Read time: 8 minutes
Highlights
Agentic shopping could compress food choice into a handful of algorithmic winners, or unlock an entirely new universe of personalised products.
- The Shortlist Economy: AI agents filter millions of products down to a small number of trusted, available and machine-readable winners.
- The Infinite Shelf: Flexible manufacturing and mass personalisation create products for increasingly specific dietary, sensory, cultural and life-stage needs.
- The strategic uncertainty: The outcome will depend on who owns the agent, how it gets rewarded and how flexible manufacturing becomes.
- The leadership imperative: Food and beverage companies should prepare for both futures through agent-ready data, flexible platforms and Strategic Foresight.
Agentic shopping could completely reshape what consumers see, what brands get chosen and how many successful food products the market can support.
Just think about this for a minute…
Today, a shopper can walk into a supermarket and face thousands of products. Online, the possible range is even larger. Search, filters, reviews and recommendations help consumers navigate the abundance, but the final decision still belongs largely to the individual.
That’s about to change.
An AI agent could soon interpret a consumer’s goals, remember their preferences, compare available products and assemble a basket with minimal human involvement. The shopper might say, “Find me high-protein breakfasts that are low in sugar, affordable, suitable for my family and available for delivery tomorrow.”
The agent won’t show everything.
It’ll decide what deserves attention.
But there are two very different ways this could unfold.
In one future, agents compress the market into a small number of trusted winners. In the other, agents make it economically viable to create an explosion of products designed for very specific people and moments.
These futures aren’t mutually exclusive. They could coexist across categories, markets and shopping occasions.
Since Futurists don’t predict the future, I want to explore both alternative futures.
Why agentic shopping changes the shelf
The first major shift is from search to conversation.
The consumer no longer needs to know the product name, category language or technical specifications. They can describe an outcome, a frustration or a personal constraint. The agent then translates that request into a set of product requirements.
Research from NielsenIQ on AI-powered personalised shopping describes a progression from AI-assisted shopping, through AI-mediated discovery, to true agentic commerce. In the final stage, the system doesn’t simply recommend products. It can bundle, reorder and potentially purchase on the consumer’s behalf.
That creates a new competitive arena.
A product has to be understood by the consumer, but it also has to be legible to the machine. Ingredients, allergens, nutrition, certifications, claims, price, availability, delivery performance and reviews all become decision inputs.
As Deloitte explains in its analysis of the algorithmic shelf, products could lose consideration without losing awareness. A brand might remain famous, well-funded and widely distributed, yet still be excluded from the moment of selection because an agent can’t confidently interpret or verify its value.
The question for food leaders is no longer only whether consumers know your brand.
It’s whether their agents can justify choosing it.
Alternative Future One: The Shortlist Economy
In this future, the visible product range becomes dramatically smaller.
The underlying catalogue might contain millions of products. The consumer, however, sees three options, or perhaps one confident recommendation. The agent has already filtered out the rest.
Welcome to the Shortlist Economy!
How the shortlist forms
The agent evaluates products against a consumer’s stated and inferred requirements. It considers price, availability, delivery time, nutritional composition, allergies, dietary preferences, previous purchases, ratings, returns, reliability and trust.
A product with complete, consistent and verified data has an advantage.
A product with missing information, inconsistent claims or unreliable availability may never appear.
This is the central dynamic. The agent doesn’t need to display every relevant product. It only needs to present the products it believes are most likely to satisfy the consumer.
The McKinsey analysis of agentic commerce points towards a compression of the consideration set. Agents can absorb a huge amount of information, but consumers still have limited attention. The machine evaluates broadly, then presents narrowly.
That could make shopping much easier.
It could also make the market much harsher.
The rise of algorithmic winners
Products that perform well inside the agent’s decision system could benefit from powerful feedback loops.
An agent recommends a product because it has strong reviews, dependable stock and clear nutritional data. The consumer buys it. The purchase generates another positive review and another reliable fulfilment record. The product becomes even easier for future agents to trust.
This creates an algorithmic flywheel.
The winners become more visible because they were already visible. They accumulate more sales, more reviews, more data and more evidence. Smaller or newer products struggle to break in because they lack the history required to earn confidence.
This would reward incumbent brands, but not automatically.
Scale alone wouldn’t be enough. A large brand with poor product data or inconsistent availability could lose to a smaller competitor with a sharper use case and better machine-readable proof.
The winners could be the brands that are easiest to understand, easiest to compare and least likely to create a bad outcome.
That’s a major change from traditional brand building.
What becomes valuable
In the Shortlist Economy, successful products are likely to have several characteristics.
They’re machine-readable. Product information is structured, complete and consistent across the brand site, retailer listings, marketplaces and third-party sources.
They have a clear job to do. The agent can easily understand whether the product is for a high-protein breakfast, a child’s lunchbox, a low-allergen household or a specific cooking occasion.
They’re reliably available. A product that’s frequently out of stock becomes a poor recommendation, regardless of how strong the brand may be.
Their claims are provable. Broad language such as “better for you” is less useful than specific, evidence-backed attributes that an agent can evaluate.
They’re operationally dependable. Delivery accuracy, substitution outcomes, refund policies and repeat purchase behaviour all become part of the brand experience.
Bain’s research on autonomous shopping and the changing customer journey highlights the importance of trust, fulfilment and data ownership. In an agent-mediated environment, a failed delivery or inaccurate product description isn’t simply a service issue. It can cause the agent to exclude the product next time.
The risks of a smaller visible market
The Shortlist Economy could reduce choice overload. That’s a genuine benefit. Consumers may spend less time searching and feel more confident about the product they select.
But here’s the problem…
A small visible range can become a small meaningful range. Products that aren’t recommended may effectively disappear from the consumer’s world.
That creates several concerns.
Discovery could decline. Consumers may never encounter unusual flavours, emerging brands, regional products or new formats that sit outside their established preferences.
Food culture could homogenise. If the same agents recommend the same products across markets, local differences could be flattened into a global average.
Commercial bias could increase. The agent may be rewarded through commissions, retail media, preferred access or platform relationships. The recommendation might appear objective while reflecting hidden commercial incentives.
Innovation could become invisible. A new product with no purchase history may be unable to compete with products that have years of accumulated data.
Brands could become interchangeable. If the agent reduces products to price, claims, nutrition and fulfilment, emotional differentiation may become harder to maintain.
The Ipsos research on shopping with AI is a useful reminder that consumers aren’t necessarily ready to surrender all authority. Many are comfortable using AI for research and comparison, but far fewer currently want fully autonomous purchasing.
That means the shortlist may be approved by a human for some time.
But the shortlist will increasingly be created by a machine.
Alternative Future Two: The Infinite Shelf
Now let’s move to the opposite future.
In the Infinite Shelf, the consumer still sees a shortlist. But behind that shortlist is a vastly expanded product universe.
The agent doesn’t simply select from a fixed range. It aggregates small pockets of demand and helps the food system produce economically viable products for increasingly precise needs.
This is mass personalisation.
From mass production to demand of one
A household might want a breakfast product that is high in protein, low in sugar, dairy-free, gentle on digestion, culturally familiar, suitable for a child and packaged in a particular format.
Historically, that combination of needs might represent too little demand to justify a separate product.
Agentic shopping changes the economics by finding similar micro-demands across thousands or millions of consumers. It can identify enough people with related needs to support a product, a subscription or a modular formulation.
This is the logic behind the Made for Me Economy. Personalisation becomes more than a recommendation layer. It becomes a product development and commercial model.
The technologies behind the Infinite Shelf
Several technologies could make this future possible.
Flexible manufacturing allows production lines to switch between formulations, pack sizes and flavour profiles with less downtime.
Modular formulations let manufacturers combine a base product with interchangeable ingredients, nutrients, flavours or functional components.
AI-assisted NPD can identify unmet needs, model consumer response, optimise formulations and accelerate testing.
Robotics can support high-mix, lower-volume production and more precise assembly.
Precision fermentation could create highly specific ingredients with targeted functional or sensory properties.
Subscriptions and replenishment models can create predictable demand for niche products.
Made-to-order production reduces the need to manufacture large volumes before demand is proven.
The result could be an infinite shelf that consumers never physically see. The agent searches the wider universe, identifies a small number of relevant options and may even help trigger production.
Personalisation becomes a production signal
Imagine an agent learning that a consumer consistently rejects certain textures, prefers particular cultural flavour profiles and needs products compatible with a specific metabolic goal.
Today, the agent might recommend the closest available product.
In the Infinite Shelf, it could identify that thousands of similar consumers share the same unmet need. That insight could support a new formulation, a limited regional launch or a subscription-based product platform.
This is a profound shift.
Personalisation would no longer be only about selling the right product to the right person. It would help determine which products should exist in the first place.
My recent thinking on nutrigenomics and NPD strategy explores how deeper biological data could eventually influence product design. Agentic shopping could become the commercial interface that connects those insights with actual purchasing behaviour.
And as I’ve explored in The Butyrate Breakthrough, AI and robotics could work together to design and produce more targeted functional food products.
The risks of an infinite product universe
The Infinite Shelf sounds exciting. I’m genuinely fascinated by the possibility of food products that serve people who’ve been poorly served by mass-market assumptions!
But it comes with serious challenges.
Regulatory complexity could explode. Every new formulation, claim, ingredient combination and personalised recommendation could create additional compliance requirements.
Data ownership would become critical. Who owns the consumer’s preference profile, health data and purchasing history? The retailer, the agent platform, the manufacturer or the consumer?
Personalisation could become unequal. Wealthier consumers may access highly tailored products while others are directed towards standardised, lower-cost alternatives.
Consumers could become confused. A supermarket aisle with thousands of variants may be replaced by an algorithmic conversation that is difficult to understand or challenge.
Personalisation could become a marketing label. A product might be described as “made for you” because the packaging or recommendation is personalised, even though the formulation itself hasn’t meaningfully changed.
Manufacturing could become more fragile. High product variety creates demands for forecasting, quality assurance, inventory management and traceability. A small error could affect a highly specific consumer group.
The biggest risk is that personalisation creates the appearance of precision without delivering meaningful outcomes.
A name on the pack isn’t personalisation.
A recommendation based on weak data isn’t personalisation.
A product that genuinely improves fit, function or experience is personalisation.
The two futures can coexist
I don’t believe the food industry will necessarily choose one future.
The Shortlist Economy is likely to dominate routine, low-risk and highly replenishable categories. Consumers may delegate everyday decisions about milk, bread, breakfast staples, snacks and household beverages to trusted agents.
The Infinite Shelf may emerge more strongly in health, wellness, sports nutrition, premium foods, culturally specific products and complex dietary needs.
Even within the same category, both futures could operate simultaneously.
A major brand may have a few hero products that win algorithmic visibility at scale. Alongside them, it may operate a flexible platform that generates small-batch products for specific consumer communities.
The visible range could shrink while the underlying product universe expands.
That’s the paradox.
The strategic question isn’t whether there’ll be more products or fewer products. It’s where the complexity will sit, who will control it and who will capture the value.
The uncertainties that matter most
I see eight strategic uncertainties shaping the outcome.
Who owns the shopping agent? A retailer-owned agent may optimise for margin, availability and private label. A consumer-owned agent may optimise for personal fit. A brand-owned agent may optimise for loyalty and its own portfolio.
How is the agent rewarded? The result changes dramatically if the agent is paid for the best match, the lowest price, the highest commission or the greatest basket value.
How flexible is manufacturing? Personalisation can’t scale if factories, packaging lines and supply chains are built only for long production runs and stable forecasts.
How much will consumers delegate? Some people will want full automation for replenishment. Others will insist on human approval, especially for health, children’s food and culturally important products.
What will regulation permit? Data use, health claims, personalised nutrition and automated purchasing will all face different regulatory boundaries across markets.
Who owns the data? The value of personalisation depends on access to high-quality consumer data, but access without clear consent could destroy trust.
Will agents interoperate? Open standards could create a broad ecosystem. Closed platforms could fragment commerce into competing data and transaction silos.
Who gets access to quality data? If only the largest companies can afford agent-ready infrastructure, the Infinite Shelf may become less diverse rather than more.
These uncertainties deserve active scenario planning now. Waiting for the market to settle could mean discovering that somebody else has already designed the rules; and you may not like those rules!
What food and beverage leaders should do now
I’d recommend following the above uncertainties preparing for both futures at the same time until the outcome becomes apparent.
Build agent-ready product data
Create a single, governed source of truth for ingredients, allergens, nutrition, claims, certifications, pack formats, use occasions, price, availability and fulfilment.
Make the information structured and machine-readable.
Then test how different agents interpret it. Don’t assume that clean data in your internal system is automatically visible or understandable in the market.
Define defensible consumer use cases
Don’t begin with “How do we use AI?”
Begin with “Which consumer problem can we solve better?”
It might be affordable high-protein breakfasts, culturally relevant family meals, personalised sports nutrition or reliable allergy-safe replenishment.
A strong use case gives the agent something meaningful to optimise.
Protect hero products while building flexible platforms
You’ll probably need both.
Hero products create scale, trust and recognisable demand in the Shortlist Economy. Flexible platforms create the ability to serve specific needs in the Infinite Shelf.
The future portfolio may look less like a fixed list of SKUs and more like a combination of trusted anchors, modular components and adaptive services.
Invest in evidence and trust
Make claims verifiable.
Explain why a product is recommended. Show the trade-offs. Give consumers control over their data and the ability to override the agent.
Trust won’t come from saying that the system is intelligent. It’ll come from demonstrating that the system is reliable, transparent and aligned with the consumer’s interests.
Measure more than market share
Start tracking agent visibility, share of recommendation, product data completeness, use-case inclusion, repeat selection, substitution outcomes and recommendation quality.
The old measures still matter, but they won’t tell the whole story.
Use scenario planning
Map what happens under both futures.
Ask what your portfolio, manufacturing network, data architecture, brand strategy and regulatory model would look like if agents reduced your visible range to five winners.
Then ask what would happen if consumers could access thousands of viable personalised products.
This is exactly where FutureCUBED™ becomes valuable. Horizon Scanning identifies the technologies and behaviours that could reshape the market. Future Foresight maps the alternative futures and critical uncertainties. Strategy and Implementation turns those insights into decisions, experiments and capability building.
The choice may not be yours
The most important point is this…
Food and beverage leaders may not get to choose whether the market becomes a Shortlist Economy or an Infinite Shelf.
Agents, retailers, manufacturers, regulators and consumers will shape it together.
But you can choose whether your organisation is ready for both.
You can make your products easy for agents to understand. You can build evidence that earns trust. You can protect your hero products while developing flexible platforms. You can experiment with subscriptions, modular formulations and made-to-order production.
And you can decide what you’ll never delegate.
Because the future of food may contain fewer successful products, more successful products or both at the same time.
The real question is whether your business will be one of the products that agents can find, trust and recommend… or whether it’ll be invisible on a shelf that technically contains everything.
To continue the conversation email me at tony@futuristforfood.com
