New Metrigy research finds AI coaching narrowed the customer-rating gap between top and low performers by 19.4%

Every contact center has them: The agents who customers rave about and the agents who generate callbacks and complaints. The gap between them is one of the most expensive problems in CX, because customers don’t choose which agent they get. Vendors pitch AI-based coaching as the fix. Their stance is that personalized feedback, generated from every interaction, helps struggling agents improve faster than a supervisor could manage alone.

So does it work? Metrigy’s Workforce Engagement Management: 2026-27 study of 108 companies globally suggests it does, though not necessarily—or equally—for everyone.

The gap is narrowing

We asked companies to rate the customer experience their top and low performers delivered on a scale of 1 to 10, before and after they introduced AI-based coaching. Both groups improved. Top performers rose from 6.4 to 7.9, and low performers rose from 5.1 to 6.8.

The low performers gained more, though. Their ratings climbed 33.2%, compared with 22.6% for top performers. As a result, the gap between the two shrank from 1.3 points to 1.0 points, a 19.4% reduction.

Companies are noticing. Among those that compare customer ratings for their best and worst agents, 46.4% say AI coaching has helped close the performance gap, up from 34.8% last year. Another 36.2% use AI coaching but haven’t seen the gap change yet.

The results aren’t automatic

The companies in our Success Group—those with the biggest measurable gains from WEM—get far more out of AI coaching. Their low performers reach an average rating of 7.3 after coaching, compared with 6.2 for low performers in the Non-Success Group. The Success Group’s top-to-low gap narrowed by 25.3%, more than twice the Non-Success Group’s 12.0%. Successful companies are also more likely to say AI coaching closed the gap: 54.3%, vs. 30.4%.

Company size matters, as well. Only 25.0% of small companies (fewer than 250 employees) say AI coaching closed the gap, compared with 59.3% of mid-size companies and 50.0% of large companies.

What sets the leaders apart isn’t simply having the tool. Successful companies are more likely to measure the gap in the first place: 71.9% compare customer ratings for top and low performers, vs. 53.5% of the Non-Success Group. They’re also twice as likely to say human coaches remain critical even with AI (43.8% vs. 20.5%). They aren’t replacing coaches with AI; they’re giving their supervisors and agents better coaching tools.

What CX leaders should do

Here are four surefire ways to get more out of AI coaching, based on what the most successful companies in our study are doing:

• Measure the gap before you try to close it – Track customer ratings for your top and low performers so you have a baseline and can prove AI coaching is working.
• Pair AI coaching with human coaches – Use AI to find skill gaps and personalize feedback at scale, then let supervisors and coaches act on it.
• Give it time, but check progress – More than a third of companies using AI coaching haven’t seen the gap move yet. Set a review point to decide whether the problem is the tool, the content, or how it’s being used.
• Don’t expect AI to fix everyone – Top performers improve too, so a narrower gap is a win even when your best agents keep pulling ahead.

AI coaching won’t turn every agent into your best agent. But, as our data shows, it can make your weakest agents noticeably better—and customers notice the difference.

Author(s)

Layne Haaksma