Anthropic is expanding how enterprises can deploy Claude inside day-to-day work, rolling out new Cowork updates and a broader set of plugins and connectors designed to turn the model into role-specific agents that sit alongside existing software rather than replace it.
The announcement, made Feb. 24 during the company’s Enterprise Agents event, centers on a push to make Claude easier to customize and govern across departments. Anthropic said enterprises can now build private plugin marketplaces to distribute approved agents across their organizations, while administrators gain tighter control over what tools those agents can access.
In a note following the event, investment banking firm William Blair said Anthropic is trying to establish Claude as a “platform-level intelligence layer across enterprise workflows,” framing the model as infrastructure that enhances existing enterprise systems rather than displacing them. The firm argued the strategy is aimed at making AI useful to broader employee populations, not only technical teams.
A key product change is a revamped administrative experience for creating and managing plugins. Anthropic said admins can start from templates or build from scratch, with Claude guiding setup by asking questions to tailor skills, commands and connectors to a specific company’s environment. Those capabilities are being consolidated into a new Customize menu that brings plugins, skills and connectors into one management surface.
Anthropic also updated the connector experience with an improved directory and streamlined controls for bundling connectors into plugins. On the distribution side, the company is adding org-specific marketplaces, per-user provisioning and auto-install features, and it is testing private GitHub repositories as plugin sources in a private beta, according to the announcement.
For end users, Anthropic said slash commands can now launch with structured forms, making workflows such as generating a report or dashboard feel closer to filling out a brief than writing a long prompt. Cowork is also adding more company branding in the interface, including a redesigned home experience tailored to the organization.
Advertisement: Scroll to Continue
Anthropic said it is also adding usage and tooling visibility so administrators can track usage, costs and tool activity across teams — features typically required for enterprise procurement and governance.
More Connectors, Role-Based Agents
To make agentic workflows practical, Anthropic is expanding connectors — integrations that pull Claude closer to the systems where work already lives. The company highlighted new connectors that include Google Workspace apps such as Calendar, Drive and Gmail, as well as enterprise tools and data platforms including DocuSign, FactSet, Similarweb, LegalZoom, WordPress, Slack (Salesforce), LSEG, S&P Global, MSCI and OpenTelemetry, among others listed in its materials.
We’d love to be your preferred source for news.
Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!
It also expanded its library of prebuilt plugin templates meant to reflect how specific jobs are performed. New templates include agents for HR, design, engineering and operations, plus more specialized agents for financial analysis, investment banking, equity research, private equity and wealth management. Anthropic also pointed to third-party and partner-built plugins, including a “brand voice” plugin by Tribe AI that can distill internal materials into enforceable guidelines.
Anthropic described plugins as portable “file systems” that customers own and can use across Cowork as well as applications built on the Claude Agent SDK, positioning this portability as a way to share approved agents across teams and, potentially, with outside experts.
One of the more concrete demonstrations focused on cross-application work. Anthropic said Claude can now handle multistep tasks across Excel and PowerPoint, passing context between the two Office add-ins to support larger projects, such as running analysis in Excel and turning it into a presentation in PowerPoint. The company called it an “early research preview” and said it is available for paid plans on Mac and Windows.
Partner Signals and Analyst Skepticism
Anthropic’s materials leaned heavily on partner validation. PwC’s Anthropic Alliance Leader, Sanjay Subramanian, said the firm is partnering with Anthropic to bring “enterprise-grade agents” into the office of the CFO, arguing this could make finance teams more strategic by giving staff tools to take on more ambitious work. Another quote, from Blank Metal COO Mark Hines, emphasized a view that Anthropic’s approach is more practical than the market’s earlier hype around “digital employees.”
Still, the William Blair note underscored a key uncertainty: As AI moves from answering questions to executing multistep workflows, it is not yet clear who captures the economic value of the “agentic layer” — the model provider, the incumbent software platform that holds the data, or both. The firm suggested the market may settle into an equilibrium where model providers supply reasoning while systems of record retain contextual data and business logic, limiting the likelihood of near-term, widespread displacement of entrenched platforms.
That tension also showed up in Anthropic’s own framing. During the event, Anthropic executives argued that transforming knowledge work depends not only on model capability but also on the data ecosystem and infrastructure that can deliver the right information at the right moment — an implicit acknowledgment that integrations, permissions and governance will be as decisive as raw model performance.
For enterprise buyers, the announcement amounts to a clearer product bet: Anthropic is pushing Claude toward being a managed, governable layer that can plug into the tools companies already pay for, with more “out-of-the-box” agents for functions like finance and research. Whether that turns into durable advantage will depend on adoption outside of technical teams — and on whether enterprises view Anthropic’s plugin-and-connector approach as the safest path to scale agentic AI without surrendering control of their data and workflows.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
Add as Preferred Source
