Data disrupted: How AI is hitting everything from comparison sites to software and analytics firms
Related
news
Shares Magazine
In early February 2026 ChatGPT rival Claude, a large language model owned by Anthropic, released a suite of specialized AI tools which sent a shockwave through certain sectors.
The software space was already under pressure from generative AI which is increasingly writing code. Anthropic CEO Dario Amodei told listeners on a recent podcast that the company itself was using Claude to write up to 80% of its own coding.
Investors zoomed into the legal services AI agent which initially knocked shares in companies that sell data and services into the legal industry such as Relx and Wolters Kluwer.
However, it quickly dawned on investors that Claude had released a whole suite of AI agents across marketing, sales and even wealth management.
It should be emphasised that Anthropic’s immediate goal was not to disrupt these sectors but to showcase how firms could build applications using its large language model.
In any case, the release of Claude’s AI tools helped crystallise AI disruption fears for investors and triggered a knee-jerk reaction, wiping out over $285 billion in global market value in a single day across a range of sectors including software and data management.
The selling hit companies operating beyond legal services including enterprise software providers, platforms like Autotrader and Rightmove as well data companies like Experian and the London Stock Exchange.
A practical example of how AI could disrupt legal services
Relx runs it legal business through LexisNexis, which is a premium subscription research platform used by law firms to access case law, court documents and citations.
Lawyers subscribe to tiered plans starting at around $150 per month, scaling up to advanced plans for multi-state coverage and advanced tools. A junior associate spends around three hours a day researching precedents and pulling case material to draft a report.
How does Claude disrupt this? Claude’s AI desktop automation tool monitors incoming client briefs which then trigger a search using publicly available court records and synthesises relevant case law before drafting a research memo into the junior associate’s folder. The associate’s role shifts from doing the research and drafting reports to reviewing and refining. The hours spent previously justified the cost of a subscription. If AI agents can automate enough of the work, firms question whether they need to pay for so many ‘seats’ (people using the subscription). There’s a risk that a manager can now justify downgrading from a firm-wide Nexis+ subscription to a shared seat. In other words, the AI agent doesn’t affect LexisNexis’s data, it attacks the human time spent using the database. Relx introduced its own AI legal assistant, Protégé in 2025 to counter the threat.
How are firms adapting?
Software companies have traditionally generated revenue by charging per number of users or ‘seats’ deployed. This revenue model is no longer viable when an AI agent can automate the work.
There are signs the software industry is moving towards a model based on the number of completed tasks or successfully taking a project from initiation to completion. A good example of a company integrating AI into its customer proposition is Salesforce.
In its quarterly earnings report on 26 February, the company introduced a new metric to investors called AWU (Agentic work unit), defined as a discrete task accomplished by an AI agent, comprising decisions made, records updated, and workflows triggered.
Salesforce said it has completed 2.4 billion AWUs on its Agentforce platform in the last two years and notched up 57% quarter-on-quarter growth in the most recent reporting period.
For professional service firms like McKinsey and Accenture the classic charging model has been billable hours.
Where an AI agent can perform a junior associate’s 40-hour data analysis in fraction of the time, the old model is vulnerable to lead revenue shrinkage, everything else being equal.
Therefore, professional firms are moving towards a hybrid model comprising a fixed ‘access fee’ to use a firm’s brand and specialised AI models, and a ‘consumption fee’ based on performance and completed tasks.
Which firms are better positioned to adapt?
Share price performance since the lows in mid-February suggests a clear split between winners and losers.
Specifically, the market appears to have given businesses which own proprietary data the benefit of the doubt.
These include firms like Relx, MONY and Thomson Reuters which have subsequently seen a significant recovery in their share prices.
The implication is that companies which own proprietary data and have private information on their customers are incentivised to build their own AI tools, thus limiting head-on AI competition.
Relx provides a good example with the shares getting back within 10% of the price before the Claude-related sell-off.
The company has launched or announced 13 products powered by AI including its legal research platform Lexis+ and integrated legal assistant Protégé.
Chief financial officer Nick Luff told Reuters: “We’re applying our algorithms, proprietary algorithms, so that we can get out the right judgments, the right inferences, and the right interpretations to professional users making high-value decisions.”
The key to long-term success will rest on how tightly companies integrate their own agents on top of proprietary data to lock customers into their ecosystems.
This does not mean that firms are immune to disintermediation, or in other words being cut out of the picture by AI, but it reinforces the importance of private and curated data.
What are the risks for Sage, Rightmove and Autotrader?
The share price recoveries have not extended to all companies hit by the Claude agent release. This suggests investors are more concerned about the impact of AI on these business models.
Investor Nick Train who manages the Finsbury Growth and Income investment trust and which has holdings in Autotrader and Rightmove, argues that these companies still own vehicle/property data, pricing histories and behavioural insights, which are not available to AI agents.
Sage’s business model is about providing accounting records and compliance software for SMEs (small and medium sized companies) which, in theory makes it harder for AI agents to get access as they typically sit on top of the whatever accounting system a business uses.
The company is building its own network of agents for MTD (Making tax digital) payroll and cashflow on the Sage Platform and Sage Copilot.
Sage already automates around 80% of the administration behind pulling records, flagging anomalies and reminding users of HMRC deadlines, leaving the final approval with the accountant.
It is also integrating open banking feeds via partners like GoCardless to pull bank transactions into Sage products, AI agents can connect directly to SMEs bank data via the same open banking links, potentially lowering switching costs.
For the uninitiated open banking was introduced by the Competition and Markets Authority in 2018 to boost competition and provide a secure way to share information with information providers.
The question of whether open banking is a tailwind or threat depends on how quickly Sage turns its system of record position into AI-driven services.
It is interesting that Fundsmith, which specialises in investing in high quality, resilient global growth companies has initiated a new position in Sage according to the fund’s January newsletter, after exiting US financial technology platform Intuit.
What conclusions can we draw?
There are two important observations which investors need to consider. The capabilities of generative AI are accelerating at breakneck speed. The adoption of AI at the consumer level has been unprecedented.
As of early 2026 there are an estimated 1.1 billion people actively using AI worldwide, or roughly 13% of the global population, with ChatGPT holding the dominant share.
The initial stock market reaction to the release of Claude’s AI applications has been indiscriminate, claiming victims across a broad basket of stocks from legal research to sales tools and analytics, all in one fell swoop.
Corporations have been slow to adopt AI and have taken a more cautious approach in contrast to the hype surrounding adoption by consumers.
For investors it feels important to figure out the strength of a company’s proprietary data and the speed at which it is rolling out AI tools, and how credible they are, to determine if it risks being circumvented by external AI agents.
Martin Gamble: Shares and Markets Writer
Martin Gamble is Shares and Markets writer at AJ Bell. He was previously the Education Editor of Shares Magazine. He has been with the business since 2019.
Martin graduated from the University of Kent in...
