Your Best Customers Are Leaving and You Don’t Know Why

It was about three years ago, and I was in a video call with a former member of my team who was now the head of customer experience at a mid-sized software company. The company had dashboards, quarterly reports, and a team that responded to every complaint within 24 hours. She was proud of her operation, including the survey response rate of about 25%, and she should have been. Then I asked a simple question: “What do you know about the 75% of your customers who never respond to your surveys?” The silence that followed told me everything.

The invisible majority

Here is the uncomfortable truth about B2B customer experience today: Survey response rates have been falling for years, and Medallia’s 2026 State of Customer Experience Report puts the average in Q3-25 at just 8.6%. In most B2B companies, fewer than one in four customers ever tells you what they think. The rest are invisible.

And invisible does not mean happy. There is another category of silent customers and this one is a bit subtle: the Passives who respond to NPS surveys. When I was at HP, our global analytics team in India studied a $25 billion business built mainly on large annuity contracts that had a 100% survey response rate. They found that it was the Passives: the silent middle who never complained, who stopped renewing when NPS declined by 15 points over three years. Not the Detractors, who were loud and got attention. The quiet ones just left.

This is not a minor gap in your data. This is the majority of your revenue sitting in the dark.

What AI actually does here

I want to be precise about what I mean by AI enhancing humans, because the phrase gets thrown around a lot. I do not mean replacing your customer success team with a chatbot. I do not mean automating your survey process so you can send even more surveys that even fewer people answer.

What I mean is this: your company already has operational and demographic data for every customer, meaning information about how they use your product, how often they contact support, how their contract terms have evolved, what industry they are in, and dozens of other signals. For the customers who do respond to surveys, you can build a model that connects all of this operational data to their expressed satisfaction. Then you apply that model to every customer who has never responded to a single survey. The AI does not replace the human judgment; it extends it to places human judgment currently cannot reach.

A study published in Nature’s Scientific Reports found that AI models using operational data achieved over 95% accuracy in predicting customer churn in telecommunications. Results will vary across industries, and the point is not the exact number; it is that operational data alone, without any survey, carried enough signal to predict what customers would do. A systematic review of 240 studies confirms that this is now a well-established across multiple industries. Most of these are not customers who responded to surveys. These are the silent ones.

It is not only about preventing bad things

Most discussions about customer retention focus on firefighting, meaning identifying unhappy customers and trying to save them. And that matters. Research by Fred Reichheld of Bain & Company shows that a 5% improvement in B2B retention can increase profits by 25% to 95%.

And here is the part that almost everyone misses. The same AI model that identifies risk also identifies what is working. When you can see patterns across your entire customer base, not just the vocal minority, you can determine which of your actions and behaviors are actually driving loyalty. Perhaps customers in a particular segment are growing because of a specific onboarding approach. Perhaps a certain pattern of account management contact is strongly associated with renewal. You would never see these patterns if you only listened to the people who filled in your survey.

AI does not just help you do less of what customers dislike. It helps you do more of what they value. That is enhancement, not replacement.

So therefore…

If you are running a customer experience program built entirely on survey responses, you are making decisions based on the opinions of a shrinking, unrepresentative minority. The technology now exists to understand every customer, including the silent majority who will never fill in your questionnaire.

The action is straightforward: start building a model that combines your survey data with your operational data, so you can extend your understanding from respondents to your entire customer base. Your customer experience and success teams do not get replaced; they get something they have never had before, meaning genuine insight into the customers they currently know nothing about.