Running an AI agent as a demo is one thing; keeping it running continuously without blowing up the cloud budget is another. That realization is what convinced a group of leading investors to inject $35 million into the American startup Sapiom, which specializes in the economic management of AI agents in production.
A Funding Round That Confirms the Urgency of Cost Control
Contents
Led by the fund Dragonfly, this Series A round brings together an impressive lineup of investors: Accel, Gradient, Coinbase Ventures, Operator Collective, Formus Capital, and VanEck Ventures, alongside existing investors such as Okta Ventures, Menlo Ventures, Array Ventures, and Anthropic. It comes just six months after a $15 million seed round, bringing the company's total funding to $50 million in less than a year of existence.
The amount speaks volumes about the scale of the problem identified: as AI agents move from prototype to production, their repeated calls to language models, tools, and third-party services can drive up cloud bills in an uncontrolled way, without technical teams having clear visibility into the spending.
Routing Every Agent Decision to the Most Cost-Effective Option
Sapiom's platform acts as a real-time router: for every task assigned to an agent, it automatically selects the model, computing power, and tools best suited to the job, taking into account cost, expected quality, latency, reliability, and the company's internal rules. According to the company, this approach can cut an agent's execution cost by a factor of ten in some cases.
Since its launch six months ago, the platform claims to have processed more than 270 million transactions and handles more than 100,000 agent runs per day. One client reportedly cut its inference costs by 75% after migrating its agents to Sapiom's infrastructure.
Why This Isn't Just a Dashboard Problem
For Haseeb Qureshi, a partner at Dragonfly, the real challenge can't be solved with just another reporting tool: agents often operate without a defined budget, while technical leadership allocates funds without visibility into their actual use. According to him, it's an infrastructure problem that needs to be addressed at the exact moment the agent acts and the money is spent, not after the fact through a dashboard.
This observation reflects a broader trend seen among companies experimenting with AI agent use cases: governance and spending control are becoming as significant a bottleneck as technical performance itself when it comes to moving from pilot to large-scale production.
What This Means for Businesses
- Real-time visibility into the cost of every decision made by an agent, rather than a global bill discovered after the fact.
- The ability to set governance rules directly at the execution level, rather than after the fact through monitoring tools.
- A strong signal to the market: investors are now betting as much on agent control infrastructure as on the agents themselves.
This type of tool is part of a rapidly maturing ecosystem, also reflected in the comparisons of AI agent tools and software now available to companies looking to scale up their deployments.
Frequently Asked Questions
What exactly does Sapiom do?
Sapiom offers a platform that directs every action of an AI agent, in real time, to the most suitable model, tool, or computing resource, optimizing cost, quality, and reliability according to rules defined by the company.
Why are AI agents so expensive to run in production?
An autonomous agent makes repeated calls to language models and external services to complete a task, which can lead to an unpredictable cloud bill if those calls aren't optimized or governed by management rules.
Who funded this round?
The $35 million Series A round was led by Dragonfly, with participation from Accel, Gradient, Coinbase Ventures, Operator Collective, Formus Capital, and VanEck Ventures, along with existing investors including Anthropic and Menlo Ventures.
Sources
FrenchWeb —
PYMNTS.com —
TheNextWeb
Author Profile
Latest entries
Planet15 November 2026CO2 Emissions: Why Transportation Remains France's Climate Blind Spot
Artificial Intelligence12 November 2026Imagine Enterprise: IBM and Microsoft Team Up on AI Agents for Large Businesses
Economy11 November 2026Livret A Rate: Understanding How It's Calculated and Its Limits
Shopping8 November 2026Black Friday: Key Dates and the Budget French Shoppers Are Planning
This article was written with the help of artificial intelligence. Editorial policy









