The Pulse
Ampersand Raises $15M to Connect AI Agents to Enterprise Systems
Ampersand raised a $15 million Series A led by Bessemer Venture Partners to help AI applications read and write data in customized enterprise systems. The San Francisco company also announced a beta release of Andi, an AI agent for integrat
AI.info Team · Written with AI ·
Ampersand announced a $15 million Series A on October 6, 2026, to help AI applications work directly with the customized customer databases and software that many companies already use. Bessemer Venture Partners led the round; existing investors Matrix and Flex Capital also participated, alongside new investors including Yelp, Tenacity Capital, CTO Fund and Mana Ventures. Bessemer partner Lauri Moore will join Ampersand’s board.
The San Francisco company is building infrastructure that lets enterprise agents read and write within systems such as customer relationship management (CRM) and enterprise resource planning (ERP) software. Alongside the funding, Ampersand announced the beta release of Andi, an AI integration agent intended to help developers with the work of adapting their products to each customer’s environment.
Customer-specific software is the integration problem
Connecting to a CRM in a demonstration is not the same as keeping an integration working across a customer base. Companies often customize fields, objects, permissions and workflows, so the same software can present different data structures to each customer. Vendors building AI applications for businesses must account for those differences before their agents can reliably reach the records they need.
Ampersand says its approach lets each customer use a custom data model, giving agents context about those fields and permissions while supporting actions in both directions. That distinction matters when an agent must not only retrieve information but also update a record: an incorrect write can affect a company’s operational data. The company is positioning its product as a way for software vendors to handle customer-specific integrations without rebuilding the connection for every deployment.
Andi targets the work before an agent goes live
The Andi beta addresses implementation work, rather than serving as the enterprise agent that uses a customer’s CRM or ERP. Ampersand says the tool helps developers get their products running in individual customer environments. The announcement does not provide beta access terms, pricing or a general availability date.
Company co-founder and CEO Ayan Barua said Ampersand is designed to sit between established systems and newer AI software. The release describes a broader offering that combines integration infrastructure with domain and forward-deployed expertise, as well as agent support. Its founders, Barua and Lauren Long, spoke with more than 100 companies building integrations before writing Ampersand’s first line of code, according to the announcement.
Bessemer points to the maintenance burden
Moore framed the difficulty as a runtime problem: integrations can break when credentials expire or customer schemas change, even after an initial connection works. “Connecting an agent to a system of record is a demo. Keeping that connection alive across hundreds of customers, with writes you can't afford to get wrong, is much harder. That's what Ampersand is building.”
A customer cited in the announcement described the commercial impact of complex integrations. John Pena, CTO at Hatch, an AI communications platform for service businesses acquired by Yelp in 2026, said Ampersand helped the company serve customers it previously could not support. “Now we've expanded our total addressable market without taking on months of maintenance headaches,” Pena said.
The $15 million round backs Ampersand’s effort to make customer-specific CRM and ERP connections part of the product-development workflow, rather than a separate project each time a vendor signs an enterprise customer. The release names Salesforce and NetSuite as examples of systems that are not going away, and describes Andi as being in beta; it gives no funding valuation or product pricing.