Ampersand Makes Enterprise Systems Legible to AI Agents

Ampersand builds the integration infrastructure that lets enterprise AI agents work inside customer CRMs, ERPs, and other systems of record. The San Francisco company is led by co-founders Ayan Barua, CEO, and Lauren Long, CTO, whose careers exposed the same expensive problem from opposite sides of the software stack.

The market Ampersand serves is becoming more important as agents move from answering questions to taking actions. An agent can reason correctly and still fail in production if it reads the wrong custom field, loses authentication, or writes into a workflow it does not understand. Ampersand matters because it treats each customer's configuration as operating reality rather than noise to be normalized away.

Founded in 2022, Ampersand has moved from a developer platform for SaaS integrations toward a broader runtime for enterprise agents. The company now has $20M in disclosed funding, current customers that include 11x, Orb, Square, and Crunchbase in company materials, and an expansion plan covering deeper ERP integrations, industry-specific systems, and forward-deployed expertise.

About Ampersand

Ampersand began with a problem Ayan Barua and Lauren Long had already spent years meeting in production. Barua co-founded Siftery and later served as VP of Engineering at G2, where customer-facing CRM and ERP integrations repeatedly pulled engineers away from the product roadmap. Long built Firebase Extensions at Google and worked at Shortwave, giving her a view of the developer experience and infrastructure required to make integrations reusable.

The founders say they interviewed more than 100 companies before writing Ampersand's first line of code. The pattern was consistent: connecting to Salesforce was possible, but connecting safely to every customer's Salesforce meant understanding custom objects, renamed fields, permissions, credentials, and workflows that had accumulated over years.

That distinction shaped the company. Ampersand does not sell a generic connector and assume the customer's system matches a standard model. It gives product teams a way to declare the behavior they need while adapting the connection to each customer at runtime.

How the Ampersand Runtime Works

Developers define integrations in version-controlled amp.yaml configuration. Ampersand organizes the work around five primitives: Read, Subscribe, Search, Write, and Proxy. Its managed runtime handles authentication, token refreshes, retries, backfills, rate limits, monitoring, logs, and the customer-specific mappings that keep an integration useful after launch.

Ampersand also supplies embeddable interface components so a customer's administrator can map fields inside the software vendor's onboarding experience. The implementation knowledge stays close to the person who understands the customer's configuration, while the application developer keeps the integration logic visible in the product's code and deployment process.

The company makes part of this technical surface public. Ampersand's open-source connectors library documents provider-specific authentication and action support, while its AI SDK and MCP server expose integration tools for agentic applications. The open repositories do not replace the managed runtime, but they make the company's architectural choices inspectable by the developers Ampersand wants to serve.

Why Enterprise Agents Raise the Stakes

Traditional integration failures created support work and delayed onboarding. Agent failures can create a more immediate commercial problem because agents operate continuously and may take actions without a person checking every step. A changed schema can make a read incomplete. An expired credential can stop a workflow. A bad field mapping can turn an otherwise capable agent into a source of operational risk.

Ampersand's thesis is that enterprise customization contains meaning. A custom object may encode how a company qualifies revenue. A permission may protect a sensitive approval. A workflow may reflect years of legal, financial, or customer-service decisions. The system becomes useful to an agent only when that local structure remains legible.

That focus separates Ampersand from an integration story measured only by connector breadth. The company is betting that depth, observability, and day-200 reliability will matter more as AI products move from pilots into recurring production work.

Leadership, Traction, and Capital

The founders' backgrounds fit the problem Ampersand chose. Barua brings the operator's experience of losing roadmap time to customer-specific integration work. Long brings developer-platform and infrastructure experience from Firebase Extensions. The combination produced a company aimed at product engineers and the solutions teams who often inherit the last mile.

Ampersand names 11x, Orb, Square, and Crunchbase in current company and hiring materials. In a Bessemer founder interview, the investor reports that 90% of completed proofs of concept become paying customers. That figure is company-reported rather than independently audited, but it shows where Ampersand believes its value becomes measurable: moving an enterprise integration from evaluation into production.

The company raised a $4.7M Seed led by Matrix in 2023, then announced a $15M Series A led by Bessemer Venture Partners in October 2026. Matrix and Flex Capital returned, while Yelp, Tenacity Capital, CTO Fund, Mana Ventures, and angels joined. Bessemer partner Lauri Moore also joined Ampersand's board.

Culture and Hiring as Market Signals

Ampersand names five values: craft, customer empathy, ownership mindset, velocity, and humility. Those words matter only if they show up in operating choices, and the founders connect them directly to critical infrastructure. The company says candidates spend half a day in its office and share lunch with the team, giving both sides a closer view of how people collaborate before making a hiring decision.

The Ampersand careers page currently lists roles across engineering, customer engineering, sales engineering, marketing, and account development in San Francisco. The mix reveals the next stage of the company more clearly than a generic hiring announcement. Ampersand is adding people who can deepen the runtime, help customers implement difficult systems, explain the category, and build a repeatable commercial motion.

That expansion also exposes the central execution challenge. A platform designed to reduce customer-specific services work cannot simply rebuild a consulting business around every deployment. Ampersand has to turn what its forward-deployed experts learn into reusable software, connectors, configuration, and operational knowledge.

What Ampersand Signals for Enterprise AI

The next generation of enterprise agents will not run above the software stack as if every customer uses the same clean data model. They will operate inside systems shaped by acquisitions, compliance rules, departmental habits, customized schemas, and old decisions that still carry business meaning.

Ampersand is building for that less glamorous layer. Its opportunity is to make enterprise variation manageable without pretending it disappears. If the company succeeds, integration stops being the project that begins after an AI product wins the customer and becomes part of the product infrastructure that helped win the customer in the first place.