DevRev’s newest version of its AI teammate, Computer, is built around a different principle: clarity and safe action, not just raw speed. The enterprise software company, founded in 2020 by former Nutanix leaders Dheeraj Pandey and Manoj Agarwal, argues that the industry’s obsession with faster outputs and higher token volumes – a practice it calls “token maxxing” – is creating a new workplace crisis rather than solving old ones.
The productivity paradox
The problem, according to DevRev, is that most enterprise AI tools lack three things: memory of a business, persistent insights, and the ability to take safe action. The result is a widening gap between executive expectations and ground reality. Research from the Upwork Research Institute found that while 96% of C-suite leaders expect AI to boost worker productivity, 77% of employees report that AI has actually increased their workload. Pandey, DevRev’s co-founder and chief executive, said: “Speed without the right context is just faster noise, noise that overloads humans in the loop, and eventually breaks them.”
Other surveys paint a similar picture. PwC UK found that 57% of UK workers believe generative AI could improve efficiency, yet 41% say their workload has risen. EY reported that 83% of UK employees now use generative AI, but 40% of potential productivity gains are lost to poor training, weak culture and misaligned incentives, with 62% of UK respondents saying their workload has increased. Workday research suggested UK employees lose more than seven hours a week simply managing disconnected AI systems, creating what it called a “copy-paste economy”. DevRev’s approach is to reject the speed-first philosophy and instead build what it calls “shared memory”.
Shared memory: the architectural foundation
The most important element in Computer’s latest release is shared memory – a curated, living picture of an organisation’s data, how it works and how its people interact. DevRev describes this as the architectural foundation that enables precision, efficiency and safety. It is powered by the company’s patented “Computer Memory” knowledge graph, which compounds over time at three levels: individual, team and organisation.
At the individual level, Computer learns how each person works, picking up where they left off with every new session. At the team level, skills and AI agents that one person develops become available to everyone. At the organisational level, institutional knowledge stays in the system permanently – when a top-performing employee leaves, their account knowledge does not leave with them. This persistent, compounded memory means that answers are sourced from real business data, cited and referenced so teams can stand behind them. It also means lower token usage and full context on the first response, because the AI does not start from scratch each time. And crucially, nothing goes out before a human approves it, with full audit trails and the ability to undo any action an agent takes. “When AI has this context, it stops guessing and starts acting like a trusted member of the team,” the company says.
DevRev, which is headquartered in Palo Alto and has raised approximately $158 million from investors including Khosla Ventures and Mayfield Fund (achieving unicorn status after a $100.8 million Series A in August 2024), is targeting a market crowded with competitors such as Salesforce, HubSpot and Zendesk. It positions Computer as a replacement or complement to traditional CRM and developer tools, and claims that the difference is built into the architecture.
New capabilities: from Q&A to content creation and multiplayer collaboration
The latest release introduces several new features that build on shared memory. Before this version, AI insight disappeared the moment it left one person’s screen. Now, with Multiplayer AI, teams can share a live Computer session where everyone sees the full context and continues the analysis together. Colleagues can question, build on and correct reasoning in real time. This shifts the unit of attribution from “what I did with AI” to “what we did with AI, together” – a response to findings from a 2025 study by KPMG and the University of Melbourne, which reported that 57% of employees admit to using AI in non-transparent ways, including hiding its use from employers. That same study found that while 66% of people intentionally use AI, less than half (46%) are willing to trust it, and 70% are unsure whether online content can be trusted.
Earlier versions of Computer answered questions and took single-step actions. The new desktop app shifts the tool into a content-producing system. From an in-app canvas, any user can generate complete, fully branded work artefacts grounded in real business data: competitive slide decks, quarterly business review reports, structured dashboards, knowledge-base articles and multi-step workflows. Outputs are available in formats including PPT, HTML, PDF and DOCX. Skills and outputs built in the canvas are saved at user, team or organisation level and become reusable across the business. Agent Studio gives any team the ability to build, test in a sandbox environment and deploy AI agents that take action across connected systems. Every action runs under individual user permissions, not a shared account, and every step is traceable, auditable and reversible – if an agent makes a mistake, it can be rolled back.
Additional capabilities include Text2SQL, which allows analytical queries across structured data in plain language without a data analyst, and connectors for Gmail, Outlook, Slack, Notion, Google Drive, Jira, Microsoft OneDrive, SharePoint and any MCP-compatible tool. Pricing is usage-based, scaling with adoption rather than headcount. The product is available across web, mobile and desktop, and existing DevRev customers can access Computer directly within their current workflow without switching surfaces. “Computer Pro” is available without a waitlist, and a free tier called “Computer Mini” is in open beta, having first launched in beta for existing customers on 9 September 2025 with a broader public release planned later.
Customer results in production
More than 250 organisations have Computer live in production, with over 1,000 users onboarded since September 2025. Customers include BILL, HDFC Bank and FAME, spanning financial services, aviation, retail and technology. The company reports that customers Pebl and Uniphore are resolving 85% of support tickets without any human involvement. BILL has achieved around $5m in operational savings. India’s largest airline went from kickoff to production in 14 days and selected Computer over Salesforce Agentforce in a head-to-head evaluation. A retail loyalty customer is saving $1.2m annually, with sales reps reclaiming six hours a week and the team reporting a 30% productivity boost. FAME is saving users more than 10 hours a week, resolving tickets approximately 40% faster and accelerating specific workflows by up to 75%.
HDFC Bank, India’s largest private-sector lender which serves over 120 million customers, has been integrating AI for years. It began its AI journey in 2017 with a rule-based chatbot called EVA and is now layering generative AI, including a dedicated GenAI Academy, to support an “AI-first” strategy for conversational customer experience, real-time risk management and internal productivity. Uniphore, itself a “Business AI Company” with a platform spanning agents, models, knowledge and data, is a customer of DevRev as well as a competitor in the broader AI market. The airline that chose Computer over Salesforce Agentforce is one of several examples DevRev cites to illustrate the speed of deployment and the ability to replace incumbent tools.
