Today I'm stepping off script to talk about something that's taking over a big chunk of the finance conversation on X: delegating tasks to AI agents. And I'm going to give a controversial take.
After spending months digging into this seriously, I've come away with the feeling that it isn't going to make my analysis better. Here's my reflection, personal, of course:
You could hand the best Claude agent in the world to a novice investor and they still wouldn't decide any better than the famous monkey from A Random Walk Down Wall Street. A hedge fund could build the best automated fundamental analysis agent on the planet, arbitraging spreads by integrating not just quantitative data but qualitative and fundamental information too, and it could probably do it better than you.
So where does the small investor's edge go?
Javier Recuenco explains that human reasoning splits into three main systems, deduction, induction, and abduction, fundamentally distinguished by how much data is available to solve a problem.
Deduction runs on classical syllogisms and propositional logic (if A is B and B is C, then A is C).
Induction works in reverse: you observe an outcome and infer the root causes behind it.
Abduction kicks in on complex problems, where uncertainty is high, there’s no prior reference point, and not all the data exists. It’s the process of taking a handful of scattered clues and forming an educated hypothesis to experiment with and navigate unfamiliar territory.
Deduction and induction are the two systems most exposed to being disrupted by AI. When all the data is on the table, AI already beats humans, and can hand out a serious beating while it's at it. Tasks that run purely on logic and historical data are only going to get better in machine hands, not human ones.
Part of company analysis can be handed off to AI, the mechanical part. But fundamental analysis is mostly abductive: you uncover what the market already knows, you learn the sector, and with only a fraction of all available knowledge, you have to build a hypothesis that will keep getting reshaped by forces and events that aren't written down anywhere and that you have to iterate on as you go. And more often than not, you have to do it against consensus. That's where the edge lives. Right where it's always been: in the ability to sit with uncertainty, and in patience.
It’s not about becoming a Luddite, there's a part of the job I do hand off but there's a part I keep doing by hand on purpose.
Even though some of the work can be fully automated, there's real value in keeping the process handcrafted. You notice details along the way, questions surface, new lines of research open up, ideas connect, things that simply wouldn't occur to you otherwise. That's how narratives get built. And that's what we call "gut feel", the thing that ends up steering the final call. And I think that's exactly how it gets trained.
This reminds me of an example, I believe from Kahneman's Thinking, Fast and Slow, about doctors who could spot disease in their patients just by observation, but couldn't explain how they knew.
Every minute spent feeding that abductive instinct means more to me than getting a rock-solid agent that spits out a hyper-elaborate DCF for any company in a single run.
The people at Horos, the Spanish value shop, have this figured out. Their motto fits me like a glove: Head, heart, handcraft.
Of course, this lends itself to debate. Curious to hear other takes. I'm reading.
Latest on the Value Pond
🔒 Bolloré SE (EPA: BOL) — The denominator paradox. Everyone runs the numerator; the argument is underneath it. Bolloré holds ~60% of its own shares through the family cascade. Count those properly and there's over 80% upside, with the family already squeezing.
🔓 Vusion Group (EPA: VU) — Why I passed. The electronic shelf label story is priced to scale, but back out Walmart's revenue from the warrant triggers and 2027 revenue falls 45%. Ex-Walmart sales have been shrinking over 20% a year, and Carrefour is a tenth of the floor space.
🔓Bayer (ETR: BAYN) — Update. Capital prefers a known, pricey ticket over an uncertain meal. The June 25th Supreme Court ruling on Durnell brings Bayer much closer to a final RoundUp number, while the Pharma division nears its inflection point and insiders are buying again.
🔒 Associated British Foods (LON: ABF) — Full thesis. Primark alone may be worth the entire group’s market cap, with the announced spin-off as the catalyst. Grocery, Ingredients, Sugar and Agriculture come effectively free.
🔓 Semapa (LSI: SEM) — Net cash, listed assets, a 49% discount to NAV, and a controlling family that already tried to take it private once. The signals point to a second attempt.
🔓Wizz Air (LON: WIZZ) — Thesis update. Short-term shocks, long-term opportunity. Fuel crisis and pass-through capacity, Q4 demand signals, and a sale & leaseback model that keeps proving the thesis right. Full dossier + Excel for paid subscribers.
Thanks for reading the Value Pond 🦆. If you enjoyed it, don’t forget to leave a like 👍 subscribe 🖊 and share 🔄 it.
DISCLAIMER: All the information provided in this document is purely informative and does not constitute a buying recommendation (according to Spanish Law Article 63 of Law 24/1988, of July 28, on the Stock Market Regulator, and Article 5.1 of Royal Decree 217/2008, of February 15). DuckPond Value Research is not responsible for the use of this information. Before investing in a real account, it is necessary to have the appropriate training or delegate the task to a duly authorized professional.
Reach us on duckpond@duckpondvr.com