01 · Understand first.
Which conversations do YOUR CUSTOMERS NEED?
An AI on the phone needs more than a friendly voice. It has to ask the right questions, get the right information and recognize when to pass a request on.That's why the Canvas starts with your day-to-day work. You describe your situation, your callers and the tasks where support makes sense. Only then does the design of your assistant take shape.You don't need any programming skills for this. The experience from your business is the starting point. Technical requirements become clear as the planning progresses.
02 · Then design.
Seven topics BUILD ON EACH OTHER.
Decisions about target groups and tasks feed into your assistant's boundaries, workflows and language.01
Problem space & motivation
How do calls work today? Note where waiting times, interruptions or system breaks occur – and which problem you actually want to solve.02
Define the target group
Your callers have different concerns and expectations. Describe who calls, what support they need and how open each group is to AI.03
Tasks & evaluation
Collect possible tasks and prioritize them. A clearly defined starting point helps you weigh benefits against effort before you work out the conversation flows.03 · Get specific.
The result: A PROMPT TEMPLATE AND A TASK LIST.
The Canvas summarizes your input for implementation. You can see which instructions the AI needs and what still has to be clarified before it goes into practical use.The template is a working basis. Together with you, we adapt it to the telephony in use and the connected applications, and test the dialogs in realistic scenarios.Your prompt template
Your assistant's role, tasks, conversation rules and boundaries in one coherent instruction.Your task list
Open prerequisites and concrete steps for the technical and organizational rollout.From research TO PRACTICE.
The Canvas is based on the scientific publication “Empirische Entwicklung eines Design Canvas für Conversational AI im Kundenservice” (Empirical development of a design canvas for conversational AI in customer service) by Pascal Slaghekke (2026).It combines a review of the scientific literature with original studies on the needs of business owners. Its development follows the design science research approach: a practical tool makes the findings usable for design.The guidance notes in the Canvas explain the background of each question and refer to the underlying literature. That way, you can make well-founded decisions.The Canvas supports planning. It does not replace technical testing or a review of individual legal requirements.