Should You Inject GenAI Into Customer Support Team?

A new study by MIT Sloan School of Management , Harvard University , and Stanford University [Generative AI at Work, 2025] gives us one of the clearest pictures of what AI really does in real jobs. Authors: Erik Brynjolfsson, Danielle Li , Lindsey Raymond

They rolled out an AI assistant (based on GPT-3) to +5k support reps at a Fortune 500 company.

Results:

1. +15% boost in productivity Support reps resolved more customer issues per hour with AI assistance.

2. Biggest gains came from junior team members Those with less experience (junior analysts) improved their productivity by over 30% — and closed the gap with veterans in just 2 months. (The red line in the 3rd chart shows that "Always Treated" employees outpaced everyone else.)

Other Things to Consider

1. With AI nudges, junior reps began to communicate like top performers: clearer tone, faster responses, more empathetic language.

2. But too much help can dull expertise. For top-performing reps, AI slightly lowered quality. Over-relying on suggestions can backfire when you already know what you’re doing.

“Generative AI doesn’t just speed things up — it levels the playing field. It turns average into excellent.” — Brynjolfsson, Li & Raymond (2025)

Takeaways

Give them AI tools

Use AI to scale consistency, not replace human judgment

Don’t autopilot your experts, keep their edge sharp

Are your sales, marketing, and support teams ready?