Optimising reporting for AI /
Introduction
At Reportl we work with teams preparing corporate reports, who recognise the dramatic changes brought by AI, and our conversations with them have inspired this white paper.
Processes using print/PDF-first software present fundamental problems for AI searches.
The rise of AI
There’s one topic that comes up in every one of our conversations. Unsurprisingly, it’s the rise and transformative power of AI. How people access and use information has already dramatically changed, with Large Language Models (LLMs), Bots and Agentic AI frequently doing the heavy lifting.
Amongst company executives, we encounter a repeated frustration – that for all the effort and resource that goes into producing an annual report or a sustainability report, its content simply isn’t reaching readers as it should. When an AI or Google search for company information uses a third-party website (i.e. not controlled by the company) someone else is, in effect, telling the company’s story and isn’t necessarily doing it accurately. Naturally, there is a desire to change this.
Since ChatGPT was launched in late 2022, much has been learned about the best way to publish content for AI to access and analyse and also the causes of AI’s biggest flaw: confidently-asserted errors euphemistically referred to as “hallucinations”.
The rise of AI tools is already affecting decisions relating to corporate reporting processes and this seems set to continue in the coming years. Decisions will likely be accelerated by AI’s well-known problems with inaccurate and inaccessible data sources.
The root cause of AI’s problems providing good data is that long-standing corporate report creation processes use print-first/PDF software. Their output presents fundamental problems for AI searches of corporate information. When a reporting design process can only deliver PDF or print-based formats, these outputs are insufficiently structured for AI to access.
'Digital-first’ simply describes the process of creating a corporate report directly from digital code - also known as ‘native HTML'.
What does “Digital-first” reporting mean?
For the avoidance of doubt, ‘Digital-first’ doesn’t mean online-only reporting. It doesn’t mean a landscape PDF, nor does it mean an additional summary of reporting content on a corporate website. Digital-first reports are not ‘converted’ from an intermediate PDF document.
‘Digital-first’ simply describes the process of creating a corporate report directly from digital code and is often referred to as ‘native HTML’. It means all data is digital at source, never converted to digital from another source.
The UK’s financial reporting regulator, the FRC, provided helpful guidance on the benefits of native HTML to fulfil UK digital reporting requirements, explaining how only a new generation of tools can deliver responsive, interactive, accessible reports optimised for search engines and enabling better web analytics.
In ‘Digital-first’, the structured code that underpins the web, HTML, is used as the source of all required formats (online, PDF, iXBRL and print). This simple shift in the design process delivers a suite of accessible corporate reporting from a single source of content that can be used to comply with a range of requirements.
This is usually delivered by the same specialist design and production team used for InDesign reports, but instead they use purpose-built digital reporting software.
Perhaps surprisingly, relatively little about the preparer’s experience changes. The proofing and sign-off process is still done using PDF output because, for now at least, PDF remains the best and most popular format for this. In this way, most existing control processes do not need to be changed when switching to ‘digital-first’ and can be readily enhanced.
The aim of this white paper
When talking about digital and AI, there is an ever-present risk of complexity, confusion and people speaking at cross-purposes. To address this we aimed to write this white paper to explain why an annual report prepared in HTML format delivers better AI results, providing signposts to relevant evidence – research studies, articles and surveys.
We have placed the evidence within relevant context so that an issuer’s corporate reporting team members, each bringing their own specialisms, can navigate to topics of particular interest within the white paper.
Specific topics include the demand for higher quality digital data expressed by many investment professionals, how the special case of XBRL fits into the overall picture and how reporting quality can be enhanced by making a change in software. We end the white paper with some predictions.
There is, of course, a commercial reason behind Reportl publishing this white paper. Reportl is a purpose-built platform which offers digital-first reporting optimised for AI. But we won’t try to persuade you of the merits of selecting our software specifically. There are other platforms available that may be better suited to your specific needs.
Nor will we suggest that PDFs are no longer relevant. Going digital-first still means producing a PDF version as part of the suite of reporting documents, and our analytics data show PDFs are still liked by many stakeholders.