New York
\n \u00a0\u2014\u00a0<\/p>\n
\n Just weeks after its new AI tools for the office shook software stocks, Anthropic is pushing even deeper into the workplace.\n <\/p>\n
\n The company is updating its Claude AI helper to perform better at tasks for specific jobs, such as design, human resources and wealth management roles, while also enabling it to work within applications like Microsoft Excel and PowerPoint apps. Anthropic made the announcement during a virtual event on Tuesday.\n <\/p>\n
\n Anthropic launched Claude Cowork in January with the goal of expanding the usefulness of its popular coding tool to more types of office work. It\u2019s made several updates since then, first by adding plugins that make it better at jobs like financial and legal analysis. Then, it upgraded the AI model<\/a> that powers the agent and launched a new tool for cybersecurity work<\/a>. Now, it\u2019s furthering that push into broader office work with even more industry-specific tools.\n <\/p>\n \n The AI company\u2019s rapid-fire updates and improvements have given Wall Street whiplash, making investors nervous that AI could soon make other businesses\u2019 products obsolete and lead to mass layoffs. Anthropic strongly denies that \u2013 it says it\u2019s not trying to replace products from enterprise software companies, which are deeply embedded in businesses\u2019 systems and maintain trusted tools to handle sensitive, proprietary data.\n <\/p>\n \n Instead, Anthropic\u2019s Scott White, head of product for enterprise, said the company is building something complementary to work with existing software and tools. Anthropic views itself \u201cas a platform, not a product, trying to own every workflow,\u201d White said.\n <\/p>\n \n But Anthropic\u2019s fast development speed remains critical in a sea of competitors breathing down its neck. Rival OpenAI is ramping up its own enterprise offerings.\n <\/p>\n \n Instead of using Claude as a separate chatbot, it can now live inside enterprise software tools, pulling context and data without users needing to leave the window they\u2019re working in. The idea is to make it so that Claude could, for example, use spreadsheet data to create slide presentations the way a person would.\n <\/p>\n \n That should make it much easier for users to interact with Claude without having to copy and paste information between apps, according to White, who claims that should make Claude \u201clike a real, fully capable virtual collaborator.\u201d\n <\/p>\n \n The new plugins for specific fields will also make Claude better at tasks such as modeling scenarios in private equity work, developing job descriptions and offer letters in human resources, putting together creative briefs for design-related work and summarizing vendor proposals for operations-related tasks. Anthropic says it worked with companies like FactSet, S&P and LSEG for the financial services plugins and Apollo for private equity tools.\n <\/p>\n \n Companies will be able to customize the plugins to work with the apps used by their organization, such as Google Drive, Gmail, Google Calendar, DocuSign and others.\n <\/p>\n \n Anthropic\u2019s announcement comes after the quiet rollout of several industry-specific Claude Cowork plugins rattled software stocks<\/a> in early February, raising concerns that the tool could challenge existing analytics and research products. A software industry ETF fell nearly 6% in a single day back then, its worst session since April. Thomson Reuters saw its biggest single-day stock drop on record in early February, plunging nearly 16%. LegalZoom sank almost 20%. FactSet dropped more than 10%. European data analytics giant RELX fell 14%.\n <\/p>\n \n IBM shares (IBM<\/a>) tumbled Monday after the AI startup published a blog post<\/a> about how AI could help modernize COBOL, a decades old programming language for business data processing. IBM offers tools for compiling and upgrading COBOL code.\n <\/p>\n \n And cybersecurity stocks were down after Anthropic debuted<\/a> a new capability for Claude Code on Friday that \u201cscans codebases for security vulnerabilities and suggests targeted software patches for human review.\u201d\n <\/p>\n