Five scenarios for retail, the probabilities behind each, and the indicators that will tell you which future is arriving
Bottom-line first: physical retail is not going to disappear. We are asked that question from time to time, usually with a reference to whichever chain filed for bankruptcy that quarter, and it is the wrong fear; retailers who spend the next decade debating it will miss what actually happens. The store has now survived the catalog, the mall’s rise and fall, e-commerce, mobile, and social commerce, and it will survive AI too. What will not survive is the shelf: the entire apparatus retail has built over a century for getting products in front of human beings who browse.
WE PREDICT that retail in 2040 will still be a large and, for its survivors, genuinely profitable industry, but that the merchandising layer as we know it will be gone. When consumers delegate routine purchasing to AI agents, the customer stops shopping, and everything retail built to influence a shopping human loses its audience: endcaps, planograms, paid placement, promotional calendars, brand advertising, the impulse aisle. Commodity retail becomes a negotiation between the customer’s agent and the retailer’s systems. What remains for humans is everything a machine cannot want on your behalf: experience, identity, discovery you didn’t ask for, and trust. The retailers positioned on the right side of that split are being determined in the next thirty-six months, not in 2039.
That conclusion comes from looking at what retail actually sells, function by function, and asking which functions AI absorbs and which it structurally cannot. It also comes from watching sectors that sit eighteen to thirty-six months ahead of retail on the disintermediation curve; travel, media, and financial services have already run the experiment of a machine inserting itself between the customer and the catalog, and the pattern is consistent. The economics migrate to whoever owns the recommendation, and incumbents who kept their customer intelligence in silos discovered too late that they had become suppliers to someone else’s demand engine.
What Retail Actually Sells
Strip the industry down, and it’s clear retail sells six things:
- access and assortment (having the thing, or a thousand versions of the thing, when the customer wants it)
- price arbitrage (buying at wholesale scale and selling the spread)
- curation and discovery (deciding what deserves the customer’s attention)
- convenience and immediacy
- trust (returns, authenticity, and blame absorption when the purchase disappoints), and
- experience and identity (the reasons shopping is sometimes the point rather than the means)
AI absorbs the first three almost completely by 2040. An AI agent with real-time visibility into every seller’s inventory makes assortment a solved query rather than a competitive moat. Price arbitrage collapses when the buyer’s agent sees every price simultaneously and is structurally immune to promotion; a machine does not feel urgency from a countdown timer. And curation, the function retail has always been proudest of, becomes the agent’s core job, performed with knowledge of one customer’s actual preferences rather than a category manager’s bet about millions.
The last three resist automation for structural rather than technical reasons. Convenience becomes a logistics war that continues to escalate, but it is a war of infrastructure, not merchandising. Trust survives because someone must absorb the blame when the purchase is wrong; a consumer who delegates a decision still wants a named counterparty when it disappoints, and that need intensifies rather than disappears as delegation grows. And experience survives because identity does not delegate. Nobody outsources to an algorithm the feeling of finding something themselves; that is why retail’s future premium sits in engineered serendipity, the one form of discovery an optimizing agent cannot deliver. The 2040 retailer is a trust-and-experience layer, not a distribution layer, and the economics of that layer favor fewer, better operators.
The Agent Question
Every wave of retail technology has raised the same underlying question of who owns the customer, and every wave has answered it a little more ominously for the merchant. Loyalty programs, e-commerce accounts, and mobile apps were each attempts to hold proprietary demand data against a platform trying to intermediate it. The AI shopping agent is the final round of that game, because the agent doesn’t just influence the decision; it is the decision, for every purchase the customer no longer cares to make personally.
The structural question, then, is whose agent it will be. If the dominant agents belong to the consumer’s platforms (the ecosystems already running their email, calendar, and assistant), then brands lose the shelf, the ad unit, and the impulse buy simultaneously, and retail margin migrates to whoever the agent trusts. If retailers or brands manage to field agents consumers actually adopt, the customer relationship survives in merchant hands. The incumbents’ record in prior rounds of this contest is not encouraging, and the honest reading is that the demand layer is more likely to be won by the platforms than defended by the merchants. The retailers that thrive in that world will be the ones who saw it coming and built to be the counterparty agents prefer: clean data, transparent APIs into pricing and inventory, impeccable fulfillment telemetry, and a returns experience that makes their listings the low-risk choice in an agent’s ranking. Being easy for machines to buy from is about to become a merchandising discipline, and almost nobody has a team for it.
Five Futures, with Probabilities
Scenario 1: The Bifurcated Store (~32%)
The base case, and the winnable one. Retail splits cleanly along the line between purchases people care alot about vs. purchases they don’t. The don’t-care majority (replenishment, staples, commodity goods) flows agent-to-API through dark stores and automated fulfillment, with the storefront disappearing from the transaction entirely. The do-care minority consolidates into fewer, better physical venues: showrooms, flagships, food-anchored formats, and stores that function as media for the brand. Total store count falls sharply; revenue per surviving location rises. Headcount shifts from checkout and shelf-stocking to hospitality and fulfillment.
The dividing line in this scenario is not brand, capital, or location. It is whether a retailer’s customer intelligence lives in institutional systems that are clean, unified, and legible both to its own AI and to the agents it sells through, or scattered across the five disconnected systems most mid-market retailers actually run. This scenario requires no platform monopoly and no coordination problem to be solved; it is every player optimizing independently, which is why it carries the highest probability.
Scenario 2: Agent Platform Capture (~22%)
Two or three agent ecosystems win the consumer relationship outright and tax everything that flows through them. The platforms that already own the assistant, the search box, and the doorstep move from influencing purchases to executing them, and retailers become licensed fulfillment operators paying rent in fees, data, and margin on demand infrastructure they do not control. The residential analogy from our CRE analysis applies with full force: merchants operating inside a portal’s flywheel. Commodity categories go first; the open question is how far up the discretionary stack platform logic climbs before it hits categories where identity outweighs optimization.
Scenario 3: Brand Direct Displacement (~15%)
Displacement arrives not from platforms but from manufacturers. With agents handling discovery and logistics networks available as a service, the brands that once needed retail’s aggregation go direct; the retailer’s historic function of assembling an assortment under one roof dissolves when the roof is an API. Retail retreats to categories where multi-brand curation still matters and to its role as physical theater for the brands that remain. The tell for this scenario is the first major CPG company reporting more than half its revenue from direct, agent-mediated channels.
Scenario 4: Full Agentic Commerce (~9%)
The maximal version: agents negotiate with agents across transparent commercial infrastructure, prices are dynamic per customer per moment, and human involvement in routine consumption approaches zero. Technically plausible by 2040 for commodity goods, but it requires consumers to delegate not just replenishment but taste. Clothing, gifts, and food are purchases entangled with identity, and that is the binding constraint; wearing what the algorithm picked is a different act than letting it reorder detergent. The governance questions around autonomous spending authority also remain genuinely unsolved. Meaningful probability in staples; near zero across retail as a whole.
Scenario 5: Muddle-Through (~22%)
Efficiency gains everywhere, structural change nowhere. Agent adoption stalls on trust and liability questions, consumers prove stickier in their habits than technologists predict, and retail in 2040 looks like retail today with dramatically better tools: AI-driven demand forecasting, automated replenishment, personalized marketing, thinner corporate staff. History gives this outcome real credit; retail has absorbed five supposed extinction events in fifty years without structural displacement, and the industry’s sheer heterogeneity slows every transformation. But muddle-through is a probability, not a plan, and it is the only scenario in which doing nothing works.
The Land Is Being Claimed Now
Read the scenarios together and one pattern dominates. In four of the five futures, the merchandising layer, the shelf in every sense, is dead or dying, and the economics have moved to two places: the agent-facing infrastructure layer, where retailers compete on data quality and fulfillment reliability, and the experience layer, where they compete on reasons for a human to show up in person. In every scenario except muddle-through, the profits accrue to far fewer companies than are profitable in retail today.
There is a path to standing in 2040, and for the retailers on it, a path to better economics than the industry has seen in decades; commodity volume without stores to staff, and experiential formats commanding premiums the old middle never could. But the survivors above the trust ceiling share a common architecture: customer intelligence that belongs to the institution rather than to channel silos, operations legible enough for machines to buy from, and a physical footprint deliberately rebuilt around the purchases people still want to make themselves. That architecture takes years to build, and the retailers conflating it with this year’s personalization pilot are making the present more efficient while someone else claims their future.
What to Watch
Scenario probabilities are only useful if you can tell which future is arriving. Five leading indicators are worth tracking:
- Agent-initiated purchase share. Watch the percentage of e-commerce transactions initiated by an AI agent rather than a human session; this is the single cleanest measure of the shelf’s decline, and it will move slowly and then suddenly.
- The first agent-channel retailer. The first major retailer to publish agent-facing pricing and inventory APIs as a named strategic channel, with an executive who owns it, marks the moment machine-legibility becomes a competitive discipline rather than an IT project.
- Ad spend migration. When measurable budget moves from consumer-facing media into agent-optimization (the AEO successor to SEO), the industry has conceded that the audience for the shelf is no longer human.
- Private label share in agent-mediated categories. Agents optimizing on specification and price erode brand premium; a sharp private-label inflection in agent-heavy categories signals that brand equity no longer survives machine intermediation.
- The first agent-pricing lawsuit. Litigation over per-customer dynamic pricing or platform self-preferencing in agent rankings will mark the moment agentic commerce is large enough to fight over; it will also shape which scenario’s rules get written into law.
The Strategic Question
The question for a retail leadership team in 2026 is not whether AI will change the industry; that debate is finished everywhere except the industry’s own trade shows. The question is which scenario your company is positioned for, and whether your current investments are building toward the trust-and-experience layer of the bifurcated future or merely making the shelf more efficient. In four of five futures, no one will be looking at that shelf. Those are different projects with different architectures, and the retailers that conflate them will discover the difference at the worst possible moment.
Innovation Vista works with retailers to answer exactly that question: translating what sectors eighteen to thirty-six months ahead on the disintermediation curve have already learned, and turning it into a positioning strategy for the split ahead. If you want to pressure-test which side of 2040 your company is building toward, that conversation is where we start.
