Every number in your valuation, traced to its source.
ValuationOS turns raw financial statements into a formula-driven Excel valuation model. AI reads the documents. A deterministic engine does the math. Your analyst signs off. Nothing leaves your perimeter.
Industrial company, FY2025 (IFRS)
Select a method to trace itIllustrative company and figures.
AI can read a 10-K. It still can't be trusted with the model.
General-purpose AI agents now build spreadsheets on request. Independent tests in 2026 show where they break: the mechanics that make a valuation defensible.
Best AI agent's score on a banker modeling test, against the weakest human analyst.
Wall Street Prep, AI modeling ranking, June 2026Tested agents that correctly handled circular interest, debt schedules and diluted share count.
Wall Street Prep, June 2026Best exact balance-sheet accuracy when models reconcile statements across documents.
FinBalance benchmark, June 2026So we split the job. Language models interpret. A rules-based engine calculates. Every step is logged and reproducible.
From annual reports to a signed-off model.
Upload 10 to 20 reports across periods. Each stage says who does the work, so reviewers know which parts are generated and which are computed.
Extract every statement
PDFs, scans and Excel packs are parsed locally. Tables, notes and footnotes keep a link to their page and line.
Map to one chart of accounts
Line items map to a standard taxonomy. Restatements, IFRS 16 leases and one-offs are flagged for review, never silently adjusted.
Build the linked model
The advisor proposes methods and assumptions. The engine writes forecast, DCF, comps and sensitivity sheets with live formulas.
Review and approve
Every change arrives as a proposal. The analyst accepts or rejects it, and the audit log records who approved what.
The right method for the business in front of you.
The advisor reads the company's profile and data quality, then recommends a primary method, a cross-check and what to avoid. It explains its reasoning. It never sets the final value.
A model your team would have built, in the file they already use.
Output is a real workbook with banker formatting conventions and live formulas. Change an assumption and Excel recalculates as usual. The add-in proposes edits and never overwrites history.
=D4-(D4-D6)*Assumptions!$C$12-D7-D8| USD m | 2024A | 2025A | 2026E | 2027E |
|---|---|---|---|---|
| Revenue | 231.5 | 250.0 | 275.0 | 302.5 |
| Revenue growth | — | 8.0% | 10.0% | 10.0% |
| EBITDA | 45.1 | 50.0 | 55.0 | 60.5 |
| EBITDA margin | 19.5% | 20.0% | 20.0% | 20.0% |
| D&A | 10.8 | 11.4 | 12.1 | 13.0 |
| CapEx | 12.0 | 12.5 | 13.8 | 15.1 |
| Change in NWC | 1.6 | 2.1 | 2.5 | 2.7 |
| FCFF | — | — | 28.0 | 30.8 |
“Add a sensitivity of equity value to WACC and terminal growth.”
- New sheet Sensitivity: WACC 8.0–10.0% × g 1.5–3.0%
- Data table linked to DCF!F24 (equity value)
- Named ranges WACC, TG on Assumptions
- Historical columns 2024A–2025A untouched
- No existing formulas overwritten
Runs where your data already lives.
Choose the mode that fits your security policy. The product, the model library and the audit trail are the same in all four.
Desktop
Independent valuers, small M&A teams, pilots.
- Data
- On the analyst's machine
- AI
- Compact local models
- Network
- Works offline
Private cloud
Boutiques, PE funds, mid-market banks.
- Data
- Single tenant in your Azure or AWS
- AI
- Private model endpoints, zero retention
- Network
- Your VPC only
On-premise
Banks and large funds.
- Data
- Your servers
- AI
- Open-weight models on your GPUs
- Network
- Internal, with SSO
Air-gapped
Sovereign and highly sensitive mandates.
- Data
- Isolated environment
- AI
- Fully local
- Network
- None. Signed offline updates
Zero external calls
No outbound traffic from the secure zone. You define any exceptions.
No training on your data
Models run inference only. Your documents never fine-tune anything.
You hold the storage
Documents, indexes, intermediate data and logs stay in your infrastructure.
Full audit log
Who uploaded what, what the AI proposed, which formulas changed, who approved.
Verifiable install
Network map, independent audit, signed updates and telemetry you can switch off.
Deal-level access
Corporate SSO, permissions per mandate, encrypted temporary files.
Built for statements outside US GAAP.
US public filings already come as structured XBRL. Private companies and most non-US reports do not. That is where manual spreading still eats analyst weeks, and where we start.
One engine, every standard approach.
Method rules follow CFA Institute practice, IVS and the IPEV Valuation Guidelines. Market data comes from your own licensed exports, such as Capital IQ or FactSet, so nothing is resold or re-licensed.
Four modules, one platform.
We ship them in order. The first two cover the full path from documents to a defensible value.
Financial Data Engine
Reports in, standardized history out, with source links, checks and review status for every figure.
Valuation Engine
Method advisor, linked DCF and comps, scenarios and sensitivity, exported as a working .xlsx.
Excel Copilot
Plain-language requests inside Excel. Every edit is shown as a diff before it is applied.
Report Generator
Approved model to valuation memo and pitch pages in Word and PowerPoint, with matching numbers.
We are choosing our first pilot teams.
We are looking for valuation practices, M&A boutiques and PE funds that value private companies and work with IFRS or local-GAAP statements.
- Pilot on your own documentsNo cost
- Deployment inside your perimeterDesktop or on-prem
- Model templates tuned to your house styleYour formats
- Direct line to the founding teamWeekly
- Pilot startQ1 2027