Sri Viswanath isn't the first venture capitalist to jump back into the founder's seat. But his timing—and the check he secured—say something about where enterprise software is heading.
The former Atlassian CTO announced late last month that his Palo Alto startup, Sycamore, has raised $65 million in seed funding to build what he's calling "the trusted operating system for autonomous enterprise AI." Coatue and Lightspeed Venture Partners co-led the round, which closed March 30, 2026 with participation from Abstract Ventures, Dell Technologies Capital, 8VC, Fellows Fund, and E14 Fund. It's one of the heftier seed rounds the enterprise AI infrastructure world has seen this year, trailing only a few weeks behind Entire's $60 million raise in February for its agent-focused developer platform.
Viswanath spent six years steering Atlassian's technical ship—January 2016 through June 2022—overseeing the company's shift to the cloud while engineering headcount ballooned from around 800 to 4,000. Then he did something unusual: he went to Coatue, joining as a general partner around June 20, 2022 with a mandate to hunt for promising AI investments. He didn't stay long. Instead, he left to start Sycamore, the very kind of company he might have funded.
The founder résumé runs deep. Before Atlassian, Viswanath served as CTO and SVP of Engineering at Groupon, led mobile computing R&D at VMware, and clocked time at Ning and Sun Microsystems. He still sits on Splunk's board.
The pitch, in brief: Sycamore is positioning itself as a full-lifecycle platform for enterprise AI agents—discovery, build, deploy, observe, evolve. The company emphasizes what it calls "trust architecture," along with permissions, isolation, audit trails, and multi-agent coordination it brands as "collective intelligence." In a blog post accompanying the funding news, Viswanath framed the product thesis around trust, adaptive system generation, continuous improvement, and—there's that phrase again—collective intelligence.
Sycamore says it's working with Fortune 500 companies as design partners. No names yet, though, which is typical at this stage.
The timing isn't accidental. Enterprises are wrestling with how to govern AI agents at scale, a problem that's moved from theoretical to urgent faster than many expected. A March report from Cognizant and HFS Research argued that companies "need an agent operating system: a unified layer that governs autonomy." Oracle has already embedded agent privilege management into its enterprise applications. AWS executives spent considerable airtime at re:Invent last December talking up AI agents as enablers of end-to-end workflows.
Lightspeed, for its part, has been placing multiple bets in agent infrastructure—Reindeer AI among them. Competitors are moving quickly: Writer launched "AI HQ," an agent builder with observability tools, late last year.
The investor roster reads like a who's who of AI and enterprise software. Bob McGrew, Lip-Bu Tan, Ali Ghodsi, Okta's Frederic Kerrest, Rubrik's Bipul Sinha, Mike Knoop, and François Chollet all participated, along with a long tail of angels from across the ecosystem.
Sycamore plans to plow the capital into expanding its engineering and applied AI teams, deepening enterprise deployments, and funding research around trust architectures, memory systems, and multi-agent coordination. The company is hiring across Palo Alto—applied AI engineers, product engineers, forward deployment engineers, research scientists.
What it's not disclosing: valuation, specific customer names, or much detail on what those Fortune 500 design partnerships actually look like in practice. The company's website went live the same day as the funding announcement, March 30.
Whether Viswanath's bet on a dedicated operating system for AI agents proves prescient or premature will depend on how quickly enterprises commit to deploying autonomous systems at scale—and whether they believe they need a new layer of infrastructure to do it safely. The money suggests investors think the answer is yes. The market will have its say soon enough.
