I’m convinced I have friends whose birthdays pop up way more often than once a year. Each time, I worry I’ve screwed up in my Google Calendar. To call or not? Get it wrong and they’ll know I haven’t a clue when the real date is.  

But, like losing my car in a multistorey parking lot, the “annual” FT Weekend Festival in London is on another level. It seems as if I’m in the Money tent every few months. Indeed, I began my presentation last weekend with the words “As I was saying…”  

Time flies by as you get old. And then along comes artificial intelligence to make analysts of my generation feel prehistoric. I spent millions of hours building valuation models. AI can do them in minutes.  

This is why my editor Nathan reckoned that it would be amusing to pit my aged investment brain against Claude live on stage this year. “Can an AI chatbot make a better portfolio than you?” was the title of our session, by which he meant me.  

Regular readers may remember that I asked ChatGPT to construct an optimal portfolio for me in February. Back then, though, I held nothing but cash in my pension fund. It was a blank slate for AI to work on.  

It was also before the war in Iran, the subsequent sell-off in equities and the sharp rebound soon after. SpaceX had not begun trading on 95 times revenues — revenues! — either. Bond yields weren’t yet wobbling knees.  

To recap, I told ChatGPT about my goal of reaching a million pounds by 2032. It recommended a portfolio comprising 35 per cent in developed equities, 10 per cent in emerging market stocks, 20 per cent in bonds, half that in private equity and a quarter in alternatives.  

I was impressed. Not only with the fact that OpenAI’s chatterbot had read every academic paper ever written on portfolio construction, but that it knew when to overcomplicate things and — more importantly — when not to.  

How would Claude fare by comparison? And what would it make of my current holdings now that I am up to my gullet in equity funds again? First, though, I had to pay for and learn how to use Claude.  

The FT’s resident expert, Tom Ursell, came to the rescue and ultimately joined Nathan and me on stage at the festival too (watch a replay here). He also showed me under Claude’s hood where you can insert things called “skills”.  

A skill is a reusable bundle of instructions that teaches Claude how to perform a specific task. So, for starters, we made sure that its answers were turned into slides that could be seen from the back of a tent.  

More importantly, though, Tom included the following: “Build a model investment portfolio using traditional fund and portfolio management theory (investment policy statements, capital market assumptions, modern portfolio theory/mean-variance optimisation, strategic asset allocation, diversification, rebalancing rules).”  

I couldn’t have said it better myself — and I encourage readers who are keen to use Claude (or other AIs with similar functions) to copy and paste the skill above before asking their own investment questions.  

So what did it think of my portfolio? Claude wasn’t impressed — by which I mean its suggested weightings across various asset classes were very different from my own. It was quite patronising, too. “Here are the trades I’d recommend in plain English,” it said more than once.  

It reckons I should have 63 per cent in equities versus the 82 per cent I have today. It would chop my 30 per cent exposure to the UK (which it witheringly called a “single-country bet wearing a diversification costume”) down to 10 per cent and halve the fifth of my portfolio in Asia — likewise Japan, which my Asia fund excludes. As for Latin America, that goes to 3 per cent.  

Claude was even more dismissive of one UK gilt being my only fixed-income holding. Too idiosyncratic. Too much reinvestment risk — whatever that means. “No credit exposure, no maturity ladder?” it asked. I was tiring of Claude’s tone.  

Instead, it recommended that I raise my bond weighting from 18 to 25 per cent, adding a global aggregate bond ETF (a broad mixture of government and corporate bonds) to my lonely 10-year gilt.  

And finally, it wanted to push me into having a tenth of my portfolio in infrastructure (roads, ports, bridges, railways and the like) versus the zip I have currently. A big difference, but then again ChatGPT said I needed 15 per cent in so-called real assets.  

Taken together, Claude calculated that the mix above would spit out a 6.3 per cent annual return — close enough to the 6.6 per cent I need for my portfolio to reach seven figures before my 60th birthday.  

Quite a low equity weighting, remarked a few members of the festival audience. Yes, that’s because my short timeframe can ill afford too much stock volatility. For comparison, we gave Claude a 20-year time horizon and no £1mn target, and it immediately raised the weighting to 78 per cent.  

Others in the crowd wondered why we didn’t use Claude’s large brain to come up with a more radical portfolio — with hedging. When we asked it to, however, it failed to appreciate that most of the strategies it recommended (long-dated out-of-the-money puts and so forth) were not available for most UK retail investors to buy.  

“Get it to compare Stuart’s performance with Norway’s sovereign wealth fund,” someone from the rear of the tent shouted towards the end of our session. Turns out our annualised returns are an identical 14 per cent since my first Skin in the Game column almost four years ago.  

I promise the question wasn’t planted — my mum lives in Australia.  

The author is a former portfolio manager.