A simulation of what I think call centre tools should look like.
I've spent six years on the phones, in telecom and in group retirement savings. ClearConnect is based on my research review of contact centres and AI, and it builds in my own call flow as 25 rules an AI can be checked against. You get to practise and see what good, clear, structured calls are like, and how they help you connect with customers while resolving their issues.
This is a portfolio project, not a live service: every customer, account and policy is fictional. My research review looked at what makes contact centre work hard and which kind of AI actually helps. ClearConnect is my answer to what those tools should look like.
Practice a customer chat with a suggested reply when you ask for one. Listen to 13 recorded calls with an AI review against the call flow: what went well first, then what to work on, with a better line to try. See your own numbers by month, led by what to practise next.
I wrote 16 test conversations, 10 of them built to trip the AI up, and had a second AI grade every reply against my rubric. 14 passed at first, and all 16 after three rounds of fixes. ChatGPT then followed the same flow and passed 15 of 16, judged by Claude. When I disagreed with the call reviewer, my judgement became a rule.
The practice chats and call reviews are live. The public chats use written customers, so there's no AI cost and nothing leaves your browser. A live call with an AI agent in my own voice is coming soon: it works on my computer, and I'm still improving how natural it sounds. Everything is fictional and in English, and the tests use conversations I wrote, so a pilot with real calls would be the next step.
Click to enlarge