For decades, manufacturers have spent billions on software to manage purchasing, inventory and supplier payments. Yet when an order arrives incomplete or an invoice doesn't match a contract, someone still has to investigate what went wrong.
Fragment, a San Francisco startup founded last year, has raised $56 million to take that work off their hands.
The funding includes a Series A led by 8VC and an earlier seed round co-led by General Catalyst and A*. Former Sequoia leader Roelof Botha joined as an angel investor, alongside Palo Alto Networks CEO Nikesh Arora and former Amazon executive Jeff Wilke.
Fragment is targeting the messy, exception-heavy operations that conventional enterprise software struggles to automate. Its initial focus is direct procurement, the process through which manufacturers purchase the materials and components needed to produce goods.
Consider an industrial distributor that receives a shipment with fewer components than ordered. Resolving the discrepancy might involve checking purchasing records, supplier contracts, delivery schedules and invoices stored across different systems.
Fragment's AI agents are designed to investigate those records, work out what happened and carry out an authorized resolution.
Its underlying technology, called the Autonomous Context Engine, connects operational data with company policies and internal knowledge. The aim is to give agents enough context to make decisions across existing enterprise systems, including established platforms such as SAP.
The company is already working with customers including industrial distributor Graybar and energy producer CNX. Graybar processes roughly three million invoices annually, and around 80% require some form of resolution, according to Fragment.
That illustrates the scale of the problem. Even modest improvements in how quickly those exceptions are handled could eliminate substantial administrative work.
Fragment was founded by Pranav Mulgund and Naveen Kashyap. Mulgund previously held product roles at Mindy and Zoom and grew up in a manufacturing family. Kashyap leads the company's AI architecture.
The founders are betting that enterprise software is entering a new phase. Systems such as SAP and Coupa transformed how companies record transactions, but much of the work needed to resolve problems between those transactions remains manual.
Fragment wants to automate that second layer. That is close to the problem CarbonAI is attacking in commercial real estate lending, where documents must be reconciled before a loan can close.
There is considerable competition in enterprise AI, and the company's technology will need to handle business-specific rules, incomplete records and supplier exceptions reliably. Its customer claims are promising, but the company has not disclosed revenue or independent benchmarks of operational savings.
For now, procurement is its proving ground. If its agents can resolve complex purchasing problems across existing software, Fragment could expand into other operational functions without requiring enterprises to replace the systems they already use.