Summary
Berlin-based B2B payments company Pliant has raised an $80 million Series C led by Seaya Nova and Speedinvest. The company plans to use the funding for expansion in North America and other markets, while opening its payments infrastructure to AI agents under defined permissions and audit controls.
From payment software to an agent execution layer
Berlin-based fintech Pliant has raised an $80 million Series C to expand its B2B payments infrastructure into a setting where software agents can initiate and manage parts of corporate payment workflows. Seaya Nova and Speedinvest led the round, with Illuminate Financial, Motive Partners and Molten Ventures also participating.
Pliant said the funding will support expansion in the United States and Canada, entry into additional markets and its artificial-intelligence roadmap. The financing comes as payment providers are increasingly positioning themselves not only as tools for finance teams, but also as infrastructure that can be called by software operating on a company’s behalf.
The strategic question behind Pliant’s move is straightforward but consequential: if an AI system can interpret an instruction, retrieve business information and move through several applications, how can it be allowed to execute a payment without exceeding the authority granted to it?
What Pliant is offering
Pliant combines commercial card issuing, payment accounts, foreign exchange and spend management on a single platform. Founded in 2020, the company is a principal member of Visa and Mastercard. Pliant said it issues commercial cards in 13 currencies across 32 countries under a European e-money licence and works with regulated partners in the UK and US. Pliant said it has more than 300 employees across eight offices.
Its customer base includes travel, insurance and AI-native companies. Banks, fintechs and software platforms also use Pliant’s infrastructure to operate their own commercial card offerings. That gives the company a dual position in the market: it provides a product for businesses managing expenditure, while also acting as a behind-the-scenes infrastructure provider for other financial and software companies.
Pliant said its own AI agents already automate recurring finance work for customers. The new focus extends that approach to external agents. Claude, ChatGPT and companies’ own automation systems can connect to Pliant through the Model Context Protocol, or MCP, and run payment workflows within broader business processes, according to the company.
Execution requires more than an API connection
There is an important difference between asking an AI system to prepare a payment instruction and allowing it to execute one. Preparing a task is primarily an information and productivity problem. Execution affects company funds, authorization, fraud exposure, financial reporting and regulatory responsibilities.
Pliant’s description of the product emphasizes permissions, controls and audit trails. Those elements are likely to be decisive for adoption. In a typical corporate workflow, an agent might read an invoice, compare it with a purchase order, check vendor details, verify a budget and submit a payment request. Each step can involve a different risk level and a different approval requirement.
A company may be willing to automate a routine payment that falls within a predefined budget and matches an approved supplier. A new beneficiary, an unusual amount, a cross-border transaction or a conflict between records may require human review. The platform therefore needs to define what the agent can access, which payment instruments it can use, when it must stop and how a person can intervene.
An audit trail is equally important. If an automated workflow produces an incorrect payment, a company needs to reconstruct which data the agent accessed, which rules it applied, what instructions it received and whether a human approved the final action. Logging is not a substitute for preventive controls, but without it, incident investigation and accountability become more difficult.
MCP can provide a standardized connection between an external model or internal automation and a payment platform. It does not, by itself, establish that a workflow is secure or compliant. Organizations still need to manage authentication, credentials, access scopes, approval thresholds, monitoring and data exposure. They must also consider model-specific risks, including prompt injection, unreliable outputs, stale vendor records and attempts to manipulate payment instructions.
Why licensing remains central
Pliant’s expansion plans also show that agent payments are not only a software integration challenge. Pliant said it operates under a European e-money licence and works with regulated partners in the UK and US. That structure reflects the fact that payment services are governed by local licensing, safeguarding, operational and compliance requirements.
The use of an AI agent does not transfer those responsibilities to the model. A company deploying the agent still has to establish who is permitted to authorize payments, how funds are protected, how suspicious activity is reviewed and how records are retained. The payment infrastructure provider, the business customer, the software platform and the model provider may each have different responsibilities, which need to be defined contractually and operationally.
Cross-border expansion adds further complexity. The US and Canada are named priorities for the new funding, while Pliant also plans to enter additional markets. Supporting companies across jurisdictions can involve different card programs, currencies, settlement arrangements, data rules and anti-money-laundering obligations. An agent may make a workflow feel global, but the underlying financial activity remains subject to the rules of the relevant markets.
Funding supports a broader growth strategy
Pliant said it doubled its business each year over the past three years and recently passed $100 million in annual recurring revenue. These figures were provided by the company and indicate that its existing B2B payments business has reached a scale that can support further product investment. They do not, however, establish that autonomous or semi-autonomous payment execution has become a mature category.
The reliability of agent-driven finance depends on more than model capability. It also depends on system uptime, data quality, exception handling, fraud controls and the ability of a company’s finance organization to review automated decisions. A workflow that works in a controlled demonstration may behave differently when it encounters duplicate invoices, incomplete information, changed supplier details or conflicting approval rules.
For Pliant, the advantage of controlling several components of the payment stack may be that it can coordinate card issuance, accounts, foreign exchange and spend controls within one operating environment. The challenge is that the same breadth increases the consequences of an incorrectly authorized action. The more financial functions an agent can reach, the more carefully its boundaries must be designed.
Payments are a demanding test for AI agents
AI agents can already assist with customer support, document processing and software navigation without directly changing ownership of funds. Payments are a more demanding use case. An agent must correctly identify the payee, amount, currency, timing, budget and approval path, then stop when the transaction falls outside the rules.
That makes B2B payments an attractive but high-risk test for enterprise AI. If the workflow is reliable, it can connect procurement, travel, insurance, subscriptions and supplier operations to the payment layer. Finance teams could spend less time on repetitive data entry and more time on exceptions and oversight. But the benefits depend on whether automation can be made observable and reversible enough for a business to trust it.
The development also points to a shift in competition. Payment providers may increasingly compete on the quality of their interfaces for software agents, the precision of their permission systems and the usefulness of their audit data, rather than only on card products or human-facing dashboards. Banks, fintechs and enterprise software vendors could all become users of such infrastructure, as Pliant’s customer profile suggests.
At the same time, the responsibility for an agent’s error remains unresolved in many real-world deployments. If an agent pays the wrong supplier because of corrupted source data, an injected instruction or an overly broad permission, responsibility could involve the deploying company, the payment provider, the software platform and the model provider. Clear controls and contracts will matter as much as the model’s ability to follow instructions.
What to watch next
Pliant’s financing reflects investor interest in companies connecting AI capabilities with regulated payment infrastructure. Its existing combination of commercial cards, accounts, foreign exchange and spend management provides a foundation for agent-enabled workflows. Pliant said its licensing structure and work with regulated partners also underscore that payments cannot be separated from governance.
The next indicators will be practical rather than promotional: how much authority enterprise customers give agents, how often human approval is required, how exceptions are handled, and how permissions and audit trails work in production. The company’s expansion in the US, Canada and other markets will also test whether its compliance model can scale across jurisdictions.
For the wider industry, the central question is not whether an AI agent can issue a payment instruction. It is whether a company can demonstrate that the instruction was generated from reliable data, executed within a narrowly defined authority and recorded in a way that supports oversight. Pliant’s new funding places that question at the center of its next phase of growth.
Source: link