Subscribe
Learn Library

In 2026, AI Starts Calling the Shots on Customer Experience

A learn article comparing 2026 reports from Deloitte, McKinsey, Twilio, Salesforce, and Omdia on AI in customer experience, covering the customer engagement paradox, trust, real-time personalization, and scaling AI, plus a step-by-step way to fix one high-traffic customer journey first.

ai-marketingevidenceworkflow
2026-08-24SupaMarketers6 min read

Have you had one of those moments lately? You explain your issue to a customer-service bot, close the window, and a marketing push notification pops up that has nothing to do with anything you just said.

I hit it once myself. It feels very real.

In that instant, the line running through your head is almost always the same: "This company — how does it feel like a few different stores stitched together?"

I recently stacked a few industry reports side by side and read them together. The more I look, the more convinced I am: that moment when everything tastes "off" is exactly the question worth wrestling with in 2026.

The reports come from Deloitte, McKinsey, Twilio, Salesforce, and Omdia. Five firms that usually each do their own math — this time, unusually, all standing on the same side.

What's the Customer Engagement Paradox?

Twilio's report has a phrase in it I remembered at first glance: the customer engagement paradox.

What's a paradox? It's digging a trap for yourself.

The brand's side thinks: We haven't skimped on messages, we haven't skimped on channels, and we haven't fallen behind on AI. So why does the customer still shrug?

The customer's side feels: There's plenty of stuff here. But none of it speaks to me.

See it? One side keeps pushing, the other keeps swiping to get away. The problem was never "did you do it" — it's "did you do it in the right way."

Twilio's suggested direction is actually very plain-spoken: don't rush to add more tools. First, make your "signals" solid.

What's a signal? Who the customer is, what they prefer, and what circumstances they're in at that exact moment. Whether that foundation is accurate directly decides whether what you hand over feels like "just right" or "let me interrupt you."

If your signals aren't solid, AI can work ten thousand times and never earn a single "this brand gets me."

Customer Wants to Be Understood, and Fears Being Seen Through

Salesforce's report draws customers in a familiar way: with one hand they want "you know me best," and with the other they're afraid "you've got me watched."

The two sides fight each other, and the situation gets complicated fast.

There's a number in the report I've remembered to this day:

88% of customers think that the more things are changing, the more trust matters.

Why does this number hit so hard?

Because the experience AI shapes runs on data. If a customer doesn't trust you, they won't hand over their data; and when data runs thin, all the "personalization" you've layered on top is just floating in the air.

So Salesforce's conclusion is blunt: before you try to look clever, stand the trust up first.

In practice, that's four things:

  • What customers give in exchange for their data has to be spelled out clearly
  • How personalization works, and what it draws on, has to be something the customer can see
  • Across every touchpoint, say the same one thing — don't contradict each other
  • Until trust is standing, keep your hands light on the data

McKinsey Puts a Definition on "Personalization"

McKinsey's report gives personalization a very measured definition: using data to adjust, in real time, the timing, content, and form of every single touch.

At first glance, you'd think that just means "knowing."

But you should press on the last two words: real time.

Right now, how far does most companies' personalization actually reach? They can get your name right, they can line up recommendations that look passable. That's already not bad. But the moment it comes to "real time," a lot of companies go silent.

Why is real time so hard?

Because real time is a coordination craft. The minute a customer's action starts to surface, the data has to be understood right away, turned into a decision right away, and delivered to their screen right away. If any one link in that chain slips, real time snaps right there.

So McKinsey's advice is practical: don't rush to lay out ten scenarios at once. Pick one journey where the change in results is actually visible, and get both steps working — "see the intent" and "catch it immediately."

Unblocking one point at a time is always better than ten points each running on their own.

Deloitte Says the Winner in 2026 Is "Scale"

Deloitte's report, Tech Trends 2026, has a single keyword: scale.

By 2026, AI is no longer sitting in the lab. It's out in the real business, fighting for real.

Riding along with it are all the holes that used to stay hidden.

You could build a chatbot over a single weekend, sure. But getting AI to look after marketing, sales, and after-sales at the same time — getting the entire journey to answer the customer with one consistent logic — a lot of companies show a crack the moment they start.

Everything Deloitte points to is "foundation work":

  • Standardize the data pipeline, so nobody's hauling data around by hand
  • Stand up governance and authorization first
  • Metrics have to be explainable clearly, none made up on a whim
  • Tear down the walls between departments, on a schedule

Turning that into plain words: an AI project that can't clear the "scale" hurdle probably won't outlast the next budget cycle.

When You Do It Yourself, Where Do You Set Your First Foot?

All of that said, when you look back at your own company, a lot of people's first impulse is "okay, I'll climb onto the big platform."

Take it easy.

The right move is to win one small battle first.

Here's an order — feel free to try it:

  1. Pick one journey with volume and obvious pain — registration, checkout, returns, rescheduling — the one with the most people wins;
  2. Mark the spots where "a context you spent real effort building up is lost the moment you hand it over";
  3. Get this journey's identity and data right — first, just this one;
  4. Then go where the friction is worst, and let AI step in — only to smooth away that one hard point;
  5. And finally, watch only the result. How many messages went out is the process; whether customer behavior actually changed is the answer.

The upside is that you don't have to tear down the whole company to test one result. And whether it turned out well, you'll be able to tell at a glance.

The Customer-Experience Road Has Already Changed Its Walk

Five reports, and they all point to the same exit: the customer-experience contest has quietly gone from "how many channels I put effort into" to "what the customer feels walking the whole journey."

Companies that respond in real time will glide smoother and smoother. A chain that's still a pile of scattered pieces will, one day or another, choke on being "a patchwork."

I always think — no wonder customer-journey orchestration has been heating up the market these last couple of years. Behind it, all it carries is people's simplest hope: stitch up the data scattered across all those systems into one honest thing that can actually make a decision for the customer.

One more thing, then:

By 2026, making the customer feel — from their first word to their last — that they're talking to the same person the whole time is the firmest card in your hand.

So when are you going to play it? Don't keep the customer waiting.