A customer has moved home and wants to update their address. The service team knows what comes next: ask for the relevant details, establish which account the request concerns, and confirm the change through the appropriate process. They also know they will have this conversation dozens of times this week.
That knowledge is the best starting point for building an AI Agent. It describes a task, the information it needs, and the experience the customer should have.
With AI Agent Studio and the new Agent Mode, powered by botario technology, you can turn that description into a first working AI Agent. Test the conversation, see what needs to be adjusted and refine it around your requirements.
Describe. Create. Test. Refine.
Agent Mode introduces a conversational way to build. Begin by describing the Agent’s goals, behaviour and requirements in plain language.
Agent Mode and the integrated AI Assistant create a first working configuration from elements such as persona, behaviour, knowledge and tools. You can test the Agent directly, then ask the AI Assistant to make changes as you develop the scenario.
For the address-change example, the first test might begin with an ordinary message: “I moved recently and need to change my details.” You can check whether the Agent understands the request, asks for the information it needs and explains the next step clearly.
The first version gives the team something concrete to assess. The people who know the service can discuss how the conversation should work while the requirements take shape.
Refine the conversation as requirements grow
Once an initial scenario is working, the AI Assistant helps you develop it further through conversation.
An energy-service demo can begin with address changes and then add meter-reading submission. The new instruction can specify which information to collect and require the Agent to read back the meter number and reading before submission.
Changes remain inspectable and reversible. Review and approval, rollback and separate development and production endpoints support controlled development and testing. You can examine a proposed change before applying it and test how it affects the conversation.
Choose the way you build
AI Agent Studio also includes Flow-Based Mode, the established approach for creating workflows visually and configuring individual steps manually. It gives you deterministic control over the sequence.
Agent Mode adds a conversational approach alongside it. You choose how to build your Agent depending on the use case, with the flexibility to extend the solution through knowledge, tools and deeper technical customisation.
Start with a service task people recognise
To make things easier, you can follow one of the demo blueprints. In insurance, an Agent can guide the first conversation about a travel claim, collect the required details and forward sensitive cases. In public services, a demo can combine answers from official information with appointment coordination. In energy, address changes provide a starting point for adding meter readings.
These examples make the intended experience visible. They also help teams identify the systems, data and support the real service will need.
Easy to create. Ready for complex requirements.
A first working demo and a production implementation serve different purposes. The demo lets you explore the conversation. A live service needs the relevant channels, connections and testing in place.
The first-contact demo scenarios do not include live third-party integration. For example, demonstrating an address-change conversation does not update a real customer account. Connecting that process to a billing system requires technical implementation.
Simple Agents with little or no third-party integration can follow a self-service path. For complex APIs, third-party systems and individual production designs, partner expertise and NFON Professional Services support you at every step.
AI Agent Studio makes it easier to get started while keeping the flexibility and control to go deeper. Begin with the question your customers ask most, create a first version and use what you learn to shape the solution.