Across marketing (and the wider world), the polarizing effects of AI are visible everywhere.

The LinkedIn posts are abundant, noting how AI is ushering in a new era for businesses, with enthusiastic examples of how it’s being used for everything from campaign automation to ordering groceries.

The concerns are equally abundant: factual inaccuracy, hallucinations, opaque data use, and real questions about what happens to human judgment and creativity when you hand more decisions to a machine.

These aren't fringe worries. Some of the recent headlines around AI errors and misuse of personal data have been genuinely alarming, and any marketer paying attention has good reason to be cautious.

Most marketers sit somewhere between those positions. We see the value AI can bring to our day-to-day, but are aware of the inherent risks it brings, too.

Regardless of which end of the AI spectrum you find yourself on, the bigger question around loyalty is ‘What do our customers actually want?’

If we start introducing AI into our workflows, will that negatively affect brand perception or, at worst, alienate customers and drive churn?

SAP’s Global Customer Loyalty Index (CLI) has revealed some bifurcated insights about the impact of AI on customer loyalty; one being that AI is making customers more loyal while also making it easier than ever for them to jump ship.

The index shows that 31% of consumers say interactions with AI assistants make them feel more loyal towards a brand. But concurrently, 30% have already switched away from a brand because of its use of AI, rising to 45% among Gen Z.

This demonstrates the nuanced complexity of understanding the impact of AI—it’s more than consumers just asking brands for "more AI" or "less AI". Instead, they're making specific asks; they’re saying, "help me, remember me, and act for me when it makes my life easier"—but don't move the steering wheel somewhere I can't reach it.

For marketers deciding how far to push automation, the opportunity isn't full autonomy, but rather earned autonomy: giving AI more responsibility as customer confidence grows.

The deal-makers and deal-breakers of customer loyalty

Before we get to AI, it’s worth unpacking one of the CLI’s key findings.

Incentivized loyalty (loyalty earned via discounts, points, and rewards) has fallen from 48% in 2025 to 38% in 2026, the first time it has dropped below 40% in the six-year history of the research. Meanwhile, true loyalty, (SAP's term for deeper brand affinity), has rebounded from 29% to 32%.

This doesn’t mean discounts have stopped working, loyalty programs are dead, or that coupon codes should be thrown into a fire.

It does mean that incentives alone are becoming a shakier foundation when it comes to the customer relationship. A discount provides a reason to buy one time; that’s very different from establishing a reason to purchase from you the next time.

The rest of the report makes that distinction clearer. Consumers point to the usual suspects when asked what builds loyalty:

• 62% point to quality products
• 42% cite trusted brand reputation
• 31% value connected online and offline experiences
• 29% point to frictionless shopping
• 23% choose memorable, personalized experiences

The irony here is that while marketing teams are busy discussing agents, orchestration, and predictive personalization, customers overwhelmingly just care about whether the product is any good.

Declining product quality is also the biggest loyalty breaker at 45%, followed by rising prices at 40%, and poor customer service at 36%.

The foundations that make loyalty work are unchanged: is the product good, is the price defensible, and can someone actually get help when things go wrong? AI cannot compensate if a brand is failing those fundamental tests. At best, it will reveal the disappointments to customers more quickly.

Not so fast: 60% of consumers want shared control with AI

The research asked consumers where they want control to sit between humans and AI, and the majority is not at either extreme:

• 20% want full human control
• 60% want shared control
• 16% are comfortable with AI leading, provided there is oversight
• 4% want a fully autonomous experience

These data points suggest that people aren't overwhelmingly rejecting automation, nor do they desire handing over the customer relationship in its entirety to a machine. What they want is for AI to do useful work while retaining control and the ability to guide strategy and decision-making.

The trust data backs that up. Human support if something goes wrong is the top trust-builder at 42%, while 36% want the ability to review or override decisions. Data privacy and security assurances sit at 35%, clear explanations at 31%, and proven accuracy over time at 29%.

The research in the CLI is saying consumers want AI assistants to remember useful context, like past purchases, regular buys, and delivery preferences.

Customers aren't asking brands to understand their deepest, most unconscious thoughts and desires; they'd just like you to remember their prior purchases and that you prefer delivery on a Thursday.

That sounds simple, but for most organizations, it isn't—purchase history lives in one system, delivery preferences in another, and the marketing team can see neither without filing a request.

What customers experience as a single relationship, most brands still manage across a dozen disconnected systems. Closing that gap is where AI-driven engagement and personalization begins.

Defeating decision fatigue: when choices are overwhelming, AI is most helpful

One of the most useful findings in the report is that consumers turn to AI most often when a decision is taking up too much time and energy:

• 23% use AI when they’re unsure what to buy.
• 22% use AI when there are too many options.
• 20% use AI when they’re short on time.
• 17% use AI when buying gifts.
• 16% use AI when they’re new to the brand or platform.

The AI benefits customers appreciate most are also the most practical AI use cases: fast advice, help finding the right products, efficient problem-solving, and a smoother shopping experience. There’s a great rule of thumb we can draw from these insights, which is simply that AI thrives where choice becomes overwhelming.

That means the better question for marketers isn’t "what can we automate?" but rather "where are we making the customer do unnecessary work?"

A catalog containing 400 near-identical products is an obvious candidate. So is choosing a gift, comparing complex plans, finding a compatible product, or navigating an unfamiliar category.

Those are moments where AI can remove friction rather than add an AI-shaped layer to an experience that was working perfectly well without one. If the customer thinks, "that saved me some effort", you're onto something.

AI autonomy is a matter of earned trust

SAP frames the relationship between AI and loyalty as an equation: Earned Trust + Human Control + Autonomous Action = Loyalty. I'd read that less as a mathematical formula and more as a permission model, because the research shows consumers want two things at once:

• 33% trust AI to complete purchases on their behalf.
• 31% regularly rely on AI to help choose what to buy.
• 56% still want the ability to reverse AI decisions when necessary.

That last figure is key to why AI autonomy has to be earned. Customers aren’t saying "don't act for me"; they’re saying "act for me, but if this goes sideways, I want to retain the ability to course correct or speak to a human." The willingness to delegate is there, but it remains conditional on confidence and control.

For marketers, this means that the aim isn’t simply to make AI capable of doing more. It’s to lower the perceived risk of letting it do more by proving accuracy over time, keeping consequential actions reversible, explaining decisions when it matters, and ensuring a human can step in.

Think about any useful assistant. You don't want to approve every tiny decision they make, because that defeats the point of having an assistant. But you DO WANT them to recognize when a decision is consequential and let you intervene before a small mistake becomes an expensive one.

That’s what makes autonomy feel earned rather than imposed. Trust gives the brand permission to personalize; control makes delegation feel safe; consistently useful outcomes can then make customers more comfortable handing over the next task.

Autonomy is not the goal in itself; less effort for more reward without a corresponding loss of confidence is.

No prophets here: Gen Z is a stress test, not a crystal ball

Gen Z makes the AI tension especially visible. 45% have already switched brands because of better AI practices, while 44% say they would trust an AI assistant to purchase an item for them.

They appear unusually willing to embrace AI and unusually prepared to punish brands when the experience falls short.

That makes Gen Z useful, but not prophetic. There’s no basis to assume older generations will behave like today's Gen Z in five years' time.

A better use of the data is to treat younger consumers as a stress test: can the experience be useful enough that somebody is happy to delegate, while remaining transparent enough that the same person does not feel tricked, trapped, or surveilled?

If both answers are yes, you're probably designing around the right tension.

What marketers should do now:

• Find the moments of cognitive friction. Look for choice overload, uncertainty, repetitive decisions, and places where customers are wasting effort.
• Build useful memory to create dazzling personalization. Remember orders, preferences, service history, and context that customers have already given you. "You don't have to tell us again" is an underrated customer experience.
• Design the escape hatch with the automation. Overrides, reversibility, explanations, and access to a human are part of what makes people comfortable delegating.
• Use autonomy selectively. The more consequential the decision, the more visible the controls should be.
• Keep the basic stuff sacred. Product quality, price, service, and reputation still do more of the heavy lifting on loyalty than most technology-occupying marketing conference agendas.

The best AI experience shouldn’t feel like an AI experience at all

The CLI makes clear that consumers are becoming comfortable granting technology more authority without giving it unconditional authority.

They'll let AI narrow options, remember useful context, and, even purchase on their behalf. But they still want visibility, an override, and a human when things go awry. The goal shouldn’t be maximizing how much AI can do. It should be maximizing the useful work customers are comfortable letting it do.

That’s a higher bar that results in a better customer experience. It makes the principle behind SAP's Autonomous Marketing and Engagement proposition (people set the direction; AI executes) more interesting, because the important word isn't "executes". It's "direction".

Earning that permission makes autonomy feel less like losing control and more like receiving good service.

Explore related content