Skip to main content
Back to Insights
AI in Professional Services

How AI Fits Into a Statutory Valuation: What the Software Does and What the Valuer Does

Published March 2026Updated 20 November 20264 min read

An earlier version of this article claimed a specific reduction in turnaround time. The claim has been removed: we have no measured figure to publish. This version describes, step by step, what Evaaluate does in a statutory valuation and what stays with the valuer.

Step 1: The Engagement and Its Legal Basis

The valuer creates an engagement by purpose: a preferential allotment, an inbound foreign investment, an income-tax net-asset value, an insolvency valuation, a scheme, an employee stock option plan and so on. From the purpose and the valuation date the software looks up, in its catalogue, the governing provision, who may certify and which methods the provision permits, and it refuses a method the provision does not allow. Every catalogue entry is disclosed as not yet reviewed by a qualified professional.

Step 2: Financial Statements

Statements arrive as an Excel upload, a PDF upload with text extraction, or JSON sent through the API. A text-based PDF is read directly. A scanned PDF needs an OCR provider to be configured on the deployment and is otherwise refused. Extraction returns a confidence score and warnings, and nothing is treated as a figure of record until the valuer has locked the inputs.

Step 3: The Gate 1 Workspace

Every input carries its origin (entered, fetched, derived or placeholder), its source and the date it describes. Industry beta, equity risk premium, country risk premium and industry multiples come from Damodaran's published datasets for the valuation date. The risk-free rate is not fetched: the valuer enters it with a citation. The forecast is entered as drivers such as revenue growth, margin, capital expenditure and working capital. Comparable companies are added by hand, pasted from an export of a database the firm already uses, or fetched from a market-data provider if a key has been configured.

Placeholders must be confirmed by the valuer, with a source, before the inputs can be locked. Locking is refused while any input is an unsourced placeholder, and it requires a written rationale.

Step 4: Research Pack and Assistant

The software can prepare a research pack: sourced facts and drafted context sections. A section is printed in the report only after the valuer approves it. An assistant can gather data, draft the pack, propose a screen of outlying peers, propose a repair to a placeholder forecast and propose method weights. It can never confirm, approve, lock or sign: the server refuses those acts from anything but a signed-in professional.

Step 5: Computation

A deterministic engine computes discounted cash flow on free cash flow to the firm, net asset value (including the income-tax net-asset formula) and a comparable-company value. No language model performs arithmetic. The same locked inputs give the same result, and each run is stored with a hash of its inputs.

Step 6: Gate 2, the Value Conclusion

The valuer sets the weight of each method, the discount for lack of marketability and any control premium, and approves the conclusion. Where the law prescribes the method, as with the income-tax net-asset formula, the software refuses a discount or premium on top of it.

Step 7: Report and Workings

The report is produced in Word and in PDF, with an Excel workbook in which every computed line is a live formula over the inputs, so the valuer can recompute the result independently. The valuer reads the report, enters the UDIN where one applies, signs and approves dispatch.

What the Software Does Not Do

It does not generate a UDIN or connect to the ICAI portal. It does not apply a digital signature. It has no template laid out to the IBBI report format. It does not compute free cash flow to equity, purchase-price allocations or option-pricing values in the report flow, and it does not fetch a government-security yield. It does not certify: the professional who signs the report does.

Why the Split Is Drawn Here

The split follows who is accountable. Reading a statement, looking up a published figure and multiplying it through a model are tasks a machine can do and a person can check. Choosing a forecast, a peer set, a discount rate and a conclusion are acts the law attaches to a named professional, so the software is built to stop and wait for them.