Professional Sceptic
In July 2022, the UK's Financial Reporting Council issued a final decision notice against auditor Grant Thornton ("GT") in relation to its statutory audits of a client in the years 2016 and 2018. The well publicised case is now old news; it centred on what the FRC deemed serious failings by GT in the conduct of the audit. What we remember most vividly was the term used by the FRC in its notice. The notice said that GT "failed to treat with professional scepticism management's assertion that Company A was not a related party". Professional Scepticism; what a great line! Economist Hyman Minsky once used a similar turn of phrase. As part of a paper that he authored in 1994 on Financial Instability, Minsky noted that "loan officers are the designated sceptics of the economy, who nevertheless make their living by accepting risks that they understand."
If Auditors and Loan Officers ought to employ scepticism in order to do their job well, it seems to me that investors should too. Both the FRC and Minsky are really making the point that an appropriate amount of scepticism is an essential tool in discerning reality. Trust, but verify as the saying goes. The relatively young profession of Cyber Security has taken this further and has started to promote so-called 'zero trust' architectures as in, 'No Trust, Only Verify'. We'd all much rather not live in a no-trust world, but when it comes to business, as Andy Grove said, "only the paranoid survive".
Reality check
A reasonable definition of reality is 'what has already happened' – reality is not what someone tells you has happened or what they think is going to happen. Good investors are the type to be always on a quest to find the truth or at least to get as close to verifying the reality of a situation as possible. The fog of war combined with an AI revolution has made this task harder. Finding reality seems to warrant an even bigger dose of scepticism of late.
AI might be both a help and a hindrance to our cause. A hindrance in terms of the higher volume of information that we are required to deal with. Fake news is already a cliché and 'Slop' was the Merriam Webster word of the year in 2025. However, AI should be a help when deployed effectively as it should boost our analytical creativity and amplify our ability to gather, analyse and interrogate information at scale.
At Phoenix and Castelnau, we have spent the last twelve months embracing AI in a spirited bid to augment our workflows and enhance our effectiveness as investors. RAGs, models, tokens and single sources of truth have become part of our vernacular. We are using AI to help us get closer to the truth.
Let's consider some specific areas of investment due diligence that are perhaps worthy of a little more professional scepticism.
Humans & spin doctors
Few areas in investing are as important or as difficult as assessing people. It is especially difficult to assess those who are masters in the art of sales and in communicating their own version of reality. What we are trying to verify here is, behind the spin, do the key decision makers we are partnering with have integrity, do they have character? Will they be honest with us when things go wrong, be candid in their dealings with us and will they act as if they are owners?
It is not practical to mimic serial acquirer Brad Jacobs (founder of XPO) who for his latest roll-up hired a CIA officer to help him assess the honesty of those that he is buying businesses from. Methods may vary but the objective is the same.
Jamie Dimon puts it well when asked what traits he seeks in leaders:
“Character is a sine qua non". Dimon goes on to say "by character I also mean that they tell the truth, the whole truth and nothing but the truth. They don't shave the truth, and they say the same thing to you that they do to you..."I always felt as a matter of principle inside a company if I allow spinning to the public, you are going to spin to me".”
How can we identify this from afar? Well, spin is the opposite of candour. What Dimon is emphasising is honesty and consistency as signals of integrity. Buffett says he applies looks for similar traits. He once recommended a related book on this topic by Laura Rittenhouse ('Between the Lines' – it is in the Phoenix Reading Room here). As Rittenhouse puts it: "Candour is the language of leaders who chose to be trusted." Regarding consistency, common law references an old concept "falsus in uno, falsus in omnibus" (Untruthful in one part, untruthful in all) which is an interesting approach although we note many juries seem to judge it too onerous a benchmark to blindly apply.
Frankly, it is really hard to know how effective AI might prove to be in definitively identifying honesty. We do know that AI's ability to trawl reams of historical evidence of behaviour (via language and video) and solve for consistency seems useful assistance for the task. Equally, assessing language style for candour seems a worthwhile addition to the analyst toolbox.
Numbers & false precision
Accounting is the lingua franca of business so there is no avoiding it. Numbers as a way of portraying reality suggest precision, but it is usually false precision. The big problem is that it is a subjective truth. And it's not just the raw numbers; it's also the calculations on those numbers.
Here is Rittenhouse again:
“We have bought into an illusion that the bottom line is the only thing that matters [in investment analysis]. But where does that come from? Accounting! Where does that come from? Judgment, because accounting rules can be applied in so many different ways. What determines judgement? The company values and how they reflect the ethical attitudes of the CEO”
In other words, financial models tell you more about the modeller than they do about the scenario being modelled. The same can be said for a set of accounts.
By definition, accounts represent the world as intended by the managers and accountants that created them and are also shaped by the culture they work in. Therefore, it makes sense to focus not just on the numbers but on the intention behind them. Often these intentions are exposed subtly via omissions in footnotes, changes in tone or even a single one-liner included by the auditor. Sometimes aggressive work cultures seep through and are reflected in the accounts. Rittenhouse was on to something – her work suggests that analysing accounts can be augmented considerably by language analysis. In the pre-AI era, we used to do this
manually by what we at Phoenix call a 'teardown' analysis – literally a detailed comparison of the language of annual reports versus the prior year. We already have proof of concept that this task looks ideal for automation an AI agent.
Information & signal to noise ratios
My very first job in a former life was to design a manufacturing test system to measure what is called the 'signal to noise ratio' (SNR) for a newly designed semiconductor chip. The chip converted an analogue signal into a digital signal, and the veracity of the converted signal was a key specification to guarantee for customers. In my current job, I wonder what might be the SNR for a daily newspaper or a Wikipedia page? Non-fiction author (and provocateur) NN Taleb, would suggest there is more truth in a novel than in a newspaper! He certainly has a way with words himself!
“Newspapers have officially the right facts but their interpretations are arbitrary. They lie with the right facts; a novelist says the truth with wrong facts”
To go a step further in this critique of newspapers' informational quality we bring in another author, Michael Crichton. Crichton coined an interesting phenomenon as the 'Gel-Mann amnesia effect'.
“The Gell-Mann Amnesia effect is as follows. You open the newspaper to an article on some subject you know well. In Murray's case, physics. In mine, show-business. You read the article and see the journalist has absolutely no understanding of either the facts or the issues. Often, the article is so wrong it actually presents the story backward—reversing cause and effect. I call these the "wet streets cause rain" stories. Paper's full of them. In any case, you read with exasperation or amusement the multiple errors in a story and then turn the page to national or international affairs and read as if the rest of at the newspaper was somehow more accurate about [x or y] than the baloney you just read. You turn the page and forget what you know.”
Crichton's point resonates and is surely extra relevant given the weight attributed to media articles in Large Language Model training model corpuses.
Stuff & getting what you pay for
Is it too philosophical or woolly to question the 'truth' or reality of a product being sold? Surely the physical form is not up for debate as to whether it reflects reality? Yet, an earlier Thought Piece on quality (here) demonstrated this fairly clearly with regard to the demise of engineering quality at Boeing. Or what about the speculation brewing in the luxury goods markets that a portion of 'Italian Made' luxury handbags are actually originating in China? Anyone remember Gerald Ratner?
In that context, product quality seems totally appropriate to critique and verify whether its quality is as good as is being touted. So too services. I read an interesting article recently suggested in passing that "a high-quality audit is a credence good" which got me thinking. A 'credence good' is basically type of product (or service) with qualities that consumers find difficult or impossible to evaluate, even after purchase and use. A colleague offers Volvo as an example where customers pay extra for the safety perception (actually the automotive markets offers many examples like insurance, car hire etc). Do consumers overpay? It's a valid
question for investors to ask. Undoubtedly, some companies take advantage of the information asymmetry of what they are selling with much scope for misleading or at worst, gouging. Can AI help us with these sorts of hypotheses? Our work has shown interesting promise for consumer sentiment analysis, one of many angles that are worth pursuing.
Conclusion
Given the importance of truth, was anyone wondered why Lie Detectors took off but Truth Detectors didn't? Presumably because lies are vastly easier to spot than truths but also lies tend to have more consequence and cost! That's one way for AI to help us: by augmenting experts to identify falsehoods, to avoid danger. As Alex Karp put it recently; "trying to be right is really hard, but you know what is wrong".
This document is issued by Phoenix Asset Management Partners Limited (PAMP), registered office 80-82 Glentham Road London SW13 9JJ, Company number 03514660. Authorised and regulated in the UK by the Financial Conduct Authority. This advertisement is directed only at persons in the UK who are (i) investment professionals as defined in Article 19 of the Financial Services and Markets Act 2000 (Financial Promotion) Order 2005 or (ii) are persons of a kind described in Article 49(2) of the Promotion Order or (iii) are persons to whom this presentation may otherwise lawfully be issued or passed on.
Castelnau Group Limited ("CGL") is a Guernsey closed-ended company listed on the Specialist Fund Segment of the London Stock Exchange. Shares traded on a stock market will fluctuate in accordance with supply and demand and may not reflect the underlying net asset value of the shares. The value of investments and any income from them may go down as well as up and investors may not get back the amount invested. There can be no assurance that CGL's investment objective will be achieved, and investment results may vary substantially over time. Past performance is not a reliable indicator of future performance. This advertisement is for information purposes only and does not constitute an offer or invitation to purchase shares in CGL.
An investment in CGL involves a high degree of risk, including the risk that the entire amount invested may be lost. Buying a concentrated portfolio of both listed and unlisted companies can result in above average volatility. Investors should therefore be open to some risk and volatility. There can be a long lag between us making an investment and its value being reflected, so CGL is intended for investors with a long-term investment horizon. Although investors can sell shares on any day, if your horizon is short-term then our approach may not be a suitable home for your money. CGL is intended for professional clients and "sophisticated" or "High Net Worth" retail investors, those considered as "advanced" but is not suitable for retail investors with only basic knowledge of investments.
The information contained within this document has been obtained from sources believed to be reliable and accurate at the time of issue. Prospective investors should not rely upon this document for tax, accounting or legal advice and should consult their own advisors prior to making any investment. The Prospectus and other regulatory documents can be found at can be found at: [www.castelnauigroup.com](http://www.castelnauigroup.com)