A Citi analyst says the company's latest event reveals how frontier AI labs are grappling with a problem that has tripped up enterprise software vendors for decades: change management
Anthropic's "Enterprise Agents" briefing this week was less a product launch than a positioning exercise.
The company showcased customer stories emphasising time savings across development and non-engineering work, while rolling out feature updates focused on admin controls, extensibility, and audit trails, the unglamorous plumbing that corporate IT departments actually care about.
The subtext was this: Anthropic is trying to embed Claude deeply enough in enterprise workflows that it becomes the default "thinking engine" for knowledge work, with Cowork serving as the natural next step for companies already running Claude Code.
But Citi analyst Heath Terry flags a structural tension. Anthropic and its peers are discovering that building a superior model is only part of the challenge.
Driving adoption at scale requires navigating procurement cycles, compliance requirements, and organisational inertia; the same friction that slowed every previous wave of enterprise software.
Partnerships with global systems integrators suggest Anthropic knows it cannot close that gap alone. Whether those relationships translate into durable revenue, as inference costs continue to compress margins, remains an open question.
His commentary chimes with that of Wedbush analyst Dan Ives, who watched the same demos.
His core argument is that the market is confusing model capability with enterprise software replacement, and that conflation is driving a software sell-off that the underlying business reality does not support.