For many consumers, this previous vacation season could have felt completely different. Perhaps an AI assistant steered presents your kinfolk would possibly truly like whereas filtering for gadgets that might arrive earlier than the vacations. Maybe it helped you navigate the specs of three completely different noise-canceling headphones or scanned 5 retailers for a selected vacation outfit, assembled a ready-to-buy basket, and politely requested, “Should I go ahead?”
This is the 12 months AI brokers stopped being an experiment and have become a part of how folks store, not in headline-grabbing methods however in on a regular basis moments—serving to buyers make sense of decisions, assemble baskets, resolve trade-offs, and transfer towards motion. Yet what appears like small comfort as we speak is an early sign of a a lot bigger shift in the best way we store. According to our analysis, even below average situations, AI brokers might mediate $3 trillion to $5 trillion of worldwide client commerce by 2030. Because brokers navigate the identical web as people—visiting web sites, participating with APIs, and interacting with loyalty packages—they will scale rapidly. And as they do, they’re reshaping how intent types, how merchandise are found, and the place worth swimming pools will be discovered.
We launched many of those themes in our report The agentic commerce alternative: How AI brokers are ushering in a brand new period for shoppers and retailers final fall. This article builds on that basis. Here, we discover what we name the “agentic commerce automation curve,” which illustrates how the consumer expertise shifts at completely different ranges of delegation, and description how retailers can put together for a world in which the shopper remains to be human however AI brokers more and more mediate key choices.
The six-level agentic commerce automation curve
The rise of agentic commerce displays the collision of three forces. First, AI brokers have reached decision-grade usefulness, permitting shoppers to delegate not solely inspiration but additionally shortlisting, meeting, and even execution. Second, the ecosystem now has rails for actual autonomy. Open-source protocols—resembling MCP, A2A, AP2, ACP, and UCP—allow brokers to learn information, negotiate with different brokers, and transact safely. The Linux Foundation not too long ago established the Agentic AI Foundation—a partner-backed effort together with Anthropic, Block, Google, Microsoft, OpenAI, and others—targeted on the interoperability, identification, and funds constructing blocks wanted to make autonomous commerce viable at scale. Third, intent is shifting upstream. Agents more and more act when client targets floor—resembling a dialog about an upcoming party, a calendar reminder for a visit, or a low-supplies sign from a tool. For retailers, the implications are stark: If your catalog, insurance policies, and worth proposition are usually not machine-readable, brokers—and by extension, buyers—merely won’t discover you, irrespective of how beloved your model is.
That mentioned, the rise of AI brokers doesn’t symbolize a single leap from human-driven buying to full autonomy. Instead, agentic commerce is unfolding alongside a curve—one outlined by how a lot of the commerce journey shoppers are keen to delegate to machines. This automation curve consists of six distinct ranges of automation, every representing a special mode of delegation—from fundamental rules-based comfort to totally autonomous multiagent coordination. Importantly, these ranges describe what brokers are technically able to doing, not what shoppers will all the time select to permit.
Further, adoption won’t essentially transfer uniformly “up” the curve. While agentic capabilities proceed to advance, two forces are shaping client delegation. The first is time and belief: As shoppers acquire familiarity with brokers and see them carry out reliably, they turn into comfy delegating bigger parts of the journey. The second is class dynamics. Willingness to delegate varies sharply by ticket measurement, emotional salience, identification signaling, and remorse threat. (This article focuses on the retail expertise; for a take a look at how brokers might affect B2B commerce, see sidebar, “How agentic commerce plays out in B2B.”)
Together, these forces decide a ceiling of delegation, in which autonomy naturally plateaus for a given class or second. In some contexts, shoppers could also be comfy delegating end-to-end execution. In others, they are going to intentionally cease brief, retaining management not as a result of brokers are incapable however as a result of human involvement is intrinsic to the worth of the expertise.
For these causes, the mannequin is finest understood as a curve somewhat than a ladder. Higher ranges of automation are usually not inherently higher or extra superior, and the purpose shouldn’t be most autonomy however optimum delegation (exhibit).
Level 0: Programmed comfort (‘set it and forget it’)
This degree is the pre-agentic baseline: Recurring replenishment for issues that run out—espresso pods, detergent, diapers, shampoo—is dealt with via subscriptions, scheduled refills, and recurring shipments. At this level on the curve, automation is rules-based—helpful however brittle and largely blind to context. When wants change, it breaks, and the human steps again in.
Still, degree 0 proves a foundational level at which shoppers delegate when automation is dependable and reversible. For instance, round 23 p.c of US Amazon buyers had not less than one lively Subscribe & Save order in 2024.
Level 1: Assist (‘the cognitive sidekick’)
At degree 1, brokers assist buyers assume and make choices, however they don’t execute. A client would possibly ask, “Find four gifts under $75 that can ship by Friday; prefer sustainable brands; and summarize trade-offs.” Or, in a extra advanced class, “Compare three noise-canceling headphones, and explain how they differ on sound quality, battery life, and comfort.” The agent’s position is analytical. It scans catalogs, parses critiques, compares options, and synthesizes choices into brief lists or suggestions. Crucially, it doesn’t decide to a configuration or resolve operational constraints. There isn’t any cart, no basket, and no readiness to transact. The human evaluates the choices, weighs trade-offs, and decides what to do subsequent.
In different phrases, degree 1 replaces search and comparability however leaves meeting and execution solely with the consumer.
Implications for retailers: Verifiable information beats advertising and marketing gloss. Agents require info they will parse and examine—structured attributes, clear eligibility guidelines, sizing and match certainty, and claims that may be substantiated.
Level 2: Assemble (‘the personal shopper’)
Level 2 marks a qualitative shift: Agents transfer from evaluation to orchestration. Here, the consumer expresses an intent, and the agent returns a purchase-ready basket. “Build a cozy winter outfit under $150.” “Stock a pantry for a vegan guest arriving tomorrow and staying for three days.” Or, extra advanced, “Put together a home office setup under $2,000 that supports dual 4K monitors and quiet video calls and has next-day delivery.” Unlike degree 1, the agent is tasked with resolving trade-offs and constraints somewhat than merely surfacing them. It selects particular gadgets, ensures technical compatibility, and balances efficiency towards value, availability towards supply pace, and promotions towards eligibility. Taxes, delivery home windows, loyalty advantages, and substitutions are dealt with by default. The output shouldn’t be an inventory of choices; it’s a coherent configuration that’s prepared to take a look at. The shopper’s position shifts accordingly from evaluating choices to approving or adjusting a proposed answer.
Implications for retailers: Success at degree 2 requires API-first merchandising. Inventory, pricing, delivery guarantees, promotions, and returns logic have to be uncovered cleanly so brokers can assemble baskets with human-level constancy.
Level 3: Authorize (‘the supervised executor’)
At degree 3, shoppers delegate not solely actions but additionally guidelines. Instead of approving every step, they authorize an agent to execute inside clear boundaries. “If groceries are under $120 and arrive Friday 6–8 p.m., place the order.” “If my preferred sneakers drop below $80 from merchants I trust, buy them.” The agent then runs the workflow finish to finish, selecting amongst eligible choices, swapping out-of-stock gadgets for accepted substitutes, making use of loyalty advantages, and escalating to the consumer for approval solely when one thing falls exterior the principles.
Implications for retailers: To help buyers at degree 3, retailers should make it potential for an agent to pay and act on a buyer’s behalf with security and transparency. That means buying authorization that may be restricted (by price range, time window, service provider, or class), exercise that may be audited (what was purchased and why), and actions that may be reversed (straightforward cancellations, refunds, and overrides when wanted).
Level 4: Autonomize (‘the intent steward’)
At degree 4, brokers function towards standing targets somewhat than one-off transactions. For instance, “Keep household essentials under $300 per month.” “Maintain my airline loyalty status at the lowest total cost over the course of 2026.” “Make sure we never run out of baby supplies.” The agent constantly displays wants, anticipates replenishment, compares choices throughout retailers, and optimizes for longer-term outcomes resembling sustaining or reaching a sure loyalty standing. The agent then handles the operational follow-through, together with adjustments, returns, and replacements. The shopper turns into episodic, stepping in primarily for significant choices or exceptions.
Implications for retailers: Competition at degree 4 shifts from successful a single buy to incomes a spot in the agent’s ongoing plan. Merchants want deeper integration—particularly round loyalty, eligibility, substitutions, and repair ensures—so brokers can cause about trade-offs and execute reliably. Put merely, it’s not sufficient to show a catalog; retailers should expose the principles and insurance policies that decide what “good” appears like.
Level 5: Networked autonomy (‘multiagent commerce’)
This forward-looking degree remains to be rising and factors to a world in which commerce turns into agent-to-agent by default. Personal brokers received’t simply work together with service provider web sites; they are going to negotiate instantly with a community of specialised brokers that optimize pricing, logistics and supply, cost authorization, and loyalty packages. Ultimately, it will outcome in multiagent marketplaces the place intent will be brokered, belief is carried via repute indicators, and transactions are settled via shared protocols—enabling “procurement as a service” to run constantly in the background.
Implications for retailers: Level 5 will likely be formed by these which can be already proficient at degree 4. Retailers that expose insurance policies, ensures, and loyalty logic in machine-readable methods will likely be positioned to affect how these ecosystems route demand. Those that don’t threat changing into interchangeable suppliers competing totally on value in machine-negotiated flows.
How the automation curve bends: Where delegation accelerates, plateaus, and reshapes worth
The automation curve describes what AI brokers can do throughout the buying expertise. It additionally might help clarify the best way that delegation can play out in follow and why automation doesn’t unfold evenly throughout classes, moments, or shoppers.
In the true world, shoppers don’t climb the curve uniformly, nor do they aspire to full autonomy in each context throughout buying classes. Instead, delegation accelerates the place automation removes friction with out sacrificing that means. It plateaus the place human involvement is intrinsic to worth, and it turns into selective amid trade-offs and uncertainty. Understanding these patterns is important for retailers deciding the place to take a position, what to show to brokers, and the way to compete in an agent-mediated world.
Where delegation accelerates: Utility, repetition, and low-regret purchases
In classes the place buying is primarily a process somewhat than an expertise, delegation tends to maneuver rapidly up the curve. Low-regret purchases resembling groceries, family necessities, and fundamental consumables are pure candidates for greater autonomy. Here, the worth of buying lies in effectivity, reliability, and predictability somewhat than discovery or expression.
As brokers show able to assembling baskets precisely, executing inside guardrails, and dealing with substitutions or supply adjustments gracefully, shoppers turn into comfy delegating execution solely. Attention shifts from evaluating choices to reviewing outcomes: Was the order on time? Did it keep inside price range? Were substitutions affordable? Over time, approval turns into implicit and intervention turns into the exception.
For retailers, this dynamic reshapes competitors. Brand storytelling and front-end expertise matter lower than operational belief. Agents optimize for delivered worth—elements resembling value, availability, service reliability, and reversibility. Merchants that expose clear stock information, predictable achievement efficiency, and clear substitution and return insurance policies turn into default suppliers, typically with out ever “winning” a standard second of consideration. In these classes, being agent-readable and reliable issues greater than being distinctive.
Where delegation plateaus: Identity, aspiration, and remorse threat
In high-consideration classes, resembling luxurious items or milestone purchases, delegation typically plateaus decrease on the curve. Here, buying shouldn’t be merely about outcomes; it’s about identification, intent, and emotional assurance. Consumers could enthusiastically enlist brokers to analysis, examine, and analyze however cease in need of absolutely autonomous execution.
Consider a luxurious purse buy: A client could ask an agent to guage how completely different manufacturers maintain worth over time, analyze resale markets, or assess how a selected type aligns with their private aesthetic. The agent could floor options, establish higher value factors in the resale market, or find in-store availability. But the ultimate choice and the transaction itself stay firmly human.
In these moments, the agent capabilities much less as an executor and extra as an analyst and curator. The ceiling of delegation is ready not by technical limitations however by emotional and identity-based concerns, resembling the need for a tactile expertise, social signaling, or the avoidance of remorse. Importantly, decrease autonomy doesn’t suggest decrease worth. In many such classes, human involvement is itself a key element of the product.
For manufacturers, this distinction is important. Competing successfully doesn’t require pushing shoppers towards full automation. It requires enabling brokers to help deliberation by exposing wealthy contextual attributes, provenance, craftsmanship, and long-term worth indicators whereas preserving human management on the level of dedication. In these classes, successful means shaping how choices are knowledgeable, not how rapidly they’re executed.
Where delegation is selective: Complexity, trade-offs, and context
Most classes sit between these two poles. In journey, client electronics, residence items, and different advanced purchases, delegation is selective and situational. Agents could autonomously deal with analysis, comparability, monitoring, and meeting whereas escalating choices that contain significant trade-offs. An AI journey agent, for instance, would possibly assemble an itinerary, optimize for loyalty advantages, and monitor for disruptions however nonetheless floor decisions that require judgment—time versus consolation, price versus flexibility. A house electronics agent could slim choices primarily based on specs and critiques however defer to the human when design, compatibility, or model desire turns into decisive.
In these classes, belief is constructed not via excellent execution however via explainability and reversibility. As autonomy will increase, shoppers wish to perceive not simply what the agent did however why it behaved in that method. Why did it select a selected possibility? Why did it make a substitution? Why did it escalate an exception? Graceful dealing with of edge instances issues greater than success on the joyful path.
This is the place metadata turns into technique. Humans infer that means intuitively, contemplating elements resembling match, really feel, temper, and suitability for a selected event. AI brokers, in fact, don’t. They depend on structured, contextual indicators. Products which can be emotionally legible to folks however semantically opaque to machines threat changing into invisible in agent-mediated flows. This requires retailers to take a position in wealthy, machine-readable attributes that allow brokers to behave with nuance—and to know when to pause and elevate inquiries to human buyers.
How worth swimming pools shift when brokers mediate commerce
Across these patterns, one shift is constant: the compression of the standard funnel. Search, comparability, and consideration collapse right into a single agent-mediated second. Continuous commerce replaces episodic choices. Loyalty turns into much less about sentiment and extra about coverage.
As a outcome, worth swimming pools migrate. Advantage accrues to retailers that may reliably execute towards agent constraints, not simply people who entice human consideration. Margins are formed by service ensures, achievement reliability, and readability of insurance policies. For some gamers, it will unlock effectivity and scale. For others, notably these depending on discovery-driven visitors, it introduces the chance of disintermediation.
Importantly, this doesn’t suggest a single finish state. The automation curve doesn’t prescribe the place each class ought to find yourself. Instead, it describes the cases the place delegation creates worth and the place it doesn’t. Retailers that acknowledge these contours early can make investments accordingly, pushing towards greater autonomy the place it reduces friction and intentionally preserving human moments the place they matter most.
The way forward for commerce shouldn’t be about maximizing automation. It is about inserting autonomy the place it enhances expertise, economics, and belief. The automation curve gives a sensible lens for making these decisions. Retailers that use it to information functionality funding, class technique, and agent readiness will likely be finest positioned to compete as AI brokers turn into an more and more central interface of commerce.
