This is an illustrative scenario, not a real client. It shows how we'd approach a situation like this. Figures describe what we'd aim to measure, not results we're claiming.
The situation
A small online and shop-floor retailer answers a steady stream of "where's my order", "do you have this in stock", and returns questions across email and social. It's most of one person's day, replies slow down at busy times, and the same answers get written again and again.
Where AI could help
The realistic opportunity is customer comms: drafting accurate replies to common enquiries from the retailer's own order and product data, for a human to approve and send. The aim is to take the repetitive load off, not to put a chatbot between the business and its customers.
How we'd approach it
- Intake. Which enquiries actually repeat, and what data a good answer needs.
- Opportunity scoring. We'd separate the genuinely repetitive questions from the ones that need a human, because not everything should be automated.
- Proof of concept. We'd build a working demo that drafts replies to a set of the retailer's real, anonymised enquiries, so they can judge tone and accuracy before any build.
- Report. Where it's safe to assist, where a person must stay in the loop, and the running cost.
- Handoff. A clear brief, including how to keep the customer's experience human.
What success would look like
In the proof of concept we'd measure how many enquiries get an accurate first-draft reply, the time saved per day, and response times at peak. Whether it's worth it depends on those numbers for this business, which is exactly what the proof of concept is for.