Former Agentforce team raises $5.1M for anti-vendor AI startup
▼ Summary
– London startup Zaro raised a $5.1mn pre-seed round led by Cherry Ventures, with angel investors including Hugging Face co-founder Thomas Wolf and GitHub chief Thomas Dohmke.
– Five of Zaro’s eight staff previously built AI agents at Convergence, which was acquired by Salesforce; the team helped ship Salesforce’s Agentforce product.
– Zaro aims to replace separate AI tools with a single workspace where a company owns its data and context, rather than feeding a vendor’s data layer.
– The platform claims to cut costs roughly 10-fold by routing routine tasks to cheaper models, and already runs its own HR, finance, and facilities on it.
– Zaro enters a market where buyers question per-seat SaaS models, but it is an eight-person startup with untested claims and a disputed framing of Agentforce’s revenue.
A London-based startup, Zaro, has officially exited stealth mode with a $5.1 million pre-seed round led by Cherry Ventures. The company’s mission is to build a single AI workspace that remains under the full ownership of the business using it, rather than being controlled by software vendors.
The investor lineup is unusually senior for such an early stage. Backers include Hugging Face co-founder Thomas Wolf, GitHub chief Thomas Dohmke, investor Charlie Songhurst, Trouva’s Mandeep Singh, and Convergence co-founders Marvin Purtorab and Andy Toulis. Cherry Ventures committed to the round before Zaro even had a product, and featured the startup at its London summit this Tuesday.
Five of Zaro’s eight employees previously built AI agents at Convergence, a startup acquired by Salesforce just 11 months after its founding. The team then helped ship Agentforce, Salesforce’s flagship AI agent product. CEO Michael Bajwa, who was Convergence’s first hire, says he grew the company from zero to £1 million in annual recurring revenue within 10 weeks.
Inside Salesforce, the founders say they spotted a critical flaw. Companies using Agentforce fed their data into Salesforce’s own layer, not their own. Every interaction, they argue, built intelligence inside the vendor, not the customer. “We built agents that worked flawlessly in isolation, and watched them fail collectively,” Bajwa explains. “The intelligence never compounds because the context never carries over.”
Zaro’s solution replaces a chaotic mix of separate AI tools with a single workspace. Agents contribute to a shared context layer, which then powers custom applications built from a company’s own files and decisions. The key pitch is ownership: the company retains its context and can take it elsewhere. Zaro also routes routine tasks to cheaper models, claiming it cuts costs roughly 10-fold compared to frontier-only setups.
The startup already runs its own HR, finance, and facilities on the platform, and ships with an app store of ready-made workflows. “Context compounds,” says co-founder and CTO Qian Zheng. “Models commoditise. The platform does not.”
Zaro enters a jittery enterprise-software market. Buyers are questioning whether per-seat SaaS can survive as AI agents become more common. Incumbents from Salesforce to SAP are racing to bolt agentic layers onto their suites. Zaro joins startups like the UK’s Deliverance AI in selling one core idea: companies, not vendors, should own and govern their AI.
Still, several claims need testing. Zaro has just eight employees and has been public for only days. Its cost and context figures are self-reported. One framing also needs careful handling: the team contributed to Agentforce after the acquisition, rather than building it alone. Zaro’s release calls Agentforce a $1.2 billion ARR product, while Salesforce’s own figures and earlier reporting place it closer to $800 million.
The core idea is simple. A company should keep the AI context it creates. Whether Zaro can deliver on that promise before a larger rival copies it remains the open question.
(Source: The Next Web)