Shares for the future: why I’ve put trading blocks on three shares

Last month’s sell-off in shares thought to be at risk from advances in artificial intelligence (AI) provoked a crisis of confidence in me.

Earlier this week, I put “trading blocks” on three Decision Engine shares, namely Autotrader Group (LSE:AUTO), Keystone Law Group Ordinary Shares (LSE:KEYS) and Softcat (LSE:SCT), while I figured out whether the chill I had caught was serious.

• Invest with ii: Open an ISA | Top ISA Funds | Transfer an ISA to ii

This puts me in an uncomfortable position. These businesses are profitable, growing and financially strong. What we know about them gives us confidence. Their futures are more blurry. What we do not know could be scary.

Expect the resolution to be messy and preliminary. My understanding of these technologies is superficial. We do not know how they will develop.

The basic fear is as old as the hills: obsolescence.

Generative AI uses computer programs called large language models (LLMs) to create intelligible text, images and sound from natural language prompts. AI Agents based on the same technology can interact with each other and software to automate tasks.

Such programmes can do things that previously required human beings, such as draft contracts, evaluate, recommend, and configure software, and scour the web for used cars.

The human touch may not be as valuable as it was.

Keystone Law: softly softly

Among the companies relying most on human capital in the Decision Engine is Keystone Law, widely regarded as the oldest platform law firm. It claims to be the market leader, with more than 600 of the 3,000 or so lawyers working for platforms.

A platform law firm provides the administrative infrastructure for a network of senior self-employed lawyers. The lawyers use Keystone’s brand, and Keystone bills clients, pays the lawyers, provides insurance, and co-ordinates the referral of work, for example, so they can get on with serving clients.

It is a low-cost model that has allowed Keystone to prosper in the “mid-market”, providing a full range of legal services to companies and wealthy people.

• Want to pay less tax? Here are five allowances to use by 5 April
• Stealth taxes and the growing case for pension planning

Self-employed lawyers typically pay less tax. Often, they operate from home or modest commercial buildings instead of swanky offices. Senior lawyers are already fully trained and earn big fees. These attributes mean there is more money to split between the company, customers and the lawyers (Keystone takes 25% of their fees). It allows lawyers to be more entrepreneurial and puts them in control of how much they work.

I last scored Keystone in August 2025, acknowledging the risk that AI might commodify some legal work. I also wrote that the company should tell us more about its technology platform. It identified technology as an enabler but did not explain what makes its technology special.

Technology might be a competitive advantage, but I’m reluctant to jump to that conclusion. Most businesses are enabled by technology. More likely, a higher degree of technology is required because its lawyers work remotely. That is something traditional law firms have got better at since the pandemic, but Keystone has relied on technology more heavily for longer to operate the business.

In September’s half-year results, Keystone told us more. It has started to adopt AI. Its lawyers can use AI to generate notes from Teams (video) meetings. They can use generative AI on all documents in their document management system. They can query the company’s operating manual using AI, and use KeyBot (a private ChatGPT “equivalent”). Keystone is subscribing to some of the AI tools offered by suppliers, perhaps for back-office software and legal information (it is not specific). I expect rivals are taking similar steps.

In a presentation, chief executive James Knight said Keystone employed three software developers although it also outsources development. For example, it is working with consultants to find out how AI agents could make the business more efficient. Work on the highest priority agents should have started by the end of the financial year in January. These may apply specifically to business processes, which would enhance its efficiency advantage without reducing billable work.

• Scottish Mortgage on AI, new holdings, and its edge over trackers
• Ian Cowie: are you braced for AI apocalypse or buying into AI optimism?

The big fear with AI is that clients will demand lower fees as AI improves efficiency. They may also be able to do more legal legwork themselves, assisted by AI. An opposing view is that AI increases the amount of work lawyers can do, and lower fees might stimulate even more work. Some senior lawyers I have spoken to say they are not using AI much, and how they use it is governed by strict policies.

Not being a technology expert or a lawyer, I make these comments with trepidation. But my experience of AI is that humans need to stay in the loop, and that generates more work.

AI makes searching annual reports for specific information or aggregating data from many years of reports much easier. But the results must still be checked. New insights often generate more queries inviting us to do more work. I think this improves my analysis, but it’s not making me less busy or requiring me to be less “expert”.

An article in the Financial Times by Tim Harford last weekend highlighted ethnographic research into how technology workers are using generative AI. He quoted researcher Xingqi Maggie Ye, who pointed to a more sinister effect: “In micro moments of prompting, iterating and experimenting, people talked about momentum and a sense of expanded capability...But when they stepped back and reflected on their broader work experience, a different tone sometimes emerged. They described feeling busier, more stretched, or less able to fully disconnect.”

The most popular comment below the article was from a lawyer. Once handwritten court submissions were necessarily short, they wrote, but the typewriter and then computer freed lawyers up to create lengthy appeals and tend to hundreds of court exhibits: “Work expands to fill the newly created void.”

These thoughts are naive and uncomfortably speculative. Working with a 10-year investment horizon is difficult during changing times.

Candidate for investment

I’m thinking about this because Keystone Law’s share price has been under pressure recently, which could make the shares a more attractive long-term investment.

Interpreting share price declines is perilous. Keystone Law shares have fallen more than 20% since early February. Since the pandemic, business has been booming at law firms, yet Keystone’s revenue growth, still strong, has weakened. The sell-off may just reflect investors’ disappointment that the company is not doing even better, or a view that the boom might be ending, and have little to do with AI.

My score already recognised the disruptive potential of AI and a second risk to the self-employed status of Keystone’s lawyers.

I know even less about this risk, but it is routinely flagged in Keystone Law’s annual report, and it was also in the company’s prospectus when it floated in 2017.

Most of Keystone’s lawyers use personal service companies (PSCs). The government has changed the rules for PSCs before, and it has shown enthusiasm for measures that would even up discrepancies in tax between the self-employed and employed.

In the worst case, changes to employment or tax law could require Keystone to employ its network directly, making it resemble a conventional law firm more closely and removing much of its competitive advantage.

I have looked at listed conventional law firms and concluded that they don’t generate enough money to satisfy growth-oriented shareholders.

The result of my deliberation is that Keystone Law’s score has dropped by 1.5 points.

Under current circumstances, the business meets all my requirements. It is performing strongly, its shares trade at a reasonable price, and it has a competitive business model.

All the deductions fall in the risks category. There are two aspects to assessing risks. How severe we think they are, and how confident we are that our judgement is correct. Most of the reduction in score is due to the recognition that my confidence in my judgement is low.

A score of 6.7 permits me to hold a modest amount of shares in the Share Sleuth portfolio. It equates to an ideal holding size of 2.7%, close to my minimum of 2.5%. On balance, that feels about right.

Trading update

A trading update in February anticipated low double-digit revenue and profit growth for the full year to January 2026, which is subdued compared to the period before the pandemic but typical of recent years.

Lawyer recruitment has increased markedly, though. Keystone Law grows by poaching experienced lawyers and their clients. Typically, this is difficult to do when traditional law firms are prospering and ratcheting up pay to retain their lawyers and poach more, as they have been.

If conditions for recruitment are improving at Keystone, it suggests the market is cooling off. However, Keystone Law benefits in that it recruits more lawyers, and consequently clients.

30 Shares for the future

Here is the ranked list of Decision Engine shares. I review the scores at least once a year, soon after each company has published its annual report. The price scores are calculated using the share price prior to publication.

Generally, I consider shares that score more than 5 out of 10 to be worthy of long-term investment in sizes determined by the ideal holding size (ihs%).

Porvair (LSE:PRV) has published its annual report and is due to be re-scored.

Click on a share's score to see a breakdown (scores may have changed due to movements in share price). Key: qual is the share’s score out of 9 for the three quality factors (capabilities, risks, and strategy), price is the price score from -3 to +1, and ih% is the suggested ideal holding size as a percentage of the total value of a diversified portfolio.

Richard Beddard is a freelance contributor and not a direct employee of interactive investor.  

Richard owns many shares in the Decision Engine. He weights his portfolio so it owns bigger holdings in the higher-scoring shares.

For more on the Decision Engine and Share Sleuth, please see Richard’s explainer.

Contact Richard Beddard by email: richard@beddard.net or on Twitter: @RichardBeddard