Post # 106
November 4, 2026
Claire Bodanis
Claire is delighted to report that AI has finally agreed with her – albeit only on FW’s new model for reporting, rather than the existential and philosophical questions about why human beings will always be superior to chatbots, that have occupied much of our gen AI/reporting debate this past year.
As many of you know, this last month I’ve been absorbed by the Government’s Modernising Corporate Reporting consultation (MCR), and indeed will be running a workshop to gather corporate feedback on the subject when this blog lands in your inbox!
The biggest challenge in this work so far has been to avoid ‘leading the witnesses’. That’s because we at Falcon Windsor are very clear on how we think reporting should be reformed to make the most of this ‘once-in-a-generation opportunity to reset the UK’s corporate reporting framework’: namely, the restructuring of the annual report around a commentary and a set of disclosures on all issues material to investors and creditors, as set out in our own new model for reporting for the age of AI. And nothing we’ve heard so far has changed our minds. But, when gathering corporate feedback, as we’ve been doing to help BIST (Department for Business, Innovation, Science and Technology), we really have to keep our opinions out of it. Challenging indeed, as those who know me well will appreciate!
But this is a blog, which means I can write whatever I like. And I simply had to tell you all about the latest convert to our new model for reporting. Yes, generative AI, in the form of Claude! This may come as a surprise to regular blog readers, who know my scepticism about AI when used for thinking and writing. As you know, I prefer to rely on my own brain – not least because once AI has suggested something, it’s impossible to ‘unsee’ or ‘unthink’ it. So how did this turn of events come about?
It all started because we were commissioned to do an annual report review for a FTSE250 company,* and make practical recommendations on how to improve its accessibility and succinctness. We did the review using our own brains, but once we’d finished I thought I’d see what the chatbots had to say about the same exercise. After all, we can’t just assume we’re better at such things than they are.
I’m pleased to say that quite a lot of what came out demonstrated that my understanding of reporting is much better than Claude’s (or ChatGPT’s, which I also tested). Neither really seemed to get how it works – although no doubt they could both produce better outcomes if I trained them properly. But what amused me greatly was point 5 of Claude’s ‘untrained’ answer to the question ‘How could the structure and content of this annual report be improved to make it more accessible to the educated business reader?’ It responded:
5. Restructure for a “layered depth” reading path
Right now the useful decision-relevant content (results, strategy, risk, outlook) is mixed with statutory/legal boilerplate across 230 pages. I’d explicitly organize into three clearly labeled tiers:
Tier 1 (10-15 pages): the true summary — results, strategy narrative, outlook, key risks. Designed to be read start to finish.
Tier 2 (40-50 pages): segment detail, sustainability, governance narrative — for readers who want more.
Tier 3 (rest): statutory accounts, remuneration policy, legal disclosures — reference material, not narrative.
Right now Tier 1 and Tier 2 content is interleaved, so a reader can’t easily stop at “enough detail for my purposes.”
This sounded startlingly like something approaching our model – apart from Claude’s having a middle, longer narrative section in addition to the summary up front. Although I suppose if you just combine the two you have our model, with the commentary part subdivided – while noting that frankly if you do Tier 1 properly, there really should be no need for another 40 or 50 pages in Tier 2. But then suspicion took over – did Claude somehow ‘know I was me’, so to speak? Had it found my own model as source material, even though I’d logged in with my online shopping email, and specified it should just use the annual report in question? Chatbots are programmed to please, after all.
So I asked it how it had come to that specific recommendation and what sources it had used to reach its conclusion. It responded as follows (it clearly has my penchant for verbosity, despite the clarity and specificity of my questions, so I’ll just give you the headlines):
Clarifying how the earlier analysis was actually derived.
That recommendation didn’t come from any external source — it’s my own analytical judgment, based on two things:
1. Direct observation of this report’s actual structure […] an empirical observation from the document itself, not something I looked up.
2. General reasoning about reader comprehension [...] A report where narrative and reference material are interleaved forces that reader to either read everything or risk missing something, rather than letting them self-select their depth. […]
So to be precise: I didn’t search for or cite any external framework, comparable report, or investor-communications research. It’s my own synthesis of what I could see in the document plus reasoning about what would make it easier to navigate.
It misses the point that of course if everything were ‘interleaved’, people probably wouldn’t read it at all; or, in reading everything, might understand little, because the real meaning would have been obscured by volumes of unnecessary information. It also hasn’t grasped that many of the disclosures in the strategic report are also reference material, not just the stuff nearer the back. But it’s getting there.
Now, as we know, current regulation doesn’t allow us to produce a separation between narrative and reference material – or in our language, commentary plus disclosures; both are mixed up together in the strategic and governance reports. But perhaps, with Claude on our side, as well as many of our lovely human supporters, our new model will see the light as a solution to the MCR’s aim of resetting the UK’s corporate reporting framework!
*Annual report review sales pitch alert! I don’t usually advertise our services in the FW blog, but since I mentioned it – and since so many of you will have to do some rethinking to accommodate the UK SRS – I thought it might be of interest. If your annual report is too long, too cluttered, too convoluted, our AR review can help. It gives you a clear set of practical recommendations for improving structure, content, language and design. By implementing them you can make your report more succinct and compelling, while also ticking the compliance box. And yes, that includes looking ahead to upcoming regulation! Just email Neil – neil@falconwindsor.com – if you’d like to find out more.