Anthropic’s AI Tools Rattle Software Stocks, Prompt Rethink of Sector Valuations

Shares of several information and professional services companies fell sharply this week following Anthropic’s unveiling of a legal automation tool that shook investor confidence in the sector’s long-term pricing power.

Thomson Reuters fell 18%, Pearson fell 7% and LegalZoom fell almost 20%, as the sell-off spread across software, financial services and asset management stocks, roughly erasing stock markets. $285 billion in terms of market value, Bloomberg reported.

The panic started after Anthropic announced eleven open-source plugins for Claude Cowork on January 30, but focused on one in particular.

That included one legal pluginthat automates contract review, NDA triage, and compliance workflows. In short, it does the heavy lifting that keeps thousands of paralegals and junior associates employed.

The panic wasn’t just about one plug-in that did document review, but about what the component represents: base model companies that started building full-fledged workflow products, prepared to tackle the enterprise software industry head-on.

“The market reaction was a signal not that AI agents will immediately replace these companies, but that investors are finally pricing in the structural risk that foundation model providers can now compete directly with the software layer.” Scott Dylanfounder of Nexatech Ventures, said Declutter. The fear is not speculative, he said.

“That’s a polite way of saying that if Anthropic can build a legal workflow tool internally, what’s stopping them from doing the same in finance, procurement or HR?” Dylan added.

If AI agents can do that, why would anyone pay per seat? That’s the business model that Salesforce, Bloomberg and every SaaS giant has built.

And now cracks are starting to appear.

Short-term FUD or structural interest rate revision?

“The selling pressure reflects a deepening structural debate,” Schroders analyst Jonathan McMullan said Reuters. “Investors are aggressively revaluing these areas as the historic ‘visibility premium’ erodes; the pace of advances in AI is making long-term valuations harder to defend, especially as AI tools enable companies to do more with fewer staff, endangering the traditional model of billing per software user.”

These concerns have spread beyond legal technology.

Advertising giants Omnicom and Publicis fell by 11.2% and 9% respectively. Australian cloud accounting firm Xero had its worst day since 2013, down 16%.

What do the people who actually do the work think?

When asked whether advances in AI agents pose a threat to legal work, Joel Simon, founder and partner of Simon Perdue, a firm with operations in Texas and New Mexico, struck a measured tone.

“We live in a world where judgment and credibility are more important than raw processing power,” Simon said Declutterarguing that human judgment still outweighs pure computational speed. “AI can sort through vast amounts of information, flag patterns and surface issues faster than a junior staffer ever could. If anything, this has been a relief because it has cleared the runway so we can focus on strategy, witness preparation, storytelling and decision-making under pressure.”

Simon said his company has already integrated AI into its daily operations, describing the technology as an accelerator rather than a replacement for lawyers.

It is already being used to draft outlines, condense discovery material, and test potential questions, while attorneys maintain control over judgment, narrative, and strategy in the courtroom. “AI is not taking a position,” he said. “We do.”

Simon predicts that in two to three years, “attorneys who embrace AI will be more valuable, not less.

The job will be leaner as fewer hours are wasted on routine work and more time is spent on case theory, client advice and execution in the courtroom.

Nexatech’s Scott Dylan had a less optimistic view.

“The honest answer is that AI agents will displace certain types of work – especially repetitive, rule-based tasks that can be well specified,” he told me. Declutter. “Contract review, NDA triage, compliance checklists. These are exactly the workflows Anthropic focuses on, and they are performed by tens of thousands of paralegals and junior associates,”

But Dylan is not entirely pessimistic. “Displacement is not the same as elimination. What is more likely is a compression at the entry level. Junior roles that used to be a training ground – associate work at law firms, analyst roles at consulting firms, frontline client support – will shrink,” he said.

Human challenges in an agentic society

Dylan said employees need to learn how to do that adapt and overcome.

“I don’t think we’re moving towards a world where people become redundant,” he said. “The scenario where agents handle all the knowledge work and people wonder what to do with themselves is, frankly, unlikely in any time frame that matters.”

In the long term, human workers will predominate in “roles that require physical presence or intensive human interaction,” such as health care, personal services and skilled trades, Dylan added.

But until society adapts, there will be a painful period for everyone, and investors are already pricing in all these elements.

IDC forecast that by 2028, pure seat-based pricing will be obsolete, with 70% of software vendors switching to pricing based on consumption, outcomes or organizational capabilities. When an agent does the work, customers expect to pay for results, not logins.

For now, enterprise software companies are experimenting with different models.

Bain & Company analyzed more than 30 SaaS vendors introducing generative AI. Almost 35% increased prices per seat with bundled AI features. Another 35% adopted hybrid models with usage-based add-ons.

The rest are experimenting with outcome-based pricing: billing per contract reviewed, ticket resolved or lead generated, rather than per seat occupied.

The challenge now is to ask customers to spend more before they see savings. A SaaS company deploying a $40,000 AI agent to replace an $80,000 salesperson faces a problem: In the short term, the customer needs both the employee and the agent in evaluating results. That is a 50% cost increase for an indefinite period.

“The problem is that most agents today rely on APIs that burn through tokens quickly, which can create expensive and unpredictable bills if not tightly controlled. Davis Housekeeperdirector of MYCO Management, said Declutter. “In those cases, you simply replace one SaaS subscription with another.”

“Unlike normal gen AIs, the risk with agents is not isolated failure, but failure on a large scale,” Householder added.

In the coming years, people can likely expect major disruptions to their working lives. Layoffs, driven primarily by fear, may occur alongside more complex automation workflows as the tooling matures.

Developing richer multi-agent ecosystems with better APIs and coordination protocols could pose a new challenge. Regulators’ attention will also turn as governments realize that autonomous agents can be weaponized or cause social instability.

In the medium term, the infrastructure could harden. There will be better regulation of work environments where people interact with agents.

We will likely see agent marketplaces with reputation systems, verified skills, and standardized protocols for autonomous agent-to-agent transactions. Expect some high-profile security breaches along the way to serve as a wake-up call.

In the long term, this will likely be a restructuring rather than an extinction event.

As AI compresses margins and commodifies basic functionality, the strongest companies consolidate their power. The real value could shift from chair-based software to proprietary data, including legal databases, financial benchmarks, compliance logic, and licensing for agent-driven systems. Service will continue to exist, but data will become the core business.

What AI agents mean for jobs: displacement or reinvention?

In the meantime, the consequences are dire.

A MIT study found that 11.7% of US jobs could already be automated using current AI technology.

Research published by the World Economic Forum in 2025 shows this almost 60% of workers worldwide will need to ‘retrain’ to remain relevant in the post-agent era.

“We need to address our education system and innovate the way we train people so they use AI to do their jobs better, rather than letting AI do their jobs entirely, which puts them at risk from employers trying to reduce costs.” Amrita Bhasin CEO of Sotira and advisor to Fortune 500 companies, narrated Declutter.

“There is no feasible way to prevent AGI,” she said. “We must support the average American worker and ensure they have the skills, training and ability to compete in an increasingly competitive and/or unstable labor market threatened by AI.”

Companies and professionals that adapt – learn to collaborate with AI agents, shift from execution to supervision, and anchor their value in judgment rather than process – are likely to do better.

Those that don’t adapt risk being revalued, just like the stocks that sold off this week.