For a procurement team, an invoice exception is more than a data mismatch. It is a phone call, a forwarded email chain, a manual lookup in a legacy ERP system, and a delay that can stretch for days. On average, 22% of enterprise invoices require this kind of manual rework, a costly friction point that has long resisted full automation [Ardent Partners, Unknown]. Fragment AI, a New York-based startup, is betting that the missing layer is not just better data, but an agent that can learn the unwritten rules of the business and act on them.
Founded in 2021, Fragment has evolved from a provider of financial infrastructure APIs into a developer of autonomous AI agents for source-to-pay workflows [TechCrunch, July 2024]. Its core proposition is an AI system that integrates directly into enterprise resource planning (ERP) software, learns from historical transactions and team behavior, and then resolves exceptions,like mismatched purchase orders or incorrect general ledger codes,without human intervention. The company has raised a total of $23 million, with a $12 million Series A led by FirstMark closing in October 2026 [Business Wire, October 2026].
The wedge: from ledger to agent
Fragment's technical foundation is a real-time, double-entry accounting ledger, an API-first system designed to give developers a reliable source of financial truth [TechCrunch, July 2024]. This infrastructure, which early customers like spend management platform Pleo used to track balances for 30,000 customers, was the initial wedge [TechCrunch, July 2024]. The company's current focus, however, is the layer above: what it calls an Autonomous Context Engine.
This engine is designed to map all related systems, documents, and,critically,the tribal knowledge of a company's finance and procurement teams into a unified model [Fragment, Unknown]. The goal is to teach an AI agent the specific business rules, approval policies, and vendor agreements that govern transactions. When an exception arises, such as a change order or an invoice discrepancy, the agent can reference this learned context, make a decision within pre-defined guardrails, and execute the correction inside the ERP. The company positions this as essential infrastructure for a future where AI agents handle more commercial operations, requiring built-in financial safety [Business Wire, October 2026].
The team and the Stripe connection
The founding team blends enterprise software experience with engineering rigor from high-stakes technical environments. CEO Thomas Neckel is a former Oracle executive and serial entrepreneur [Citybiz, October 2026]. CTO Varun Mohan, who is identified as co-founder in recent coverage, is a former engineer at Robinhood and self-driving vehicle company Nuro [Fortune, October 2026][Forbes, August 2024]. An earlier co-founder, Omi Chowdhury, was listed as CTO in the company's initial launch coverage [TechCrunch, July 2024].
Fragment's investor roster signals strong conviction in its shift toward agentic automation. Stripe led an $11 million seed round, and the Series A was led by FirstMark with participation from BoxGroup [Fortune, October 2026][Business Wire, October 2026]. The angel list includes a who's who of fintech and operator talent, such as Plaid's Zach Perret, Uber's Dara Khosrowshahi, and Coinbase's Emilie Choi [TechCrunch, July 2024].
Traction with a focused enterprise roster
By October 2026, Fragment reported having between 50 and 60 paying customers [Fortune, October 2026]. Its roster has expanded from early adopters like TruckSmarter and Nala to include more established financial and commerce platforms. The company lists Ramp, BILL, and Whop among the businesses using its platform to support "internet-scale financial systems" [AIPressRoom, Unknown].
The company's hiring plans reflect its go-to-market phase. Open roles include a Founding Agent Strategist and a GTM Engineer, indicating a push to refine and deploy its AI agents alongside customers [Fragment Careers, October 2026][Varun Mohan LinkedIn, 2026].
| Round | Date | Amount | Lead Investor |
|---|---|---|---|
| Seed | July 2024 | $9M | Unknown [TechCrunch, July 2024] |
| Seed | 2025 | $11M | Stripe [Fortune, October 2026] |
| Series A | October 2026 | $12M | FirstMark [Business Wire, October 2026] |
The path to autonomous resolution
The ambition is clear: to move from a system that flags problems to an agent that solves them. For a procurement officer or accounts payable clerk, the standard of care today is a heavily manual triage process. An exception triggers a workflow that often involves logging into multiple systems, searching for original contracts, and emailing internal stakeholders for clarification. This rework consumes an estimated 22% of the average invoice's processing timeline [Parseur, 2026]. Fragment's bet is that an AI, trained on a company's unique operational context, can autonomously navigate these steps, apply the correct business logic, and close the loop. The patient population, so to speak, is every finance team drowning in exception-driven busywork.
The most credible risk for Fragment is that it is selling a solution for a problem that enterprises are not yet ready to fully delegate. Letting an AI agent autonomously change a purchase order or correct a general ledger code requires a profound level of trust in the system's accuracy and understanding. A single high-profile error could stall adoption. The company's answer appears to be its foundational ledger, which provides a verifiable audit trail for every agent action, and a gradual, rules-bound approach to autonomy that keeps human oversight in the loop for the most critical decisions.
What to watch in the next year
The next twelve months will be about proving that autonomous resolution drives measurable ROI. Key milestones will likely include published case studies detailing time and cost savings from early adopters, and potentially an expansion into adjacent financial operations workflows. Given its recent Series A, the company is well-capitalized to invest in both R&D and sales. The market is watching to see if Fragment can transition from being a sophisticated ledger API to becoming the default nervous system for AI-driven commerce within large organizations. Its success hinges not just on the technology, but on convincing CFOs that an AI agent can be a more reliable and faster accountant than a human facing a 22% exception rate.
Sources
- [TechCrunch, July 2024] Fintech Fragment eases ledger problems, nabs $9M | https://techcrunch.com/2024/07/22/digital-ledger-fragment-9m-banks-balance-sheets/
- [Fortune, October 2026] Fragment raises $12 million to help keep AI agents from going rogue with company money | https://fortune.com/2026/10/14/fragment-raises-12-million-to-help-keep-ai-agents-from-going-rogue-with-company-money/
- [Business Wire, October 2026] Fragment Raises $12 Million Series A to Turn Agents into Financial Operators | https://www.thecanadianpressnews.ca/businesswire_press_releases/fragment-raises-12-million-series-a-to-turn-agents-into-financial-operators/article_c2e85d97-170a-5844-8530-329be650800b.html
- [Fragment, Unknown] Fragment | The AI engine for procurement | https://fragment.ai/
- [Ardent Partners, Unknown] Average enterprise invoice exception rate
- [Parseur, 2026] The industry average invoice exception rate is 22% for teams without automation
- [Citybiz, October 2026] Fragment raises $12m to put accounting controls behind AI agents | https://www.citybiz.co/article/920316/fragment-raises-12m-to-put-accounting-controls-behind-ai-agents/
- [Forbes, August 2024] Forbes Next Billion-Dollar Startups 2024 | https://www.forbes.com/sites/amyfeldman/2024/08/13/next-billion-dollar-startups-2024/
- [AIPressRoom, Unknown] Fragment supports internet-scale financial systems |
- [Fragment Careers, October 2026] Careers at Fragment | https://fragment.ai/careers
- [Varun Mohan LinkedIn, 2026] Varun Mohan LinkedIn profile | https://www.linkedin.com/in/varun-mohan/