By Tamara O’Brien, TMIL’s roving reporter
When even the bros at Google DeepMind, OpenAI and the like issue a statement that ‘Mitigating the risk of extinction from AI should be a global priority’ – reviewing the complexities of corporate reporting regulation might bring to mind angels on pinheads, or deckchairs on Titanics. Certainly, ‘risk of extinction’ trumps most of the risks companies have to worry about and report on.
So what’s the point? If we’re all going to die anyway, why are Claire and all the other good folk here assembled, seizing the chance offered by this government review, to establish good principles for reporting in the age of AI?
I see two reasons. One is the rallying cry of changemakers across the ages: if not us, who? Progress doesn’t happen organically. It needs people prepared to grapple with legislation, inertia and the status quo. And the other is, humanity has always been at risk of extinction, either in our heads or in reality. Yet still, somehow, we rise. Perhaps part of being human means not waiting passively to be overwhelmed by the latest existential threat.
In a change to the usual programme, the first in this special webinar series was hosted by Diana – AI pragmatist, and collaborator on ‘Your Precocious Intern’, the report that started it all. Claire took a seat on the panel of reporting experts, because, as Diana explained, she was there to introduce her proposal for the radical reform of reporting.
This initiative-in-progress has already caught the attention of government. And today’s guest speakers have been influential in its development, either by prompting it (Jonathan from the IFRS standard-setting world), or commenting on it (Company Secretary Ida, bringing the report-preparer’s point of view).
A perfect storm for reporting reform
Claire explained that the convergence of three things – the imminent adoption of global sustainability reporting standards in the UK; her own research into AI in reporting; and the government’s intention to modernise reporting – has given her the opportunity to fulfil a long-held ambition. Which is, to re-structure the annual report so that it fulfils its ultimate purpose of building trust between companies and their stakeholders.
The scale of this ambition can get lost in the minutiae of what a restructure means in practice, so let’s just take a moment to appreciate it.
The AR as we know it was established in the 1990s (with ongoing legislative tinkering since the 1930s). Since the 2010s, companies have responded, fitfully, to various efforts designed to integrate non-financial information into their reporting. Recently these efforts were consolidated by the ISSB into new, voluntary, international sustainability disclosure standards. The UK will be adopting the first of these, IFRS S1 and S2 (as the UK Sustainability Reporting Standards) soon… very soon.
An oversimplification of course, but you can feel the weight and difficulty of what it’s taken to get this far. So rethinking the AR – the canonical work that sits atop this vast and ancient pyramid – was almost unthinkable.
But, as Archimedes nearly said: give people like Claire a lever long enough, and a fulcrum on which to place it, and they shall move the world.
She explained her proposal with the help a diagram devised with sustainability consultant and Falcon Windsor associate Hilary Eastman. It still breaks the AR into two parts, but not the familiar ‘front end’ and ‘back end':
1.Disclosure
This part contains all the required material information and data, subdivided by type (strategy, environment, governance, remuneration etc).
The data will be machine readable, so AI can fill its boots here. With the data crunching aspect of report compilation sorted, humans will have more time in the reporting window for analysis and commentary.
2. Commentary
This section consists of:
Management commentary – narrative which explains performance against strategy that year. Refers to KPIs in the disclosure section only as evidence for claims or assertions.
Board commentary – the Chair’s opinion of how the business performed against strategy, and what the Board itself contributed in terms of governance and oversight.
AI is verboten in the commentary section. It has no business replacing the well-thought-through analysis of the management, or pretending to be the genuine opinion of the Board.
Claire had more to say: on why the purpose of the AR must be codified for different types of company, so they only report on what’s material to them (rather than resorting to the ‘disclose everything’ approach that makes analysis impossible). And on creating a dedicated section of the company website to publish the annual report together with other required disclosures, that are nevertheless not material to the company. Plus anything else the company chooses to report voluntarily, in a clearly marked subsection.
AR evolution: the missing link for integrated reporting?
Safe to say, Jonathan’s a fan – not only of Your Precocious Intern (‘Required reading that, thanks to Claire’s presentation to the IFRS Foundation’s Integrated Reporting and Connectivity Council, has really informed our thinking’), but also of Claire’s initiative in broadening the scope of the government’s consultation to include consideration of AI in reporting.
And he applauded her proposal as a way to increase the value of the narrative report, and shake off its image as the Cinderella of corporate reporting: ‘Claire’s proposal will do that, because it connects the data in the financial statements and sustainability disclosures with the disclosures about the business model and strategy that lead to long-term value creation.’ He also appreciated the care that’s been taken to use existing definitions and principles, including the IFRS’s own definition of materiality. ‘This is a huge boost to companies and preparers, because they’ll already be familiar with these concepts.’
In his view, agreed principles, and a common language with which to talk about them, are vital as the profession becomes ever more specialised. He’s concerned by the increasing silo-thinking he sees in governance, reporting and risk management. ‘It's holding companies back from exploiting opportunities across the business, whether that's the identification of data, or the quality of data that may already exist within the business. We're not getting the benefits of integrated thinking.’
He welcomes generative AI as potentially ‘helping us to see valuable information, unobscured, for investors. Not jumbled up for different users in the same part of the report, where it's difficult to understand who the information is for, and whether it’s actually decision-useful.’
How companies can take back control
Ida summed up the reporter’s paradox: technological and regulatory change are incredibly fast-moving, and yet not fast enough.
‘It’s a constant battle to stay on top of new reporting requirements and the uncertainty created by last-minute changes. It all adds to the time and effort that goes into reporting,’ she said. Besides which, there’s much more of it: ‘In the last ten years, the increase in regulatory requirements, and the demands of different stakeholders, have meant our annual report has pretty much doubled in length.’
For reporters, knocking the ever-expanding mass of information into meaningful shape for investors and other stakeholders is nigh-on impossible. But also imperative, since, as Ida pointed out, robust and stable data are the building blocks of trust. Then there’s the heft of the AR as a kind of company compendium. ‘There’s still the feeling that if it’s not in your annual report, it doesn't count.’
Ida welcomes the government's review and consultation as starting to address some of those practical challenges, while bringing the AR back to its original, trust-building purpose. They switched to a three-report approach a few years ago, driven by a desire to give stakeholders the information they want, in an appropriate format. ‘It feels that these proposals take us a step further on from this point, which is great! And seeing the corporate regulators come together to drive it forward is really positive.’
Ida is also optimistic about the role AI can play in solving the ‘volume’ problem – but as a governance professional, recognises that the devil is in the detail. At which point should information go into report preparation? And where should there be guardrails, so as not to invalidate the solid processes and procedures established over decades? ‘So Claire’s call for companies to adopt responsible AI practices is really important – and we all have a stake in that conversation.’
What AI can and can’t do for us
In answer to the question How do you see the proposed model for AI being adopted?, Jonathan cautioned that, while AI’s potential ‘heavy lifting’ function will deliver benefits, its output very much depends on the quality of data input. AI produces confident statements that can appear qualitative, which adds an extra layer of risk. So the data-gathering process will still require human oversight. He also suggested that AI might help to highlight information that’s decision-useful for investors, and could be useful in managing the AR process itself.
But what excited him most was something quite different. ‘A huge, underrated benefit of using AI is that it can amplify human attributes; releasing people to do things like test the quality of management and governance processes, engage with the board, understand strategy…’ To support this, there’ll need to be changes in how we train and prepare people for the business world. And restore some of the human skills – creativity, judgement, empathy – that we’ve lost, in our increasing reliance on technology to come up with all the answers.
Claire agreed, adding that that’s why Your Precocious Intern places so much emphasis on training. Companies need to be quite clear and directive about its use: ‘you can use AI for this, but not for that.’
What are the benefits and pitfalls of your proposed approach? someone asked. Deftly reframing this, Claire explained that the problem with reporting today is a writing one. ‘The regulators have asked us to write a short story, the commentary, which shows readers what the company is all about – but you must also bung in 50 encyclopaedia entries in specific places! How is that ever going to be coherent or readable?
‘But,’ Claire continued, ‘if you separate this information out, it becomes a sensible task that fulfils its objective. The detailed data needs to be in a structured, comparable format. And the commentary on that data, from the management and the board, will help us decide whether we think they are good stewards of the business.’ In this way, AI will help us provide the accurate information needed for the ‘encyclopaedic’ part, while we safeguard the truths contained in the ‘short story’ part.
The pitfall? The massive change it will take, in the first couple of years, to pull reporting apart and rebuild it differently. ‘But I promise you, in the long term, the gain will be worth it.’
Ida too foresees a period of ‘misalignment’ between what corporates say in light of new regulation and restructuring, and what investors and proxies are expecting to see. Which is why it’s so important that companies are clear about the purpose of the annual report, and who they’re reporting to.
Diana closed by announcing that the next webinar, on 15th January 2026, will be on the thorny question of whether companies should disclose their use of generative AI in reporting… and if so, how? (‘In my model it would be really easy!’ interjected an enthusiastic Claire. Why not sign up now and find out…?)