Funding Alert: Ampersand Raises $15M Series A to Let AI Agents Actually Write Into Enterprise Systems

🌏 中文版

Deal Terms

What the Company Does

Ampersand builds integration infrastructure for AI agents — letting AI-native software safely read from and write into a customer's systems of record, like CRMs and ERPs.

Its product lets developers declare an integration once, in version-controlled code, using five primitives: read, write, subscribe, search, and proxy. At runtime, Ampersand handles each customer's quirks — custom fields, objects, permissions, workflow differences — while the customer's own admin maps fields during onboarding, so engineers don't have to hand-build each connection. The company also announced a beta of Andi, an AI integration agent that helps developers finish the implementation work needed to deploy inside a customer's environment.

Its customers are mostly AI-native SaaS companies connecting to systems like Salesforce, SAP, NetSuite, and Workday — solving the last-mile problem that stalls agents between a demo and production. Bessemer partner Lauri Moore, who joins Ampersand's board, flagged exactly this risk when announcing the round: connecting an agent to a system of record looks simple in a demo, but keeping that connection accurate across hundreds of customers is the real challenge.

What This Round Signals

What It Means for the Agent Ecosystem

Whether an agent is useful is increasingly less about model reasoning and more about whether it can safely write into a company's existing systems. This round reflects integration becoming its own infrastructure category — similar to how unified APIs (Merge, Paragon) carved out a niche around read access, except Ampersand is betting the harder, more valuable problem is the write path agents actually need.

What Investors Are Betting On

Bessemer's Lauri Moore previously built a voice-AI startup herself and has felt firsthand how an "externalized model that looks good enough" often isn't. Her thesis here is that agents make the integration problem more urgent, because an agent can't absorb a customer's custom configuration the way a solutions engineer can on the fly. The presence of a strategic angel like Yelp among the new investors also suggests buyers themselves feel this pain around agent write access.

Numbers Worth Watching

• From a $4.7M seed (led by Matrix Partners in 2023) to this $15M Series A — a 3.2x jump that tracks the company's shift from "user-facing integrations for SaaS products" to "integration infrastructure for AI agents"
• New investors include non-traditional strategic angels like Yelp, suggesting some buyers are themselves evaluating whether to build similar agent-integration layers in-house
• The company's narrative shifted from "unified API" (its 2023 framing) to "agent integration infrastructure" (its 2026 framing) while the underlying architecture — the five primitives — stayed the same, a clean example of how the AI wave is repackaging existing infrastructure companies

Watchlist Status

Ampersand is not currently on the watchlist. Recommend adding it to section D11 (Workflow Automation) — tracking whether this integration layer can hold its accuracy and margins across customers as agent volume scales up.

Today's Takeaway

Most agent-infrastructure narratives focus on making agents smarter. Ampersand's round is a reminder of a more basic constraint: an agent that's smart but can't write into a company's systems of record has no ability to act. "Integration" sounds unglamorous and old-fashioned, but it might be the last wall standing between agents and real deployment.

References

• Ampersand closes generation gap between agents and the enterprise software stack, backed by $15 million from Bessemer Venture Partners
• Ampersand Raises $15 Million Series A To Expand Integration Enterprise Infrastructure Connecting AI Agents
• Meet the founders of Ampersand: Ayan Barua and Lauren Long