DevRev’s AI agent hangout targets worker productivity, data integration

Imagine a city hub where a multitude of AI agents can gather, interact, and converse. Office workers are more like sentries — seeing and interacting with the information generated by agents at the hub.

Buses move information from various “neighborhoods” to the city hub, where workers can build AI agents and skills, and use productivity applications that leverage the growing knowledge base. Those sources could be from Slack, Microsoft or even legacy systems.

That’s how Dheeraj Pandey, co-founder of DevRev, described “Computer,” which his company unveiled last month. (Pandey previously co-founded and served as CEO of Nutanix.)

“We are the bus,” Pandey said in referencing the enterprise integration approach. The “bus” term is also a reference to Tibco’s service bus, which did bidirectional data transfers between data sources.

DevRev’s Computer aims to bring connective tissue to disparate systems, effectively unlocking previously unused data — generally referred to as unstructured data — and delivering new context to the information used by companies to make key decisions.

“What you build on top of all of this is basically agents and skills. We also build apps,” Pandey said. 

Those insights are made accessible to office workers through a conversational user interface. The DevRev stack adds more context through its own reasoning, indexing, and data processing.

“Computer…unifies both structured and unstructured data with extreme depth of organization to eliminate these silos and give AI agents full context,” Pandey said.

Box, Microsoft and Google in the last year have released AI features that can conduct deep research and add context to unstructured data. Those companies are plugging AI features into their legacy strengths in storage and productivity. 

DevRev’s Computer functionality goes beyond simple systems integration, Pandey said. “We like to take the agent approach before we go to the apps. We are the only AI company that can replace apps as well,” Pandey said.

Pandey described Computer as a kind of Switzerland where all AI offerings are treated equally. There is no lock-in to specific platforms or data sources, where only a limited amount of data is shared between systems. It provides a platform where workers can draw input from all data.

DevRev’s Computer creates its own knowledge graph from the information transported by the bus to the city hub. The goal is to turn “enterprise data into like a living network that maps complex relationships between teams, customers and products,” Pandey said.

A technology called AirSync is a key component powering the technology. It is a bidirectional synchronization technology that moves information back and forth from various systems to feed the AI-native knowledge graph.

Workers build agents that access the hub through a search engine, a SQL engine, and an MCP (Model Context Protocol) gateway. Orchestrators continuously extract and load data from external systems, while agents apply reasoning to automate workflows and present information to users through a conversational interface.

“The SQL engine runs on users’ devices at the edge, enabling fast queries without relying on cloud infrastructure. That’s a design choice that makes the system more affordable and responsive,” Pandey said.

Workers need to invoke workflows via large language models (LLMs) through an engine that’s being published through MCP.  “Basically you’re taking all your automations and annotating them, labeling them, and really publishing it through the MCP protocol,” Pandey said.

For customers, it takes just “seven lines of code” to deploy Computer into a web portal or application.

“Our customer’s employees will use this through a desktop app. It will auto-upgrade whenever it has to. It basically starts with a search bar. And the search bar eventually becomes a ChatGPT-like interface.”

APIs connect to existing systems, and DevRev’s marketplace already has many connectors built in. “Once they have deployed the marketplace connectors, everything just flows in,” Pandey explained.

The product name “Computer” was inspired by Peter Drucker’s 1967 article “The Manager and the Moron.”  (In the article, the moron caricature was a computer.)

“In 1967, he was conceiving what the computer could do,” Pandey said. “And we said, look, we can look at this as a more intelligent computer. It’s a more intelligent moron, but it’s not here to take your job.”