Anthropic targets core business systems with new Claude plug-ins

Anthropic is expanding its push into the enterprise market with a new set of “coworker” plug-ins designed to embed its Claude AI directly into tools used by investment bankers, HR teams, and engineers, signaling a shift from standalone assistants toward AI agents that operate inside core business workflows.

In a blog post, the company said new connectors are available for widely used enterprise platforms, including Google Workspace tools such as Calendar, Drive, and Gmail, as well as DocuSign, FactSet, MSCI, and LegalZoom, while partners such as Slack, LSEG, and S&P Global have built plug-ins for joint customers.

The integrations allow Claude to assist with tasks such as reviewing deals, analyzing portfolios, drafting HR materials, and supporting engineering and design work, Anthropic added.

“Admins can now set up plugins from starter templates or build them from scratch, with Claude guiding you through setup by asking questions to tailor skills, commands, and connectors (MCPs) to your company,” Anthropic said in a blog post.

The rollout comes weeks after the company introduced a legal-focused integration that drew intense attention across the technology sector.

While Anthropic has previously said its goal is to improve outcomes for customers rather than replace software providers, the speed of its enterprise expansion highlights growing competition among AI vendors to become the underlying engine powering corporate workflows.

Earlier this week, OpenAI said it had formed multi-year partnerships with consulting firms including Accenture, Boston Consulting Group, Capgemini, and McKinsey to help enterprises integrate systems and scale AI deployment globally.

The move unfolds against a politically sensitive backdrop, as US Defense Secretary Pete Hegseth has reportedly given Anthropic a deadline to revisit AI use restrictions linked to government contracts.

Implications for enterprises

For CIOs and IT leaders, the development raises practical questions about integration complexity and data governance.

From a functional perspective, the new plug-ins may not introduce entirely new capabilities but could enhance productivity and accelerate delivery timelines, according to Faisal Kawoosa, founder and lead analyst at Techarc. That is the yardstick many CIOs are likely to use when assessing their value.

“CIOs should take an outcome-driven approach to evaluate these offerings directly against capabilities already embedded in core SaaS platforms,” said Charlie Dai, VP and principal analyst at Forrester. “They should measure whether the plug-ins reduce cycle time, manual effort, or error rates in specific, high-friction workflow steps rather than simply generating similar outputs in a different interface.”

Dai said plug-ins built with stronger business controls and governance, and designed to work across multiple enterprise systems, are more likely to deliver lasting value than tools that simply repeat basic summarizing or drafting features already found in existing software.

However, “Most enterprises are still cautious and primarily deploy AI agents in advisory or assistive roles rather than granting autonomous authority over mission-critical processes,” Dai added.

Governance and risk concerns

Connecting AI agents across core enterprise systems introduces concentrated risk around identity management, access control, and potential data leakage, particularly when those systems span finance, HR, and collaboration platforms.

“Broad permissions, opaque agent actions, and unintended data blending across finance, HR, and collaboration systems can create compliance exposure and audit gaps,” Dai said. “Enterprises must embrace zero-trust principles to enforce least-privilege access, maintain detailed action logs, and apply policy-based constraints on what agents can read, infer, and execute, especially in regulated or irreversible workflows.”

Kawoosa said many enterprises are likely to begin by deploying such plug-ins in non-core functions, treating them as experimental tools before expanding their role. Building trust in these systems could take nine to ten months or longer, he said, depending on internal governance maturity and risk tolerance.

“But at the same time, there is immense pressure on CIOs to do something with AI within their enterprises,” Kawoosa said. “Many CEOs have started believing that they could be wiped out if they don’t act now. This pressure could force some enterprises to move toward quick implementations, even in critical tasks, to avoid any kind of FOMO.”