The bottleneck isn't the analysis, it's the data collection
Anyone who builds a comparable analysis knows the problem: the analyst's value lies in judging the comparability of the peer set, in normalizations, and in reading the multiples. Yet most of the time is spent before any of that, sourcing the raw numbers. Open the consolidated statements of an issuer listed on Euronext Milan, hunt for adjusted EBITDA in the annual financial report, find net financial position buried in the notes, reconcile market capitalization to the correct date, then repeat for twelve comparables. These are hours of low-value work where every copy-paste introduces a risk of error.
The result is a comps table that ages quickly and that no one really wants to refresh ahead of a closing. The promise of vertical AI for finance is not to replace the analyst's judgment: it is to remove the data-collection time so that judgment can be applied to fresh, traceable data.
How Amaril builds the peer set and reads the filings
Amaril starts from the universe of issuers listed on Euronext, Borsa Italiana, and Deutsche Börse and from their filed documents: annual and half-year financial reports, IFRS and OIC statements, earnings releases, prospectuses. Once the peer set is defined, the engine does more than keyword search. It interprets the structure of the document, distinguishes reported EBITDA from adjusted EBITDA, locates net financial position across balance-sheet lines and explanatory notes, and reconstructs Enterprise Value by combining market capitalization with net debt.
From there it computes the multiples the work requires: EV/EBITDA, EV/EBIT, EV/Sales, P/E, and the relevant sector KPIs, from organic growth to margins, from capex-to-sales to cash conversion. For a bank or an insurer the set changes, and the system knows it: for an issuer subject to Solvency II it makes sense to reason in terms of Solvency ratio and combined ratio, for a credit institution in terms of CET1 and cost-to-income. Every number lands in its cell of the comparables table, ready for the comparison.
What separates a finance-grade tool from a generalist chatbot is discipline on the data: values are expressed in the reporting currency, reconciled to the correct reference date, and tagged with the fiscal year. A comparable built by mixing different time perimeters isn't an annoyance: it is an error that propagates all the way to the price.
Source citation: every multiple is verifiable
In finance, a number without a source does not exist. This is the non-negotiable principle Amaril is built around. Every figure in the comps table, from EBITDA to net debt to the final multiple, carries a precise reference to the document, page, and line item from which it was extracted. The analyst does not have to trust: they click and verify the row of the consolidated statements that generated that figure.
This solves the problem that kept generalist models out of critical workflows: hallucination. Amaril is designed not to invent values. If a data point is not present in the filing, the system flags it as missing rather than producing a plausible, untraceable estimate. In an investment committee memo, a credit memo, or a due diligence exercise, this difference separates a usable tool from an operational risk. Source citation is not a peripheral feature: it is what makes the comparable defensible before an investment committee, an auditor, or a regulator.
There is also an immediate practical benefit: updating. When a comparable publishes its new half-year report, Amaril re-reads the document and refreshes the multiples while preserving traceability. The table stops being a snapshot that ages and becomes a live view of the peer set.
EU cloud, GDPR, and on-premise: trust as a precondition
No M&A team, private equity fund, or credit desk puts sensitive data and data-room material on infrastructure whose jurisdiction it does not control. Amaril, developed by Hoplo S.r.l., runs on EU cloud with end-to-end encryption, zero retention, and full GDPR compliance, with an on-premise option for those who must keep data within their own perimeter. For firms subject to MiFID II, MAR, AIFMD, or DORA, data sovereignty and operational resilience are not a technical detail: they are compliance requirements that decide whether a tool goes into production or stays an experiment.
Takeaway
Comparable analysis was never the real obstacle: the manual data collection that precedes it was. Amaril moves the analyst from copy-paste to judgment, reading the filings of issuers listed on Euronext and Borsa Italiana, extracting EV/EBITDA, multiples, and KPIs, and populating the comparables table in minutes, with every number anchored to its source. No hallucinations, verifiable citation, EU cloud and GDPR: the three pillars that turn AI from a curiosity into infrastructure you can rest a valuation on. The time saved is not the goal: the goal is to walk into committee with comps that are defensible, current, and traceable.
