An answer is often just the beginning
An invoice is missing. A document has not been submitted. A question concerns a specific contract. A convincing sentence is not enough: the next step must fit the actual case.
Our developed customer-service agent FRED connects answers to actions. The existing published case study describes knowledge, system access and human responsibilities. This insight explains that workflow without publishing new customer data.
From the request to an allowed action
One previously documented example: someone asks for their latest invoice on WhatsApp. FRED checks who is writing and can send the invoice by email through the intended process. That completes a task rather than merely telling someone where to search.
| Workflow stage | What matters |
|---|---|
| Understand the request | What information or action is needed? |
| Retrieve context | What do permitted sources say about the case? |
| Use a tool | Which action is the agent allowed to perform? |
| Check the result | What actually happened? |
Available interfaces are another limit. If a required capability is missing, AI prepares the case for people.

Handover is a specific capability
The published case page distinguishes channels: on WhatsApp, FRED can create a case with a summary for the team. In website chat, it directs the customer to another contact route. A blanket claim that AI automatically hands over everything would hide these differences.
A cancellation request is another example: AI can record it, while the decision belongs to the team. A handover should only be confirmed once the relevant case has actually been created.

What reaches the person matters
A useful workflow reduces the need for the team to gather context again and for the customer to repeat it. Whether that works in another business depends on the task, data, tools and responsibilities.
The existing case study reports historical metrics with a measurement period. “Without handover” measures handovers, not proven resolution. This insight adds the workflow perspective and makes no claim of a new operational measurement.
Tell us about a recurring task. We first assess where AI can work usefully and where a person is needed.
Sources.
- Vorhandene SolvraONE-Fallstudie KundenserviceOwn source · As of
- SolvraONE: research and source reviewOwn source · As of



