San Francisco-based Ampersand announced Sunday it has secured $15 million in Series A funding led by Bessemer Venture Partners, bringing its total capital raised to $20 million, according to the company. The round included participation from Matrix Partners, Flex Capital, Yelp, Tenacity Capital, CTO Fund and Mana Ventures, though the company declined to share its valuation.
The startup tackles a problem that has become increasingly urgent as companies race to deploy AI agents: getting those agents to actually do something inside the labyrinth of enterprise software. Reading data is one thing. Writing back into Salesforce, manipulating records in NetSuite, or triggering workflows in SAP requires a different order of access and reliability.
Ampersand's infrastructure handles the messy realities of enterprise integrations at scale. The platform manages credential lifecycles, isolates data by tenant, adapts to schema changes, and maintains event subscriptions across more than 130 different systems. Lauri Moore, a partner at Bessemer, will join the company's board.
"Your AI agent is vaporware without deep integrations into systems of record," CEO Ayan Barua wrote on LinkedIn when announcing the round.
Building for execution, not just observation
The distinction matters more than it might seem. Most AI applications today can read from enterprise systems—pulling customer records, scanning invoices, summarizing email threads. But executing transactions requires write access, and that means navigating the quirks of how each platform structures its data.
Ampersand's 22-person team has built five core integration primitives: read, write, subscribe, search and proxy. The approach preserves each system's native data structure rather than forcing everything into a flattened, lowest-common-denominator format. "We don't flatten," the founders explained in a Q&A published by Bessemer on Sunday. "We read and write in each system's native structure."
That architectural choice reflects the founders' own scars from previous battles with enterprise software. Barua previously served as vice president of engineering at G2, the software review platform, and co-founded Siftery before its acquisition by G2. Co-founder and CTO Lauren Long led Firebase Extensions at Google and worked at email startup Shortwave.
The company says its customer list includes 11x, Orb and Square, which rely on Ampersand to handle actions inside Salesforce, SAP, NetSuite and Workday environments. Hatch, a company in Yelp's portfolio, also uses the platform. "With Ampersand, we're able to serve new customers previously unavailable to us because of their complex integration requirements," said John Peña, Hatch's CTO, in the announcement.
According to Bessemer, 90 percent of Ampersand's completed proofs-of-concept convert to paying customers—a conversion rate that suggests the company has found genuine product-market fit in a crowded space.
The expanding integration battleground
Ampersand isn't alone in recognizing that AI agents need better plumbing. Integration-platform-as-a-service incumbents like MuleSoft, Workato and Tray.ai have all added AI agent orchestration and governance features over the past year. Unified API vendors such as Merge offer standardized endpoints for categories like HRIS and accounting, though they typically handle fewer per-customer customizations, according to a September analysis by developer tools company Nango.
The competitive dynamics are shifting quickly. Bessemer's May 2025 roadmap on "AI systems of action" argued that legacy systems-of-record moats are eroding as agents create new entry points into enterprises. Yet those same agents still depend on deep CRM and ERP integrations to execute the transactions that matter.
Ampersand announced a beta version of "Andi" alongside the funding round. The AI integration agent helps developers build connectors and troubleshoot implementations, a nod to the reality that even infrastructure for AI agents sometimes needs its own AI assistance. The company also maintains open-source repositories on GitHub, including an AI SDK launched in September and a connectors library with conventions for managing rate limits, retries and webhook ingestion.
Barua said the startup will use the new capital to expand ERP coverage beyond NetSuite, SAP and Microsoft Dynamics. The company also plans to hire forward-deployed experts for pre-sales work and customer implementations, roles that have become critical as enterprise buyers demand proof that integrations can handle their particular edge cases before signing contracts.
The broader question is whether startups like Ampersand can maintain an advantage as larger platforms wake up to the same opportunity. For now, at least, the founders seem confident that depth beats breadth when enterprise buyers need their AI agents to execute real transactions rather than just talk about them.
