The model reads the complaint and suggests a category, an urgency level and a likely resolution. An agent always makes the final decision.
Agents see only the cases assigned to them, update the status and record the final judgement.
On resolved cases, the agent's final category matched the AI's top suggestion in 8 of 10 cases, and its top two suggestions in 9 of 10.
Resolved cases feed back into the model when it is retrained.
Sample figures for the prototype.