Daniel Okafor runs support at Lumenpay. Two years ago his team measured success by deflection. Today they measure resolution — and 68% of tickets are resolved without a human touching them.

Why deflection stopped working

Customers who needed something done still waited. The chatbot was good at answering questions and useless at refunds, card replacements and limit changes — the reasons most people write in.

Giving the agent real tools

Lumenpay gave its agent the same permissions as a week-two hire and the same policy document. Every action writes an internal note with a trace link, so the team can see what happened when a customer follows up.

The question used to be ‘did the bot answer?’ Now it’s ‘did the customer get what they needed?’

Read the full case study for the rollout plan and the numbers.