Articles2 Aug 2026
Where a chatbot stops paying for itself in Arabic-first service
Deflection is easy to celebrate and easy to over-count. Here is where the return curve flattens, and what to measure instead.
Read moreWhen a utilities contact centre came to us, its customer satisfaction score had been stuck between 66% and 68% for three consecutive quarters. Nothing had changed on the surface — same volume profile, same channels, same team size. What had changed was the mix of contacts: more complex, multi-step requests arriving with higher stakes.
The first move was not a new tool. It was measurement discipline: rebuilding the quality rubric so that it measured resolution quality, not just call handling. Once auditors and team leads agreed on what a good outcome looked like, coaching finally had something concrete to aim at.
Forecasting came next. Historical interval data was hiding a mid-morning peak that staffing had quietly absorbed as overtime. Re-cutting schedules around the real curve recovered hundreds of hours a quarter without adding headcount.
Six months later the score crossed 94%. The order mattered: measure properly, then coach, then forecast. Teams that invert the sequence usually buy technology to solve what is, in the end, an operating-discipline problem.
Keep reading
