Are the books cooked? Find out in hours, not weeks.

“AI-powered fraud detection” usually means a document in and a summary out, or a clever prompt with a skill bolted on. Transparently.ai is built the other way round: more than 200 models built in house on years of academic and practitioner research run first, as code; generative AI only explains what they found.

  • Every number computed, never generated
  • Every figure with the page it came from

The forensic review most deals never get

A forensic accounting engagement takes weeks and is priced to match, so most private companies' accounts are never examined closely. Upload the accounts and have the examination done in hours: every figure is extracted with its page, 166 indicators are computed and compared with the company's peers, and only then is a cited report written.

  • It readsPDF, scanned PDF, Excel, Word and XBRL, every page
  • You receiveA rating from A+ to F for each fiscal year, the findings ranked by severity, and the page behind each one
  • It costsPay as you go, with free credit to start amounts are read from the app
Fiscal year

No evidence, no rating

The documents are read, every figure is checked, the models run, and only then is a letter given.

  1. Annual report 2025

    Revenue412.6p.61

    Trade receivables31.0p.163

    Cash from operations21.4p.221

    1,284 figures read, each with its page and line

    1Read

    Every page of every document. Each figure keeps the page and the line it was printed on.

    • Revenue412.6✓agrees with the segment note
    • Trade receivables31.0✓agrees with the balance sheet
    • Cash from operations21.4✓agrees with the cash flow statement
    • Other income3.9✗does not add up: left out, reported as missing

    2Check

    A figure must agree with the statements it belongs to. One that does not is left out and reported as missing.

  2. 0of 166 indicators, in code

    Each one compared with 212 listed food producers, 2025

    peersthis company: above 94% of them

    3Compute and compare

    166 indicators, each calculated in code and compared with listed companies in the same industry and year.

    • 2021B+
    • 2022A-
    • 2023B+
    • 2024C-
    • 2025E

    4Rate

    A letter from A+ to F for each fiscal year, with the findings behind it ranked by severity.

Lyonesse Foods Limited, Annual report 2025. Illustrative.

Not a PDF dropped into a chatbot or a clever prompt

Transparently.ai is built on predictive analytics. Everything below runs before a single sentence of the report is written, and none of it is done by language models.

    One of 166 inputs to the 2025 rating E

    Not AI slop

    Written after the number, from the number. A separate check then strikes any sentence the figures do not support, before you see it. Illustrative figures and definitions.

    166 indicators in 14 categories, from more than 200 models. The one above is one of them.

    Don’t take our word for it. Turn to page 163.

    The math raises the findings. Each one is then tied to the page of the documents it rests on, so you can check it against what the company printed.

    Lyonesse Foods Limited, Annual report 2025. Illustrative.

    Strategic report, pages 1 to 58: two findings. Governance, 59 to 104: none. Financial statements, 105 to 118: none. Notes to the accounts, 119 to 296: seven. Shareholder information, 297 to 312: none.

    Built on research you can read

    A study with university researchers of the methodology behind Transparently.ai, now under peer review for journal publication. wording unapproved; counsel to see

    Working paper paper title, from the PDF
    authors and university, from the PDF
    including Hamish Macalister, co-founder

    Recognition

    RegTech100 yearRegTech Awards APAC, finalist yearthe other awards, once counsel confirms we may show them

    Need the whole market, not one company? It is already analysed.

    Everything above is the company review: one company, from the documents you upload. The dataset is sold separately. It is the same analysis, already run on every listed company and recorded point in time, so each year shows what was known then. name on the site to confirm: Point in Time dataset

    • It coversOver 85,000 listed companies, point in time since 1997
    • It is forAsset managers, funds and supervisors who need the same reading across the whole market
    • To get itTalk to us about what you need

    Two services. Start with either.

    The company review

    For one company you are looking at: family offices and private-market investors weighing a deal, lenders and credit teams, and the advisers who run diligence for them.

    Upload its documents and read the review the same day.

    Review a company free
    The listed-market dataset

    For the whole listed market: asset managers, funds and supervisors who need the same reading across every listed company.

    Tell us what you need from the dataset.

    Contact us about the dataset