Your next annual report has a new reader, and it isn't human

Your next annual report has a new reader, and it isn't human

Investors have a new research analyst. It sits in their browser, works for free and distils your entire annual report in seconds. It's Artificial Intelligence (AI), and it's answering questions about your investment case and your free cash flow before anyone opens a single page of your report.

That answer is now part of your IR strategy, whether you wrote it or not.

If your disclosures are buried in image-only PDFs, scattered across a microsite from 2019 and contradicted by a Wikipedia entry, the model improvises. It pulls from Glassdoor, an old press release and a sub-domain nobody has updated since the IPO, then hands it all to a prospective investor with the confidence of a verified source.

The organisations getting ahead of this are treating information governance the way they already treat financial compliance. They're auditing what exists, retiring what is stale and establishing a machine-readable corporate hub that tells AI exactly who they are and what the numbers show. These four steps make all the difference.

Four steps to prepare for AI-ready reporting:

• Structure your data. Publish HTML and XBRL versions of the annual report alongside the PDF, with headings that mean something and tables built as real tables rather than images. If a junior analyst can't pull your year-on-year EBITDA out of a structured table, neither can an AI.
• Set one source of truth per fact. Use the same free cash flow number across the report, the IR site, the press release and the investor deck, because discrepancies are where hallucinations are born.
• Establish a corporate hub. Give your key facts one permanent home, in plain text on pages a crawler can reach, and update it the day anything changes. When a model goes looking, that page should be the most authoritative thing it finds.
• Audit the perimeter. Search for your company the way an outsider would, and see what comes back from old sub-domains, third-party profiles and legacy pages nobody owns any more. If it ranks, it speaks for you.

Where it gets harder

None of this means stripping your report back to grey pages of data. The best reporting still looks like something a person wants to read, with a strong visual identity and a clear structure. You're writing for two audiences now and one document can serve both.

Governance and ESG are where we see most of the trouble. Board composition, related-party transactions, Scope 1 to 3 emissions and diversity progress tend to sit in dense tables and footnotes, often with several years of figures side by side. An AI has no way of knowing which column holds this year's figure. It picks one, reports it as current, and nothing in the wording gives the mistake away. You need to label every figure with its year and whether it's in millions, tonnes or percentages, so there's nothing left to guess at.

Reach, Trust and Truth, the three tests your disclosures have to pass

Reach is whether AI can find your information at all. If it sits behind a login, inside a scanned PDF or on a page search engines never visit, you're invisible at the moment the question gets asked.

Trust decides whose version wins. AI has already met your Wikipedia entry, your Glassdoor reviews and a news story from four years ago. You earn the right to be chosen by keeping your own pages current, consistent and written in plain text rather than locked inside images.

Truth is whether your story survives the retelling. A figure reported in millions in one place and thousands in another, or a growth rate quoted without the period it covers, is how a strong set of results ends up described as something you never claimed.

How we can help

We work with clients across the GCC on 360 AI governance audits, stronger digital presence and reports built to perform for every reader, human or AI. What you get is a clear Reach, Trust and Truth roadmap for AI readiness.

The next reporting cycle is the moment to act. AI is already reading your story, so let's make sure it reads the one you wrote.

We'd love to hear from you. Contact us mena@emperor.works