Danske Bank Lets Business Customers Bank Through AI Agents

On 21 October 2026 Danske Bank announced that business customers can now bank through AI agents such as Claude and ChatGPT. Danske says it is the first bank in the Nordic countries to offer this. A finance team can ask its AI assistant for today's balances or a cash-flow forecast and get the answer from its own bank data.

The main points:

1. Customers bring their own AI tool. Business customers can connect the AI solutions they already use, such as Claude or ChatGPT, to their banking data. Danske presents this as giving customers more control over how and where they use their financial data.
2. Questions instead of portals. Customers can ask questions, create tailored financial views and, "ultimately, take action with a simple prompt". Danske's examples include a summary of account balances, analysis of recent transactions, a cash-flow forecast, and information from several APIs combined in one interaction.
3. MCP is the bridge. A Model Context Protocol (MCP) server connects the customer's AI solution to Danske Bank's APIs. Customers no longer need to switch between bank portals, treasury systems, accounting platforms and spreadsheets.
4. Same infrastructure as other integrations. The agent connects through the same secure, authenticated infrastructure as other customer integrations. Access to a customer's own data requires onboarding, certificates, credentials and subscriptions.
5. Help for teams without developers. The agent can help customers explore Danske's APIs, read the documentation, test the APIs with mock data and set up the connection in their AI tool, including customers without traditional development expertise.

Danske's developer documentation adds two points worth knowing. For now, agents can only read data in production. And the customer is responsible for what its agent does: the bank does not review, approve or monitor the agent's actions.

Starting with business customers and with reading data is a sensible first step. Treasury teams have mandates and approval workflows, and reading balances and transactions is useful from day one with limited risk. The harder questions come when agents start to act: how consent and strong customer authentication work when an agent initiates a payment, and who is liable when it gets something wrong. EU payment law has no specific rules for AI agents yet. In the UK, the FCA has said it will work on a trusted framework for agentic finance.

Sources

• Danske Bank, "Danske Bank becomes the first bank in the Nordic countries to enable business customers to bank through AI agents", press release, 21 October 2026
• Danske Bank Developer Portal, MCP Service
• Osborne Clarke, "Agentic Payments: a new challenge for Europe's payments ecosystem", 6 March 2026

Frequently asked questions

What has Danske Bank launched?

A service that lets business customers connect AI agents such as Claude or ChatGPT to their banking data, through a Model Context Protocol (MCP) server linked to Danske Bank's APIs. Danske says it is the first bank in the Nordic countries to do this.

What can business customers do with it?

Danske's examples include a summary of account balances, analysis of recent transactions, tailored financial views, cash-flow forecasts, and information from several APIs combined in one interaction.

Can an AI agent make payments through Danske Bank?

Not yet. Danske says customers will ultimately be able to take action with a simple prompt, but according to its developer documentation, agents can for now only read data in production.

Who is responsible for what the AI agent does?

The customer. Danske's developer documentation says the bank does not review, approve or monitor the agent's actions, and that the customer is responsible for them and for their consequences.

What is the Model Context Protocol (MCP)?

An open standard for connecting AI assistants and agents to other systems, such as a bank's APIs, so they can look up information and, where allowed, take actions there.