An enthusiastic response to an AI concept can hide disagreement about how much the product does. “Takes care of your weekly shop” might mean a prepared basket to one person and an order placed without asking to another. Both can score the concept highly and still want different products.

A customer can want the benefit of an AI feature and still hesitate at the moment it acts for them. Recommending something, preparing it for approval and deciding without asking are different promises, and research into an AI-enabled product needs to know which decisions people will hand over along the way.

Consider an illustrative household service that helps with food shopping. It suggests meals, notices ingredients likely to run out and prepares a shopping basket. That could remove a tedious task. If it can adapt to changing plans and preferences, it offers something more useful than a fixed reminder or a standard weekly order.

Now let it place the order automatically. The intended benefit is still convenience, but the experience has changed. It is spending money and committing the household to particular choices. A person who welcomed the prepared basket might want to approve the substitutions. Another might consider approval an irritating extra step that defeats the point.

What is the customer actually handing over?

A product concept can describe an outcome attractively while leaving the degree of autonomy vague. That ambiguity matters when interpreting an enthusiastic response: some people imagine assistance they can direct, others imagine a task disappearing entirely. Product and marketing both need to know which promise has been made.

What decides where people draw the line is mostly what a wrong action would cost, how easily it can be undone and who else it commits. An unsuitable meal suggestion can be ignored. An unwanted delivery takes more effort to undo.

At home the stakes rise. A system that changes the heating or locks the door acts for everyone in the house, and the person who set it up may not be the one who lives with its choices. Who in the household approves is part of the proposition.

“AI” doesn’t settle the value proposition

In experiments led by Washington State University researchers with more than 1,000 US adults, product descriptions that mentioned artificial intelligence lowered stated purchase intentions, and the drop was larger for higher-risk products such as costly electronics, medical devices and financial services. The label doesn’t carry the proposition; the benefit has to.

That is no reason to conceal how a product works. People need clear information about relevant functions and implications. They also need to understand why the experience would be better for them.

In the shopping example, the benefit might be less planning, fewer forgotten ingredients or an easier way to accommodate a change of plans. Those are different promises, for product and for marketing. A product that is excellent at one may disappoint someone who bought it for another.

More control isn’t automatically a better experience

It would be easy to conclude that every AI action should require confirmation. That could make an otherwise useful service exhausting.

A person buying relief from a recurring task has little appetite for supervising each step. Equally, automatic action can be an unwelcome reading of a promise to help. The value often lies in preparation more than completion, or in handling routine situations while making an exception easy to recognize.

The line also moves. A reassuring first demonstration doesn’t show how someone will feel after an unexpected result, and trust after the first mistake depends on how visible and how easy to undo that mistake was. Research that stops at a first reaction measures a relationship that hasn’t started yet.

The proposition becomes clear when the team can explain both what the product will take care of and what the customer still wants to decide. Our work on product and technology innovation starts from that question.

Common questions

What decides how much control customers will give an AI feature?

Mostly what a wrong action would cost, how easily it can be undone and who else it commits. People hand over routine, reversible choices more readily than ones that spend money, change the home or speak for someone else, and that line moves after the first time the product surprises them.

Does calling a product AI make people more likely to buy it?

Not by itself. In experiments with more than 1,000 US adults, Washington State University researchers found that mentioning AI in a product description lowered stated purchase intentions, and more so for higher-risk products. The benefit has to be clear on its own terms.

Sources

  1. Washington State University, “Using the term ‘artificial intelligence’ in product descriptions reduces purchase intentions,” WSU Insider (July 30, 2024). The university’s account of experimental surveys with more than 1,000 US adults, led by Mesut Cicek and published in the Journal of Hospitality Marketing & Management; the measure is stated purchase intention, and the negative response was “even stronger” for high-risk products and services.