On 5 January 2026, Ben Rogoff and Alastair Unwin, managers of Polar Capital Technology (PCT), published a think piece entitled “2026: The year AI’s impact becomes impossible to ignore”. It looks as though they were right.

AI had been talked about for decades, but it was the release of ChatGPT4 in March 2023 that lit a fire under the stocks associated with it. For example, despite the occasional bump in the road, Nvidia is up almost 7.6x since then.

Similarly, while most of the excitement has been about the vast impact of building the infrastructure to support the growth of AI, agentic AI – AI that goes beyond just responding to prompts and queries and moves on to designing and creating systems that can run autonomously to perform tasks and automate workflows – was largely talked about as something that was coming in the future. However, Anthropic’s Claude Cowork, which launched on 12 January, lowers the bar for users rather than programmers to code, deploy, and run software.

There was no immediate market impact, but this week saw the launch of Claude Cowork tools that help automate legal workflows such as reviewing documents. Investors panicked about the prospect of highly-rated Software-as-a-Service (SaaS) and data-driven businesses being bypassed by AI, and many share prices dived.

Such stocks make up a large proportion of Finsbury Growth and Income (FGT)’s portfolio. RELX, which owns LexisNexis was an obvious casualty, but investors reasoned the accounting software company Sage, credit scoring company Experian, and even the London Stock Exchange (LSEG), which makes far more money out of data provision than listing fees, could also be disintermediated.

The falls this week were severe but zooming out reveals that many of these stocks peaked early last year and have been falling since: RELX is 48% off the top, for example, and LSEG off 37%.

FGT’s manager Nick Train addressed the AI issue in his December factsheet, observing that: “Sage, Autotrader, Experian, RELX and Rightmove materially grew their businesses in 2025 and that, if anything, the opportunities for future growth are even brighter.”

Then, as we reported yesterday, Train has since argued that the market’s “panic” over Anthropic’s threat to his key data and software holdings was mis-placed. In particular, he feels that the value of businesses such as LSEG and RELX lies in their databases.

FGT wasn’t the only trust affected. HgCapital Trust (HGT), which has been a star performer for well over a decade, largely on the back of its success in building software businesses, gave back all the gains it had made since June 2021 in a few days. I confess that, although I have top-sliced my position over the years, I am still a shareholder in the trust.

A story in the FT from last Monday (behind a paywall) highlighted the problem that Hg may be facing. Exponent, a rival private equity firm, was forced to cancel its planned sale of UK accounting firm Xeinadin after failing to get bids at Exponent’s £1bn+ target price. The message is that these software and professional services firms don’t actually have to run into problems before NAVs start falling, it may be enough that confidence has been undermined.

Again, as we covered, HGT’s manager has sought to reassure investors about the prospects for its portfolio. In a statement released on Wednesday ahead of today’s announcement about HGT launching a share buyback, Matthew Brockman chair of Hg’s investment committee argued that AI adoption will be measured and likely led by incumbents. He highlights the considerable work that Hg’s investee companies are already doing to embed AI tools within their businesses.

I can see some logic to this in that these companies already have the customer relationships and hopefully insight into their needs. The SaaS firms can also use AI to cut their own cost base and look to pass savings onto customers as a way of trying to retain the business.

In addition, there are well-founded concerns about AI’s history of hallucinating data. As much as AI’s reliability has improved, it will still take time to convince business leaders of the merits of using agentic AI for critical systems.

But are we just talking about delays to an inevitable decline? Mark Sheppard, manager of Manchester & London (MNL), says “If Application Software wishes to recover it has to switch its revenue model from a Seats based model to an Outcome based model. This will take years to achieve just as the historic shift from license to seat took in the early 2000s. At this point, our guess is that bottom fishers will get fried.”

That would also spell bad news for HGT, but private equity may yet profit. Sheppard says: “The end game for Application Software names is that they will become so, so cheap that the private equity names will pounce. Then starts the hard process of rebuild.”

He was concerned enough about the potential hit to enterprise software businesses to slash his position in Microsoft and exit a 3.1% position in Synopsys. That decisive action ought to help calm nerves and narrow MNL’s overwide discount. It is still the best-performing trust in the technology sector over three years and not that far behind sector-leader PCT over one year.

A few commentators have pointed to Scottish Mortgage (SMT) and Schiehallion (MNTN) as beneficiaries of this as they both hold Anthropic. I’d point out though that these positions are not that large, 0.7% for SMT and 1.5% for MNTN. Also, surely Anthropic needs to win business from incumbent players to justify its valuation.

Then a couple of other burning questions formed in my mind. Yes, Claude Cowork might be furthest advanced with this agentic AI, but how many other firms will catch up (keeping one eye on what is happening in China, for example) and how long will it be before this area is commoditised?

And, in a similar vein, as inference costs (the expense of running trained models) fall, won’t that put downward pressure on revenues from agentic AI?

But, for me, perhaps the most important question is what happens to the global economy as IT specialists, junior lawyers and accountants, and many other roles are eliminated by AI. It feels to me as though this is already showing up in youth unemployment numbers, but the layoffs may accelerate.

My concern is that while the AI impact may be impossible to ignore, not all of the changes it brings will be positive.