An M&A data room is a collision of thousands of heterogeneous documents: IFRS and local-GAAP financial statements, half-year reports, commercial and financing contracts, guarantees, term sheets, appraisals, litigation files. The deal team has weeks, sometimes days, to turn that mass into a defensible view of risk and value. The expensive part is not judgment: it is reading. Paging through PDFs, reconciling figures across statements and notes, hunting for the change-of-control clause buried on page 47 of a supply agreement. This is where a finance-vertical AI changes the economics of due diligence, under one non-negotiable condition: every statement must trace back to a precise source.
From a stack of PDFs to questions with cited answers
Amaril's first job is not to generate text, it is to read. The platform ingests the entire data room, normalizes formats, recognizes the structure of a financial statement (balance sheet, income statement, cash flow statement, notes) and indexes every paragraph, table and footnote while keeping the pointer to the document, page and line.
From there the analyst stops searching for files and starts asking questions. "What is net financial debt at year-end, and how does it reconcile with the leverage covenant in the loan agreement?" Amaril pulls the figure from the financial debt note, retrieves the Net Debt definition from the loan agreement (which often differs from the accounting definition) and shows both sources side by side. Every number in the answer is clickable and lands exactly on the spot in the document it came from.
This is the pillar that separates a finance tool from a generalist chatbot: no hallucinations. If a piece of information is not in the documents, Amaril says so instead of inventing it. For a memo headed to an investment committee or a credit committee, the gap between a plausible answer and a verifiable one is the gap between usable and useless.
Extracting numbers, clauses and red flags at scale
The real leverage is running the same question across hundreds of documents at once. Across a contract portfolio, Amaril consistently identifies and tabulates the clauses that shift deal risk:
- •Change of control: which contracts let the counterparty terminate or renegotiate on a change of ownership, with the clause cited.
- •Financial covenants: leverage thresholds, interest cover, capex limits, and their status against the latest reported numbers.
- •Guarantees and indemnities: caps, baskets, survival periods, the scope of representations.
- •Exclusivity, non-compete, MFN, penalties: terms that move valuation and post-deal integration.
On the financial side, the platform builds multi-year time series from filed statements, flags discontinuities (a jump in trade receivables, an accounting-policy change disclosed in the notes, aggressive capitalization of development costs) and links every line back to its origin. Red flags do not arrive as opaque verdicts but as traceable observations the analyst can confirm in seconds by opening the source.
A concrete example: for an issuer listed on Borsa Italiana or Euronext, Amaril can compare the annual financial report against price-sensitive announcements and subsequent filings, consistent with Consob and ESMA transparency obligations and the MAR regime on inside information. The same holds for a German target supervised by BaFin or a French one under AMF: the European regulatory perimeter stays the reference point, not an afterthought.
Comparability, comps and the memo that almost writes itself
Once the data is structured, comparable analysis stops being a manual copy-paste exercise. Amaril extracts multiples and operating metrics across a set of European peers from their own statements and reports, keeps the calculation base explicit and flags where definitions diverge between local GAAP and IFRS, so the comps rest on like-for-like numbers rather than approximations. From there, the draft IC memo or credit memo is born already populated with the right figures and, crucially, with the citations attached: whoever reviews the memo can trace every number without asking for the underlying file.
This sharply cuts the dead time between collection and decision, and moves the analyst to where they are actually needed: interpreting, challenging assumptions, negotiating. The machine reads and cites; the person judges.
Why it matters where the data lives
A data room holds highly sensitive information, often under strict confidentiality undertakings. That is why the architecture is part of the value proposition, not an IT footnote. Amaril runs on EU cloud, with end-to-end encryption and a zero-retention policy: documents do not train third-party models and do not leave the agreed perimeter. For institutions with stricter requirements, banks, asset managers and funds subject to DORA and to EBA and national-regulator expectations on operational resilience and outsourcing, on-premise deployment is available. GDPR compliance and the handling of any personal data inside the contracts are managed by design.
Takeaway
AI does not replace the judgment of the people running due diligence: it removes the bottleneck. Reading an entire data room, extracting numbers, clauses and red flags, and presenting them with the source always attached turns weeks of reading into hours of analysis. The three pillars remain the precondition for any of this to be usable in a European finance context: every answer cited to source, no hallucinations, and data held on EU cloud or on-premise in line with GDPR and the regulatory framework. On that footing, faster due diligence is not a gamble, it is a process you can defend in front of a committee.
