Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

Customer experience teams sit on a goldmine of unused data

AI can help make sense of it.

[Photo: Getty Images]

Profile photo of Brad Birnbaum

Co-Founder & CEO · Kustomer

Every business leader, myself included, sits in rooms full of smart people and spends hours trying to figure out why customers were leaving. Everyone had a theory, usually backed by selective data scraped together in the search for clear causality.

That instinct to look to the data for answers is right, but limited. Leaders squinted at surveys and dashboards full of traditional customer experience (CX) metrics to find some actionable pattern. Meanwhile, the most valuable source of information —actual support conversations with customers— only got a sliver of attention.

Every customer service interaction is an unedited account of what customers are feeling, in their own words. A company can have thousands of these conversations every week. But they’ve rarely been a big part of the analysis, because most companies never had a practical way to make sense of that many conversations at once.

That once made sense. No team could realistically read every conversation, so a sample became the unquestioned standard. But AI has changed how much CX teams can see.

THE SAMPLE ILLUSION

For as long as I’ve worked in this industry, “monitoring the customer experience” has meant pulling a handful of call records, spot-checking a folder of tickets, and rolling up everything that’s been reviewed into a tidy dashboard.

While valuable, this process is like reading one page from every chapter of a book and then telling yourself you know the whole story.

For many teams, that’s the only method that’s ever been available. But it’s a sample standing in for the whole, dressed up with enough process and rigor that nobody stops to ask how representative it really is.

Sampling works well when you’re looking for something common and likely to show up wherever you look. It fails when you’re looking for something rare. And problems worth catching early, like an emerging complaint or a policy that’s starting to break down, are almost always rare before they become common.

By the time a problem is showing up enough that a small sample catches it reliably, it’s usually not a small problem. So the review process most companies miss exactly what it was designed to find: trouble still small enough to fix cheaply.

WHY FULL COVERAGE WAS NEVER REALISTIC BEFORE

Structured data that teams build in advance can only capture what you plan for ahead of time. A conversation is the opposite of that: open-ended, unpredictable, full of things you can’t put in a database field.

The only way to make sense of it was to read the data manually, and there has never been enough human time in the day to do that across even a meaningful fraction of a company’s customer conversations. Full coverage was a ceiling CX teams couldn’t get over.

So the only manual review that was possible focused on the narrow job of basic quality assurance. Did the support agent say the right thing, is the AI agent responding how it’s supposed to, and so on. While reasonable, it’s not the same as understanding what customers are actually experiencing.

For that, companies reached for traditional surveys and hoped enough people answered honestly for it to add up to a real picture. Analyzing conversations at scale was never an option on the table.

3 WAYS AI CHANGES THIS

Reading conversations at scale is now among the most viable, useful things that AI can do for a CX organization. Three shifts happen to companies that put all of their conversation data to work.

The first is coverage instead of sampling. AI reviews every conversation, which changes what a quality score means because it looks at the whole, not just a fraction.

The second is proactive signals instead of reactive surveys. Satisfaction, intent, and root cause can be read directly within what a customer actually says in the moment during support interactions.

The third is pattern recognition instead of patchwork. When AI automatically reads and analyzes all conversations, it identifies patterns. A person reading one conversation at a time can, at best, notice a pattern by memory. The same complaint that popped up three weeks ago, in a different channel, is showing up again today.  

Altogether, these shifts mean a CX team is finally seeing the root cause behind a dozen complaints, or catching a product issue building for weeks before it turns into a support spike.

WHAT THIS WAS ALWAYS ABOUT

In my many years in the CX space, I’ve believed that the businesses that understood their customers best would keep them. That belief is one of the few things that hasn’t changed throughout decades of technological innovation.

What’s changed is the gap between what customers tell us and what we can do about it. That gap has closed.

It may sound simple, but understanding customers is the whole job of a CX team. And the clearest place to find that understanding was the customers’ own words, and it always has been. Those words are now a lot easier to hear. 

Brad Birnbaum is the cofounder and CEO at Kustomer.