Anthropic sets its sights on the corporate world: why 2026 is the year of AI agents in business
AI has stopped being individual productivity and become operational infrastructure for businesses. According to Forbes, 90% of Anthropic's internal code is already written by agents. See what this means for your business.
What Anthropic announced for the corporate world
In a Forbes report, Henrique Savelli, Anthropic's AI architect, opened his talk at StartSe's AI Festival with a figure that sums up the market shift: 90% of all the code produced internally at the company today is written by AI agents. Not as a target, but as the reality of the operation.
The message is that artificial intelligence has stopped being an assistant for research and text and has become operational infrastructure. To back that bet, the company is seeking a funding round of between US$ 30 and 50 billion, which could raise its valuation to as much as US$ 950 billion, according to The New York Times, as cited by Forbes.
Claude Code
An agent that receives a problem, breaks it into smaller tasks, executes each one, and fixes its own errors along the way. According to Anthropic, 100% of its own code is already written by itself.
Cowork
Described as the first AI agent to work end-to-end with areas such as sales, legal, and finance, accessing the company's CRM, databases, and documents.
Claude Design
An agent for website, app, and presentation layouts that analyzes the company's code and design and exports to PPTX, Canva, PDF, or HTML.
Managed Agents
Infrastructure for companies to build and publish their own agents on Amazon Bedrock, Google Vertex, and Microsoft Foundry, without managing models or hardware.
From individual productivity to operational infrastructure
The timeline helps explain the movement. In 2023, with the first chatbots, AI became popular as an assistant for research, text generation, and brainstorming. In 2024, the focus shifted to applications and process automation. The bottleneck, according to Savelli, was human dependency: when an input strayed from the pattern, the flow broke and someone had to fix it by hand.
It's this bottleneck that agents promise to solve. Instead of a well-structured prompt asking for a task, the user hands over a problem, and the agent breaks it down, executes it, verifies it, and corrects it. That was the principle behind Claude Code and, later, Cowork, brought to non-technical teams in sales, legal, and finance. Integration with Google Workspace, Microsoft 365, and Canva extends that automation to areas such as finance, marketing, support, operations, and HR.
What changes for businesses
The promise is appealing, but the takeaway that applies to any business is the same one Gartner reinforces in its AI Hype Cycle for ITSM: the value doesn't come from the hype of full autonomy, but from well-governed agents acting on trustworthy data. An agent that accesses CRM, databases, and documents is only as good as how well that information is organized.
In practice, adopting agents across the whole company requires three foundations: standardized data and processes, governance (access control, audit trail, oversight), and use cases with clear ROI. Without this, automation fails at scale. With it, it reduces rework and frees people for what requires judgment.
How Central IT already delivers agentic AI with governance
The movement Anthropic describes, of agents working with the entire company, is exactly what Central IT operationalizes in Brazil, with governance and local context. It's the principle behind the CITSmart X² platform:
- Agents that act across business areas. Just as Cowork targets sales, legal, and finance, CITSmart ESM brings AI-powered service to IT, HR, Finance, Legal, Procurement, and Facilities in the same console.
- Governed agents, not experiments. The Hyper Agents execute triage, routing, and resolution with auditing, access control, and oversight. It's agentic AI in production, with a trail.
- Data as the foundation. Catalogs, SLAs, and an up-to-date CMDB give agents the trustworthy foundation that real automation requires.
- Measurable ROI. SLA, volume, and productivity dashboards make each initiative comparable, in the spirit of pilot projects with clear returns.
While the big players move the global market, Central IT already puts these agents to work inside the operations of Brazilian companies and governments, with the compliance care and the experience of a company that has worked in service management for more than two decades.
How to adopt AI agents without becoming a hostage to the hype
- Organize your data first. An up-to-date CMDB, standardized catalogs, and processes by area.
- Start with a case with clear ROI. Triage, routing, and recurring responses deliver quick returns.
- Build in governance from the start. Access control, auditing, and human oversight on the agents.
- Measure and expand. Compare before and after, and take AI to new areas based on results.
Frequently asked questions
What did Anthropic announce for the corporate world?
According to a Forbes report, Anthropic presented a strategy of heavy investment in the corporate market, with products such as Claude Code, Cowork, Claude Design, and Managed Agents, plus integrations with Amazon Bedrock, Google Vertex, Microsoft Foundry, Google Workspace, Canva, and Microsoft 365 to automate finance, marketing, sales, support, operations, and HR workflows.
What does it mean for AI agents to work across the entire company?
It means moving from AI as an individual assistant (research and text generation) to AI as operational infrastructure, with agents that execute tasks end-to-end across multiple areas, accessing systems and data. According to Anthropic, 90% of the code produced internally is already written by AI agents, and 2026 would be the year agents start working with the rest of the company.
What products has Anthropic launched for businesses?
According to Forbes: Claude Code (a coding agent), Cowork (an agent for sales, legal, and finance), Claude Design (creation of layouts and presentations), and Managed Agents (deploying your own agents across the leading clouds). The company is also seeking a funding round of US$ 30 to 50 billion, which could raise its valuation to as much as US$ 950 billion, according to The New York Times.
How can Brazilian companies adopt AI agents safely?
By prioritizing governance and trustworthy data before broad automation: standardized catalogs and processes, access control, audit trails, human oversight, and pilot projects with clear ROI. An agent can only decide well on a well-organized foundation.
How does Central IT apply agentic AI in the corporate environment?
Central IT operates agentic AI with governance through the CITSmart platform: the Hyper Agents execute tasks in areas such as IT, HR, Finance, Legal, Procurement, and Facilities, on trustworthy data and with auditing, access control, and measurable ROI, adapted to the Brazilian context and compliance requirements.
This article is a Central IT analysis based on the report "A Mira da Anthropic Agora É o Mundo Corporativo," published by Forbes Brasil. Data and statements are attributed to Forbes and the sources it cites. Trademarks and products mentioned belong to their respective owners.
Put AI agents to work in your operation
Discover CITSmart and the Hyper Agents: governed agentic AI, built on a mature data foundation with measurable ROI.