Seventy-seven percent of employees say AI has increased their workload, even as 96% of C-suite leaders expect the opposite. DevRev, an enterprise software company based in Palo Alto, is shipping a major update to its Computer product around a diagnosis: the problem is not model quality — enterprise AI has no memory.
The Upwork Research Institute data underpins what many IT leaders already sense: organisations are investing heavily in AI tooling and getting more noise, not less. DevRev calls the pattern "token maxxing" — optimising for speed and output volume without building the contextual infrastructure that would make those outputs reliable. The result is knowledge workers spending hours hunting for information across disconnected systems while AI assistants repeat the same mistakes each session.
Computer, by DevRev, is the company's AI teammate for enterprise teams, and this release is built around a concept the company calls shared memory. At the individual level, Computer learns how each person works and carries that knowledge forward between sessions. At the team level, skills and agents built by one person become available to colleagues. At the organisational level, institutional knowledge is retained permanently — when a top-performing sales rep leaves, their account context stays.
Speed without the right context is just faster noise, noise that overloads humans in the loop, and eventually breaks them. Computer is built on a different philosophy: work softer. Give AI the enterprise memory and shared context it needs to perform reliably, and then let it take action. Only then will your people have the confidence that the AI they use is performing the tasks they need it to, more accurately and with less handholding.
Three new capabilities ship with this release. Multiplayer AI allows teams to share a live Computer session with full shared context, so colleagues can question, build on, and correct reasoning in real time rather than working in isolation. A new desktop app shifts Computer from answering questions to generating complete work artefacts grounded in actual business data — competitive presentations, QBR reports, knowledge base articles — in formats including PPT, PDF, and DOCX. Agent Studio gives any team the tools to build, sandbox-test, and deploy AI agents that take action across connected systems, with every step auditable and every action reversible.
More than 250 organisations now have Computer in production. BILL has recorded approximately $5 million in operational savings. India's largest airline went from kickoff to production in 14 days, choosing Computer over Salesforce Agentforce in a direct evaluation. A retail loyalty customer is saving $1.2 million annually; its sales team has reclaimed six hours a week and reported a 30% productivity improvement. Support teams at Pebl and Uniphore are resolving 85% of tickets without human involvement.
CEO Dheeraj Pandey frames the product philosophy around deliberate pace over raw throughput: "Speed without the right context is just faster noise, noise that overloads humans in the loop, and eventually breaks them. Computer is built on a different philosophy: work softer. Give AI the enterprise memory and shared context it needs to perform reliably, and then let it take action. Only then will your people have the confidence that the AI they use is performing the tasks they need it to, more accurately and with less handholding."
All capabilities announced — shared memory, multiplayer AI, the desktop app, and Agent Studio — are generally available. The product runs across web, mobile, and desktop and connects to Gmail, Outlook, Slack, Jira, Notion, Google Drive, SharePoint, and any MCP-compatible tool. Pricing is usage-based.
