Helping a legacy manufacturer reduce inquiries to their internal support desk
An industrial agricultural manufacturer needed help reducing inquiries made to their support desk.
- Role
- Lead
- Year
- 2024
Approach
We ran generative user research with prospective end users, and a quantitative analysis of over 300,000 support tickets to understand major pain points and use cases for an automated solution.
Result
An interactive chatbot which leverages GPT. This bot is currently being used by 34% of the companies international workforce with 62.5% ticket resolution success rate, and 50% success for escalated routing, saving our client significant overhead costs. Lessons: These are scenarios to use tools like ChatGPT and not to use ChatGPT for research synthesis. It’s sometimes best to do synthesis by hand instead of relying on an LLM. There’s also no need to try to synthesize thousands of data points, when synthesizing hundreds will give you a result that you can trust just as well (e.g. 99% confidence).