UBS Analysts Warn AI Models Threaten Enterprise Software

UBS analysts attended the HumanX conference and concluded that models from Anthropic and OpenAI are moving from underlying infrastructure toward acting as full application platforms. They observed customers deploying Claude and ChatGPT to build production applications and autonomous agents, and reported a broad increase in investment in AI agents beyond coding helpers such as Copilot. The UBS note flags enterprise software vendors as vulnerable where AI can subsume existing workflows, while singling out firms that manage or secure corporate data as relatively safer.

The analysts cited customer checks showing practical app and agent builds on Claude and ChatGPT, and pointed to plugins and vertical integrations that automate domain tasks. Key technical patterns they highlighted include:

• •AI agents that chain model calls and external connectors to automate multi-step business processes
• •Coding and developer-assist agents that reduce routine engineering toil, potentially compressing demand for some dev tooling
• •Vertical plug-ins that embed model reasoning into specific workflows, for example legal review automation exemplified by Anthropic's Cowork

This is an investor-driven assessment of product-market risk. The observation that large foundation models are being used as application platforms tracks a multi-year trend of models moving up the stack. Where enterprise vendors sell workflow automation, document processing, or task orchestration, model-native agents threaten to replace packaged functionality. The UBS view helps explain volatile market responses to new AI product releases, including the notable selloff linked to Anthropic's Cowork plug-in launch earlier this year. At the same time, data governance, security, and systems that integrate and curate enterprise data retain strategic value because they control the inputs, compliance, and auditability agents need.

Monitor adoption metrics for Claude and ChatGPT in enterprise environments, the pace of plugin and agent deployments, and vendor responses that emphasize data control, audit trails, and tighter integration with secure backends. The competitive pressure will push incumbent software companies to either partner with model providers or evolve into secure data and orchestration platforms.