Ex-Agentforce Team Raises $5.1M for an ‘Anti-Vendor’ AI Workspace

The more AI tools a company adopts, the more its institutional intelligence scatters across disconnected applications. A London startup founded by two engineers who saw this fragmentation from inside Salesforce just raised $5.1 million to solve it.

Zaro, which emerged from stealth in June 2026, is building what it calls an AI-native workspace — a single environment designed to sit underneath all the disparate AI tools a company uses . Led by Cherry Ventures, the pre-seed round drew a heavyweight angel list that underscores how seriously the enterprise world is taking the platform problem .

The problem: AI tools that forget everything

Enterprises are adopting AI agents and assistants at record speed, but each new tool arrives with its own isolated memory. A decision made in one workflow gets lost when a different assistant takes over. Data trapped in one application never informs the next task. The result is exactly the opposite of what AI promised: organizations become less coordinated, not more .

Cherry Ventures described the core issue bluntly: Zaro is building “shared memory for AI agents” — a layer where company knowledge compounds for the business itself rather than for any individual software vendor .

How Zaro’s platform works

Zaro’s platform operates as a persistent context layer that every connected agent and application can read from and write back to. Rather than asking companies to rip out their existing tools, Zaro sits underneath them and does a few specific things :

• A single shared workspace holds company data, decisions, workflows, and operational history in one accessible place. When an agent completes a process, the context it produced becomes instantly available to the next agent — not lost in a silo .
• Smart model routing sends routine, low-stakes tasks to cheaper AI models and automatically escalates complex or high-value workloads to advanced frontier models, helping enterprises control costs without sacrificing performance on the work that matters .
• Custom applications and pre-built workflows let teams build their own tools on the shared context layer or deploy ready-made workflows, tackling fragmentation from the bottom up rather than patching it with more standalone apps .

The pitch is deliberately different from most enterprise AI startups. Where typical vendors sell another tool that captures more company data and locks it inside a new interface, Zaro’s “anti-vendor” position argues that the company — not the software provider — should own the workspace where its intelligence lives .

Who built Zaro — and why the team matters

Zaro was co-founded by Michael Bajwa (CEO) and Qian Zheng . Both were the very first employees at Convergence, the London-based AI agent startup that Salesforce acquired in May 2025 to power its Agentforce platform .

Convergence, founded in April 2024 by Marvin Purtorab and Andy Toulis, had raised a $12 million pre-seed round led by Balderton Capital before Salesforce closed the deal on June 11, 2025 . Bajwa and Zheng spent the next year helping build and ship Agentforce from inside Salesforce — and that’s where they saw the enterprise fragmentation problem firsthand .

They left to start Zaro. Today, the team is eight people, and five of those eight previously built AI agents at Convergence before its acquisition . Two of the most notable angel investors in Zaro’s round are Convergence’s own co-founders, Purtorab and Toulis, who now run Agentforce inside Salesforce .

The round’s other angels include Hugging Face co-founder Thomas Wolf, GitHub CEO Thomas Dohmke, investor Charlie Songhurst, and Trouva’s Mandeep Singh — a senior backer list for a pre-seed stage European startup .

What the $5.1 million will be used for

Zaro plans to use the capital to expand its engineering team, accelerate product development, and onboard early enterprise customers . The company remains in its earliest commercial phase, but the combination of a tight team with direct experience shipping an enterprise-scale AI agent platform and the backing of both the investors who financed that platform and the operators now running it suggests the market is paying attention.

As of this writing, Zaro’s workspace is an early tool in a category that barely exists yet: AI operating environments that enterprises own and control themselves rather than licensing from a growing stack of autonomous tools . Whether the product can deliver on its architectural vision remains to be seen, but the funding — and the people betting on it — signals that the next layer of enterprise AI infrastructure is being built right now.