risks. The lawyer always maintains control and verifies the output according to current NOvA guidelines of late 2026.

In short

• AI agents search enormous data rooms in a fraction of the time.
• You immediately identify hidden risks and non-standard contract terms.
• NOvA always requires manual verification of the generated AI output.
• Transparency regarding the use of AI is mandatory under the European AI Act.

What does due diligence with legal AI entail?

During an acquisition, parties share numerous business documents [2]. You typically find these files in a strictly secured data room. Previously, lawyers manually read through every single document. This consumed a lot of time and increased the chance of human error with large volumes.

With the advent of legal AI, this working method changes fundamentally. An AI agent can analyze and structure thousands of pages in a short timeframe. You give the agent a specific assignment. For example, you search for specific exclusion clauses in hundreds of contracts simultaneously.

The AI agent highlights relevant clauses directly for you in an overview. The system also flags deviations from your own standard clauses. As a result, you start your legal analysis with a clear dossier. You spend less time on manual searching and more time on legal interpretation.

For lawyers and civil-law notaries, this efficiency is essential. It enables you to advise faster and sharper on the risks of a business acquisition.

Which documents does an AI agent analyze during an acquisition?

During a due diligence investigation, you analyze the complete legal administration of an enterprise. An AI agent excels at comprehending this unstructured data.

Consider employment contracts, management agreements, and personnel files first. You look for exceptional arrangements, such as a heavy non-compete clause, golden handshakes, or highly divergent pension schemes. The agent reads all documents simultaneously and distills these provisions flawlessly into a clear dashboard.

Supplier contracts, licensing agreements, and rental contracts are also crucial in this process. You always want to know if 'change of control' provisions are present. Such provisions can mean that an important contract automatically ends upon an acquisition. An AI agent tracks down these specific clauses at lightning speed in the data room.

Additionally, the agent analyzes specific legal documents, such as a settlement agreement from a recent dismissal or ongoing litigation files. By automatically categorizing these documents and providing a summary, you immediately gain insight into potential disputes. This significantly accelerates the preparatory phase of your investigation.

What does the screening look like in practice?

To understand the impact of AI agents on due diligence, a comparison with the traditional working method is clarifying. A targeted use of technology shifts the lawyer's focus from searching to analyzing.

This table shows that the lawyer's role does not disappear. The nature of the work simply shifts to a higher, more strategic level.

Why do NOvA guidelines require human verification?

In the second half of 2026, the NOvA guideline for AI dictates your duty of investigation. The Dutch Bar Association sets clear limits on AI use [1]. The core message is that you, as a professional, always remain ultimately responsible.

You may never blindly adopt the outcomes of an AI agent or paste them directly into your report. The AI helps you collect, filter, and structure data from the data room. However, you must thoroughly check the facts, quotes, and legal context yourself afterward. This touches directly on your fundamental professional responsibility.

If an AI agent flags a potential risk, you always test this risk against the original text. You check whether the clause in question actually appears in the contract in the stated wording. You also weigh independently whether this is legally relevant for your client's specific transaction.

This manual verification protects your client against incorrect assumptions and hallucinations by the AI model. It ensures that your advice continues to meet the high quality standards of the legal profession.

How do you prevent data leaks during screening?

Confidentiality during a business acquisition is absolute. The premature or unintended leak of a possible merger immediately has major strategic, legal, and financial consequences. Therefore, managing data flows during due diligence is highly critical.

NOvA explicitly warns against the use of public or open AI models [1]. If you paste confidential contracts into such a consumer model, you are effectively sharing trade secrets with the tool's provider. This is an immediate and direct violation of professional secrecy.

A secure screening therefore always requires a shielded business environment. The sensitive data from the data room must never leave your secure office environment uncontrolled. You therefore connect the AI agent exclusively via strictly secured connections and closed systems.

This principle is called 'privacy by design'. You only share the absolutely necessary information with the AI system. Furthermore, you must be able to guarantee contractually that the inputted data is never used to train the provider's external models.

Which steps do you follow for a secure AI screening?

A successful and safe implementation of AI agents requires thoughtful preparation. Follow these concrete steps to execute the due diligence investigation efficiently and according to the rules:

• Determine the specific risk areas and search terms for your due diligence in advance.
• Ensure a shielded and secure environment without access to public AI models.
• Set clear, specific, and defined search queries for the AI agent.
• Have the agent clearly and systematically categorize all non-standard contract terms.
• Assign an experienced legal specialist for the manual verification of the generated output.
• Document internally exactly which AI tools you used during the entire investigation.
• Record all verified findings with a clear and traceable source reference to the original document.

How it works at Prudai

Deploying AI in complex legal trajectories requires a reliable and structured approach. Workflows in LEO and LEO Tax follow a fixed pattern: intake, plan, approval by a human, parallel specialists, a counter-role (counterparty, inspector, or supervisor), reviewer, editor, deliverable. LEO has 8 professional roles: lawyer, corporate lawyer, civil-law notary, tax specialist, mediator, bailiff, director, patent attorney.

Writing tools in LEO always request approval first via an approval card. Our method relies heavily on one fact in one place, information alignment, and source referencing. This aligns seamlessly with the strict requirements surrounding verification and control in the current legal profession. You maintain full control over every step in the process.

For the specialist

The use of AI in due diligence investigations directly relates to the European AI Act (Regulation (EU) 2024/1689) [3]. This legislation sets strict transparency and risk management requirements for deploying AI. You must be able to clarify internally and to your client when and how you use AI systems for analyzing legal texts.

Additionally, the tightened NOvA guidelines regarding confidentiality and data security play a major role in 2026. Under no circumstances may you enter confidential client data into public models. This constitutes a direct and unacceptable risk to legal professional secrecy. A professional data room screening therefore always requires a fully closed IT architecture with clear data boundaries.

The professional verification duty also demands a conclusive audit trail of your work. You must be able to demonstrate in the file how an AI proposal or signal came about. You also accurately record who manually checked, weighed, and approved the proposal. Read more about the NOvA guidelines for AI in the legal profession for further depth into these indispensable frameworks.

Information security is also crucial for law firms. Prudai is certified to ISO/IEC 27001:2022 and NEN 7510-1:2024. The certificates (NL 3143.1.1 and NL 3144.1.1) were issued by Brand Compliance, accredited by the Dutch Accreditation Council (RvA C 548), and are valid until 29 September 2029.

Frequently asked questions

Can an AI agent assess a data room completely independently?

No, that is absolutely not possible and also not legally permitted. AI functions as a tool that identifies patterns and documents possible deviations. You must always critically assess, verify, and legally interpret the output yourself. Only then do you guarantee the required legal accuracy for your client.

Does AI help with checking a settlement agreement?

Yes, an AI agent can rapidly compare the specific clauses in a settlement agreement with your own firm standards or earlier agreements. Any deviations or missing articles light up directly in your overview. You then decide yourself, using your legal expertise, whether the deviation poses an actual risk.

How do you stay compliant with regulations regarding AI use?

For compliance, you always follow the current NOvA recommendations for your profession and the frameworks of the AI Act. Exclusively use secured and closed AI systems. Always verify the found facts manually using the original source document. Finally, ensure a clear and traceable source reference in your final due diligence report.

Sources

• [1] Recommendations AI in the legal profession (NOvA), 2026.
• [2] Legal matters when selling a business (Ondernemersplein).
• [3] Regulation (EU) 2024/1689 (AI Act), applicable from August 2026.

Would you like to know how AI contract analysis with LEO strengthens and accelerates your due diligence? View the solutions on the LEO website.

Updated on 12 October 2026