Blog Post
- Research Topics:
- Industry Channels:
Co-Pilot, Not Autopilot: How Much Control Shoppers Will Give Grocery AI
Shoppers want AI to help them make better decisions, rather than take over the shopping trip. The insight for the industry is to offer AI as a co-pilot, not a decision-making autopilot.
New research from the Coca‑Cola Retailing Research Council asked shoppers how much of their shopping they would hand to AI. The answer is clear, and it sits well short of automation.
| Preferred level of AI control | Share |
|---|---|
| AI gives advice; shopper decides | 37% |
| AI builds the cart; shopper approves | 20% |
| AI shops automatically, including checkout | 4% |
| Would not use an AI assistant | 39% |
Nearly 60% of shoppers make a similar point: they would let AI help advise or build the cart, but not make decisions for them.
The shopper intends to stay in charge of the budget, the brand choices, the dietary restrictions, and the final cart. That does not show a reluctance toward technology, but rather a desire to use it as an aid.
The cart-builder: a high-leverage opportunity
An AI cart-builder feature deserves particular attention because it sits between the extremes in the table above: more than advice the shopper has to act on alone, well short of the automation almost nobody asked for. It gives AI a meaningful role while protecting shopper agency.
It also connects naturally to infrastructure most retailers already have:
- Past purchase history
- Loyalty data
- Digital lists
- Meal plans
- Substitutions
For many retailers, this is the best near-term path to employing AI.
What a reviewable cart has to show
The AI does not have to be right every time, but it has to show its work. A cart that answers these four questions leaves the shopper in control, and that is how an AI feature moves from trial to habit:
- Why did you add these items?
- What am I saving?
- What did you substitute?
- What can I change?
Reviewable, explainable, correctable. A shopper who can audit the cart will tolerate an imperfect one. A shopper who cannot will abandon a good one.
Get substitutions right
Whenever a shopper cannot find the exact product, size or variety they want, substitutions are necessary, and the study finds that this can be a visible pain point for an AI tool. But it is also one of the clearest chances to build trust.
The fix is to stop inferring what an acceptable swap looks like and let shoppers define it in advance based on:
- Price
- Size
- Quality
- Nutrition
- Brand preference
A swap that follows the shopper’s own rules is service; a swap that surprises them is a reason to stop delegating.
For brands, this is where shelf logic changes. When a shopper approves a cart instead of picking every item, the substitution rule decides which product actually ships. The study’s guidance to CPG partners is to understand when and why their products get swapped out, and where quality or value proof is needed to prevent it.
Keep a strong non-AI path
The largest single group in the data would not use an AI grocery assistant at all. They matter, and the study is direct about why: a strong AI strategy needs a strong non-AI path beside it. Trust is damaged when shoppers feel pushed into a system they did not ask for.
That is a practical constraint on the roadmap, not a footnote. If the AI path becomes the only maintained path, roughly two in five shoppers get a worse experience than they had before.
This is one theme from the CCRRC AI in Grocery Consumer Study, a survey of 3,014 U.S. primary grocery shoppers conducted with Think Blue and NIQ research support. Value First covers what shoppers most want that assistance to deliver. The full study publishes this fall at ccrrc.org. Two findings left in this series: the planning window where the basket takes shape, and the households adopting fastest.