Fragment raises $12 million to help keep AI agents from going rogue with company money

Keeping the books straight is a business headache.

As companies turn to AI to automate routine financial tasks, they face a new worry: whether they can trust bots with company money.

Fragment, a New York-based startup, is developing software designed to reduce costly accounting errors.

Founded by serial entrepreneur and CEO Thomas Neckel and CTO Varun Mohan, a former Robinhood engineer, Fragment announced on Wednesday that it raised a $12 million Series A round that closed in mid-2025 to build software that lets AI agents safely handle billing and accounting.

Early-stage venture capital firms FirstMark and BoxGroup co-led the financing.

The round follows an $11 million seed round led by payments company Stripe, with participation from BoxGroup and tech executives including Uber CEO Dara Khosrowshahi and Plaid CEO Zach Perret.

Neckel declined to disclose the company’s valuation following the Series A.

It works like this: Developers plug Fragment’s tools into their software.

The system uses double-entry accounting, a centuries-old method that records where money comes from and where it goes, to check that the numbers balance before money moves.

“Just like Claude Code or Cursor is a set of software blocks for agents to automate coding, we are a set of financial building blocks or financial primitives to automate commerce,” Neckel told Fortune .

Fragment’s push is part of a broader shift toward agentic commerce, in which AI agents shop and make payments on users’ behalf.

In June, Visa announced a partnership with OpenAI to let agents make purchases within spending limits set by users, while Mastercard introduced a service that lets AI agents more easily pay each other.

Three months earlier, Stripe and Tempo launched a shared standard that lets agents pay to access online tools and data without a person completing checkout.

But giving AI agents more control can also expose businesses to costly mistakes.