Thanks to “AI City”, all Danske Bank employees can now create AI agents without needing to know how to code

Danske Bank has built an internal AI framework called "AI City" that lets any employee create AI agents without programming skills. Chief AI officer Kasper Tjørntved Davidsen says the aim is to democratise the technology and raise skills across the organisation. As an example, he points to an in-house hackathon in June, attended by 700 staff, where a customer advisor who had never touched a terminal built five AI agents to handle part of his daily work in two days.

The core tool is an Agent Developer Kit that runs on a declarative engine. Employees describe in plain language what the agent should help with, and the system guides them through the process. Davidsen says this is ten times faster than building an agent from scratch, and that agents automatically get the bank's built-in security, guardrails and compliance features. He says this gives traceability and quality and helps the bank avoid "shadow AI", where staff turn to unapproved tools.

The bank has also launched an internal AI marketplace where employees can find solutions already developed and deployed. Most were created by employees rather than the IT department. An AI Innovation Lab produces a "proof of value" and tests each proposed solution before it is shared. AI City runs on standardised cloud components, including AWS Bedrock and AWS Agent Core, and consists of 50 interchangeable modules. Because modules are replaced centrally, all solutions and agents update automatically.

To build skills, the bank offers training, e-learning portals and demo days. Its annual GenAI Week, held in week 41, is an in-house conference with around 70 sessions and workshops. It drew 8,000 staff last year, and 10,000 are expected this year.

The effort fits the updated Forward '28 strategy. The bank expects AI and other technologies to add roughly DKK 2 billion in value by 2028, while annual spending on technology, AI platforms and consultancy rises from about DKK 4 billion to about DKK 4.5 billion. One reader commented that the article shares no concrete use case of an employee-built agent beyond the hackathon example.