Computer use agents automate legacy enterprise workflows

OpenAI identifies two main categories for enterprise automation: tedious workflows in systems of record and creative work. Key use cases include automating procurement processes where agents navigate vendor software and spreadsheets, financial reconciliation across ledgers without APIs, and sales prospecting by populating CRMs. James Sun of OpenAI highlights a specific example where agents use computer use to photograph restaurant back storage, analyze inventory against traffic projections, order groceries via procurement software, and adjust employee schedules in ERPs. The agents can also proactively monitor Slack and email, flagging critical feedback and missing calendar invites, and integrate with tools like Sentry to pull logs and analyze errors. OpenAI notes that while legacy software from 10-40 years ago will persist, computer use allows agents to operate these interfaces directly, reducing the need for new API development. OpenAI executives argue that 'boring is opportunity,' suggesting that automating tedious, repetitive tasks in legacy systems offers the most immediate value for enterprises.

A16Z partners Sima Ambel and Fabrizio Serafini discuss the opportunity for startups to build on OpenAI's computer use capabilities. Sun, a former founder, advises that startups still have a moat in deep industry expertise, regulatory understanding, and go-to-market execution, even as AI labs advance. He notes that vertical industries like restaurants have slower AI diffusion, creating opportunities for startups to build specialized tools that connect disparate software via computer use. Sun emphasizes that for average small and medium business owners, understanding, connecting, deploying, and servicing AI tools remains a significant barrier that startups can address. The discussion highlights that while OpenAI provides the underlying models and infrastructure, the application layer for specific, complex workflows remains a green field for new companies.