TL;DR

Wall Street’s sell-off across brokerage, property, and software stocks reflects a deeper shift: AI model-builders Anthropic and OpenAI are moving directly into enterprise territory. The central question is no longer whether AI will disrupt these industries, but who will control the agents doing the work.

A Full-Frontal Attack

The hammering of wealth management stocks this week — triggered by fintech upstart Altruist — illustrates how generative AI has given a new twist to familiar disruption narratives. Altruist launched an AI-powered service helping investment advisers analyse portfolios and recommend strategies, the kind of thing fintechs have done for years but now supercharged by large language models.

The real alarm, however, comes from the AI model-builders themselves. Anthropic repurposed its coding agent into Cowork, a general-purpose agent with plug-ins for tasks like legal contract analysis and marketing content production. OpenAI followed with Frontier, an enterprise platform designed to control, orchestrate, and optimise all AI agents accessing a company’s systems.

The Agent Control Question

For incumbent software companies, the threat is existential. If AI agents created by Anthropic or OpenAI start performing the work customers value most, existing providers risk becoming utilities — storing data that other companies turn into valuable services.

The AI companies describe themselves as partners rather than challengers, which is unsurprising given that a large part of their revenue comes from selling model access to potential competitors. But their product roadmaps point in one direction: controlling the orchestration layer where agents interact with enterprise systems.

Some incumbents are already fighting back. Salesforce blocked third-party AI services from drawing data through its Slack communications platform. But restricting access risks alienating customers who want the best available tools.

Looking Forward

This battle is only starting to take shape. For UK businesses relying on established enterprise software, the practical question is whether their current providers can build competitive AI agent capabilities, or whether Anthropic and OpenAI will set the terms. The answer will likely determine which companies thrive and which get pushed into the background.