Customer support agents are increasingly moving beyond answering questions. Plaid and Decagon are bringing bank-account connectivity directly into AI-powered customer support conversations, letting users securely connect or reconnect financial accounts without ever leaving the chat.

This partnership embeds Plaid Link into Decagon’s AI agents, enabling consumers to complete bank-verification and account-connection steps within a support interaction. The companies say the approach can help resolve issues involving failed payments, disputed charges, fraud cases, insurance claims, and loan applications.

This development points to a broader shift in financial-services support: AI agents are moving from providing information toward helping customers actually complete the underlying financial task.

Putting bank connectivity inside the conversation

A common problem in financial customer support happens when an issue requires access to a customer’s bank account. A support agent may be able to explain why a payment failed, for example, but resolving the problem can require the customer to authenticate with their bank. Traditionally, that means leaving the support conversation, opening another application or website, and then returning to the original interaction.

Plaid and Decagon are attempting to remove that break in the journey. With Plaid Link integrated into Decagon’s AI agents, customers can get prompted to connect or reconnect their bank accounts from within the conversation itself. The announcement describes this as consumer-permissioned access, with users controlling the permissions granted through Plaid.

Decagon’s own announcement says the integration lets customers securely connect their bank accounts within a conversation and take action on bank connections, failed payments, and disputed charges, all in one place.

AI support agents can now handle more financial tasks

This partnership matters because bank connectivity is often a prerequisite for completing a financial-service workflow. A customer applying for a loan, for example, may need to verify an account before the application can progress. An insurance customer may need to provide banking information as part of a claim. A consumer reporting a payment problem may need to reconnect an account before the issue gets resolved.

Instead of stopping the conversation at that point, the AI agent can initiate the bank-connection process directly. Decagon says the integration can support longer-running tasks, letting an agent resume a claim, application, or dispute with the relevant context after the initial interaction ends.

That changes the role of the support agent considerably. Rather than simply telling a customer what to do next, the AI system can potentially guide the customer through the required financial action and continue the workflow from there.

Plaid Link adds a permission layer

Security and consumer control sit at the center of this integration. Customers grant specific permissions through Plaid, and those connections can be managed or revoked through Plaid Portal or the relevant service provider. Decagon also provides auditable agent logic and monitoring designed to keep autonomous interactions within defined guardrails.

Plaid’s own materials similarly emphasize permission and consumer control when financial accounts get connected to AI experiences. The company says consumers choose whether to connect their accounts and can manage Plaid-powered connections through Plaid Portal at any time.

This permission model becomes particularly important as AI agents gain the ability to do more than simply provide information. An AI system that can access financial information needs clearly defined authorization. An AI system that can initiate or support financial actions needs additional controls around what it can do, when it can do it, and what information it can access.

From chatbots to transactional AI agents

This partnership illustrates a broader evolution happening in customer-service technology. Traditional chatbots primarily answer frequently asked questions. More advanced AI agents, by contrast, can maintain context, interact with enterprise systems, and execute multi-step workflows.

Financial services make for a particularly important testing ground, since many customer-service requests connect directly to actual financial events. A customer may not want an explanation of why a payment failed. They want the payment problem fixed. A borrower may not want instructions on how to verify an account. They want the verification completed. An insurance customer may not want another form to fill out. They want the claim to move forward.

By combining Decagon’s AI support layer with Plaid’s financial connectivity infrastructure, the companies are targeting that exact gap between answering a question and completing the task.

Multi-step workflows could run across days or weeks

Another notable feature here is persistence. Decagon agents can maintain context across multiple sessions, letting them manage financial tasks that extend over days or weeks. That could include insurance claims, loan verification, and dispute resolution.

This differs meaningfully from a conventional one-session chatbot. A longer-running AI agent can potentially remember where a customer stopped, identify the next required step, and continue the workflow once the necessary information becomes available.

For financial institutions, that could reduce how often customers need to repeat information or restart a support interaction from scratch. The effectiveness of this model will ultimately depend on how institutions configure permissions, escalation processes, and human review.

Plaid expands its role in AI-powered finance

This partnership also fits Plaid’s broader push into AI-enabled financial experiences. In September, Plaid announced an integration with Meta’s Muse AI agent, letting consumers securely connect financial accounts and use financial information within conversations. Plaid said the integration can provide access to balances, transactions, investments, mortgages, and other financial information based on consumer permission.

Plaid also announced in June that it was integrating Plaid Link directly into Fin, another AI customer-service agent, so users could connect or reconnect bank accounts without leaving the conversation there either.

This Decagon partnership therefore extends a broader direction for Plaid: becoming financial connectivity infrastructure for AI experiences generally, rather than limiting account connections to conventional fintech applications alone.

Decagon targets financial services support

Decagon is also expanding its presence in financial services specifically. The company says financial services is one of its fastest-growing verticals, with AI agents already in production at companies including Square, Affirm, Chime, and Valon.

That focus reflects the particular value of automation in financial customer support. Financial institutions handle large volumes of repetitive requests, but many of those requests also involve sensitive data and consequential actions. An AI agent therefore needs more than natural-language capability. It needs access to the right systems, clear permissions, monitoring, and escalation mechanisms built in from the start.

This Plaid integration addresses one of those infrastructure gaps directly, by bringing bank connectivity into the support environment itself.

Security becomes more important as agents gain access

The move from conversational AI to transactional AI creates a much larger responsibility for financial institutions. An AI agent that only answers a question has limited ability to affect a customer’s financial position. An agent that can connect an account, check transaction information, or support a payment workflow carries a much greater operational role.

That makes permissioning and auditability absolutely critical. This partnership’s model keeps the consumer involved in granting access through Plaid Link, while Decagon says its agents operate with guardrails and auditable logic throughout. The wider industry will likely focus increasingly on these controls as AI agents gain more authority inside financial workflows.

The bigger shift in financial customer support

This Plaid-Decagon partnership highlights a broader change in how financial institutions can think about customer service. The objective is shifting from a simple model, where the customer asks and AI answers, toward something more complete: the customer asks, AI authenticates, AI accesses authorised data, and AI completes the workflow.

That distinction matters quite a bit. The value of an AI support agent isn’t necessarily measured by how naturally it communicates. In financial services, the bigger question is whether it can safely resolve the underlying problem. Bank connectivity is one of the infrastructure layers needed to make that possible.

What comes next

The initial use cases focus on bank connections, failed payments, disputed charges, and longer-running financial workflows. Decagon says the partnership is designed to let agents help customers complete these processes without repeatedly sending them outside the conversation.

As AI agents gain access to more financial infrastructure, additional use cases could emerge across lending, insurance, payments, and banking support. Broader adoption, however, will depend on security, customer consent, institutional risk policies, and the ability to demonstrate that automated actions remain within clearly defined boundaries.

For financial institutions, the opportunity is significant: fewer disconnected support journeys and more tasks resolved in a single interaction. But the central requirement remains trust, above everything else.

The bottom line

Plaid and Decagon are bringing bank verification into AI customer support, giving users a way to connect financial accounts through Plaid Link without leaving the conversation. The bigger development here isn’t simply another AI chatbot integration. It’s the movement of AI customer support toward transactional workflows, where agents use permissioned financial connectivity to help resolve real problems.

As AI agents gain access to increasingly sensitive financial infrastructure, the combination of consumer permission, controlled access, monitoring, and auditable actions will become just as important as the conversational experience itself. The direction of financial customer support is becoming clearer: AI agents are being built not only to explain what customers should do, but increasingly to help them get it done.

Plaid and Decagon partner to create more helpful financial experiences for customers