Retail has always evolved in response to how customers shop. Catalogues gave way to department stores, malls to ecommerce, desktop to mobile — each transition rewired the industry around a new buying behaviour. The next transition is different. The customer is no longer human.
AI agents are entering the buying journey at every stage — discovering products, evaluating options, negotiating prices, completing purchases, and managing post-purchase experiences. McKinsey's research on agentic commerce projects this shift could generate up to $1 trillion in US retail revenue by 2030, with global projections reaching $3 to $5 trillion. In B2B, the stakes climb higher still: Gartner projects AI agents will command $15 trillion in B2B purchases by 2028.
These are not aspirational forecasts. Google has launched agentic checkout across Search and Gemini. Amazon's "Buy for Me" agent shops competitor websites. OpenAI's Operator and Google's Project Mariner complete multi-step purchases autonomously. In B2B, buyer-side agents are negotiating with seller-side agents through API flows with clean audit trails. This article maps the use cases driving agentic commerce across both retail and B2B — the workflows agents are taking over, the data they require, and the infrastructure businesses must build.
Key Takeaways
- McKinsey projects agentic commerce could generate up to $1 trillion in US retail revenue by 2030, with global projections of $3 to $5 trillion. Gartner projects AI agents will command $15 trillion in B2B purchases by 2028.
- Agentic checkout is live: Google AI Mode, Gemini, Amazon's "Buy for Me," OpenAI's Operator, and Google Project Mariner are all shipping products — not research prototypes.
- Roughly 70% of online shopping carts are abandoned before purchase. Agentic checkout eliminates every friction point that causes this — the agent already has payment, shipping, and preferences stored.
- Eight retail and five B2B use cases are moving from pilot to production — spanning conversational discovery, autonomous checkout, automated replenishment, agent-led negotiation, and machine-readable procurement.
- The winners will not be the largest or cheapest — they will be the most agent-accessible, with structured product data, API-accessible catalogues, machine-readable terms, and transparent pricing.
What Makes Agentic Commerce Different
Agentic commerce is the progressive delegation of purchasing decisions from humans to AI agents — agents that interpret intent, evaluate options, negotiate terms, execute transactions, and manage post-purchase workflows. Every manual step is a candidate for automation, and the more steps, the more value unlocked.
The delegation is fundamentally different in consumer and B2B contexts. In consumer commerce, an agent acts on behalf of a single person, with simple guardrails: personal budget and preference. In B2B, delegation is institutional — authority flows from procurement policies, budget owners, risk teams, and legal frameworks. A purchase order can draft itself, verify pricing against contract terms, check supplier eligibility, negotiate within defined boundaries, and leave a clean audit trail, without anyone chasing approvals that sit in inboxes for days.
What unites both contexts is that the buyer is now a piece of software. It does not browse visually or admire hero images. It parses data, evaluates structured fields, and moves to the next option the moment yours returns noise instead of signal.
Retail Use Cases
The retail journey — from the moment a need forms to the moment a delivery arrives — is being rewritten one use case at a time.
Conversational Product Discovery
The most visible shift is how products get found. Traditional discovery relies on keyword search and filter-based browsing — the process forces the buyer to do the thinking. Agentic discovery inverts this. A shopper tells an AI agent "I need trail running shoes for muddy terrain, under $150, with good ankle support" and the agent handles the rest — querying data sources, evaluating specifications, cross-referencing reviews, and returning a curated shortlist. It interprets the use case, physical requirement, and budget simultaneously.
Microsoft's analysis of agentic commerce describes this as the new "front door" to retail — the AI conversation replaces the homepage. Google's Shopping Graph now connects AI agents directly to merchant product feeds, and shopping-related searches on generative AI platforms grew dramatically between 2024 and 2025.
Products that win in agentic discovery are not those with the best keyword optimisation. They are those with the most complete, structured, machine-readable information. An agent cannot cite "premium quality materials" — it can cite "recycled ocean plastic, 600D denier, water-resistant coating." Products structured through Schema.org Product markup give agents citable material. This is the same principle that determines whether AI search engines can find your business at all.
Context-Aware Personalisation
Traditional personalisation shows you products based on what you browsed last week. Agentic personalisation is different — the agent carries the customer's full context: wardrobe inventory, calendar, budget, sizing across brands, style preferences. When this agent approaches a store, it already knows what fits.
A shopper tells their assistant: "I need an outfit for a rainy outdoor wedding in Scotland next month." The agent cross-references weather, style history, available inventory, dress code norms, and budget — then returns a complete outfit, sized correctly, available before the event. Bain & Company's research found that 81 percent of shoppers prefer brands that personalise their experience, and agentic personalisation goes further — it anticipates needs the shopper has not yet articulated. Product data must include the contextual attributes agents use for matching: occasion, weather appropriateness, compatibility. Retailers who optimise their structured data for AI engines gain a decisive advantage.
Virtual Fit and Compatibility Analysis
Returns are the hidden tax on ecommerce, and fit uncertainty drives a significant share of them. Vertical AI agents address this directly — virtual try-on agents visualise how a garment fits a specific frame, Fit Analyser agents cross-reference body measurements with brand-specific sizing, compatibility agents verify that accessories work with existing devices, and spatial agents check whether a sofa fits through a doorway. The common thread is that the agent eliminates the guesswork between browsing and buying — and enabling it requires detailed, machine-readable specifications: precise measurements, material compositions, compatibility matrices, and standardised attribute schemas.
Dynamic Price and Promotion Evaluation
Price comparison is one of the first tasks consumers delegate to AI agents, but agentic price evaluation goes beyond finding the lowest sticker price. The agent evaluates total cost of ownership — product price, shipping, delivery time, return policy, loyalty points, promotions — across multiple retailers simultaneously. Retailer A might have the lowest price but charges for shipping; Retailer B includes free shipping and 10 percent cashback; Retailer C bundles a free carrying case. The agent presents the total-value comparison, not the price comparison.
Retailers competing on price alone may lose to those competing on total value. For brands preparing for agentic commerce, pricing and promotion data must be structured, accurate, and machine-readable. Hidden fees, unclear shipping terms, or inconsistent loyalty calculations cause agents to deprioritise or skip a retailer entirely.
Autonomous Checkout and Purchase Completion
The shopping cart has been the centrepiece of online commerce for twenty years. Agentic checkout makes it obsolete. When an agent has evaluated options, confirmed fit, and selected the best offer, it proceeds directly to purchase — applying loyalty points, selecting the optimal payment method, choosing a delivery slot, and completing the transaction without a single click from the buyer.
A consumer tells their assistant: "Order me the same running shoes I bought last year, in the new colourway, from whichever store has the best price with free shipping." The agent searches across retailers, compares prices and shipping, navigates to the cheapest option, fills in the address, applies discount codes, processes payment using stored credentials, and confirms the order. The consumer never visits a website, sees a checkout page, or types into a form.
Google launched agentic checkout across Search (AI Mode) and Gemini, enabling agents to execute purchases directly on merchant websites. Amazon's "Buy for Me" agent shops competitor websites when products are unavailable on its own platform. OpenAI's Operator and Google's Project Mariner complete multi-step purchases across the web. These are production features processing real transactions.
The cart abandonment problem evaporates. Baymard Institute's research shows that roughly 70 percent of online shopping carts are abandoned before purchase. Forced account creation, excessive form fields, unexpected costs, payment friction — agentic checkout eliminates every one of these barriers. The agent already has payment details, shipping address, and preferences stored. There is nothing to fill in, nothing to abandon.
The deeper structural change is that the cart itself becomes optional. If an agent can go from product selection to completed purchase in a single automated flow, the "add to cart → review → checkout" sequence is unnecessary overhead. Agents also do not tolerate what humans patiently click through — dark patterns, CAPTCHA walls, and JavaScript-heavy interactions cause an agent to abort entirely. Transparency is mechanically required, not just ethical. The technical readiness for agentic commerce includes ensuring checkout APIs work for autonomous agents, not just human browsers.
Automated Replenishment and Recurring Purchases
Consumable products follow predictable consumption patterns, and agentic commerce turns these patterns into automated purchasing. Instead of the shopper remembering to reorder, the agent monitors consumption rates and places orders at the optimal time.
This is not a subscription. Subscriptions deliver on a fixed schedule regardless of consumption. Agentic replenishment adapts — if the household used less coffee while travelling, the reorder is delayed. If they hosted a dinner party and consumed more olive oil, it is accelerated. The agent also optimises across retailers for each cycle. For grocery and FMCG retailers, this shifts the competitive landscape from shelf placement and impulse buying to data accuracy and fulfilment reliability. Loyalty is earned per transaction, not assumed per subscription.
Post-Purchase Experience Management
The agentic journey does not end at checkout. Agents increasingly manage the post-purchase experience — tracking deliveries, initiating returns, scheduling installations, requesting warranty service, providing setup guidance. Real-time merchandising intelligence also flows back through these interactions. When 12 percent of shoppers ask whether a summer collection comes in petite sizes, that signal reaches merchandising teams automatically — in real time, not through a quarterly survey. Order status APIs, return policy schemas, warranty terms, and product documentation become inputs for agent-managed experiences. Brands that build trust with AI agents through consistent post-purchase data earn repeat recommendations.
Agent-to-Agent Negotiation
The most forward-looking retail use case is machine-to-machine negotiation. A consumer's personal agent interacts directly with a retailer's agent to negotiate pricing, delivery terms, bundles, and loyalty rewards — within parameters set by both parties. The consumer sets guardrails: "I want this TV, but I won't pay more than $800 including delivery, and I need it before Saturday." The retailer's agent evaluates margin flexibility, inventory, delivery capacity, and competitive pricing before making a counteroffer. The negotiation happens in seconds — or the consumer's agent moves on.
IBM's analysis of agentic commerce describes agent-to-agent interaction as the logical evolution of autonomous purchasing. SAP, Salesforce, and Shopify are all building infrastructure that enables merchant-side agents to respond to buyer-side agents with structured offers. Fixed product-page prices become one input among many. Dynamic pricing engines and structured negotiation protocols become essential infrastructure for capturing agentic transactions.
B2B Use Cases
Consumer agentic commerce gets the headlines, but the bigger transformation is happening in B2B — where the stakes per transaction are higher, the workflows are more complex, and the efficiency gains are measured in millions. Forrester analysts expect a material share of B2B sellers to face agent-led quote negotiations before the end of 2026.
Autonomous Procurement and Reordering
The most immediately deployable B2B use case. An agent monitors inventory, usage patterns, or consumption signals and triggers reorders before stockouts occur. It validates contract pricing, checks vendor eligibility, applies negotiated discounts, and places the order — escalating to a human only when something falls outside defined parameters.
In manufacturing and MRO, this eliminates parts arriving too late. The value is continuity — the right part arrives before downtime hits. Medical and pharmaceutical procurement adds another layer: agents enforce strict rules on approved items, maintain audit trails, and provide full explainability. These constraints make agents more valuable, not less — they enforce compliance consistently in a way manual processes cannot.
Agent-Led Price Negotiation
This is where B2B diverges most dramatically from its consumer counterpart. When terms are machine-readable and boundaries are explicit, AI agents on both sides of a transaction can negotiate autonomously. A buyer agent needs 1,000 units. The supplier API returns a base offer at $100 per unit. The buyer counters at $92 with a four-week window. The supplier responds at $98 with three-week delivery. The buyer offers $95. The supplier accepts — within its guardrail — and the final terms lock into the order. Every counteroffer is logged. Enterprises can orchestrate hundreds of these interactions simultaneously, producing a cleaner audit trail than most human-led negotiations.
Intelligent Quote-to-Order Workflows
B2B buying cycles are notoriously long. Agentic workflows compress this. A business buyer instructs an agent to rebuild a cart from a contract catalogue — the agent pulls the right SKUs, applies contracted pricing, checks compliance against purchasing policies, and routes the quote for approval within spend limits. If the total falls under the auto-approval threshold, the order goes through. If it exceeds the threshold, the agent packages the request with full context and routes it to the right approver. McKinsey has highlighted that autonomous sourcing agents can meaningfully compress procurement cycle times while simultaneously improving compliance consistency.
Payments and Settlement Automation
A significant share of B2B payment workflows will leverage AI agents by the end of 2026. This is not invoice-processing automation — it is making payments a strategic backbone. Agentic payment workflows handle dynamic discounting, real-time reconciliation across suppliers and currencies, and automated exception handling for mismatches between POs, invoices, and goods received. Marketplaces that embed flexible payment orchestration — virtual cards, real-time settlement, dynamic terms — become the infrastructure that enables fully autonomous trade.
Wholesale and Distribution Order Validation
In wholesale electronics, agents validate component compatibility, apply tiered pricing rules, check inventory allocation, and place complex multi-line orders — catching errors that would otherwise result in returns, restocking fees, and production delays. Fewer returns means more reliable supply chains and lower operational costs across the network.
The Adoption Gap — and the Early-Mover Window
Not every industry moves at the same pace. Technology firms lead in familiarity with agentic AI. Services firms follow. Goods and manufacturing trail noticeably behind. A separate gap exists within businesses already using AI — a large share of B2B suppliers have adopted AI in sales, but far fewer have deployed autonomous agentic AI specifically. That gap is where early movers build durable advantages. For suppliers in industrial, manufacturing, and construction sectors, the window to establish agent-friendly infrastructure before competitors is wide open. The businesses that structure their data and expose their catalogues through APIs now will be the ones AI agents discover and prefer when autonomous procurement scales.
What Agents Need From Your Business
Whether the agent is buying running shoes on behalf of a consumer or reordering industrial components on behalf of a procurement team, the requirements converge. Agents parse data, not design.
Structured, machine-readable product data. Every specification — dimensions, materials, certifications, compatibility, lead times, minimum order quantities — available as structured fields, not buried in PDFs. Implement Schema.org Product schema with complete attributes. This is the single most important factor in whether an agent can evaluate your offerings at all.
API-accessible catalogues and pricing. If your catalogue exists only as a human-browsable website, you are invisible to autonomous buyers. The shift from "website as storefront" to "API as storefront" is the infrastructure transformation agentic commerce demands.
Machine-readable terms and policies. Pricing tiers, volume discounts, payment terms, return policies, and shipping options need to be in parseable formats — not natural language policy pages.
Compliance and audit capabilities. Procurement in regulated industries requires explainability. Agents must log every decision and demonstrate policy compliance.
Transparent, real-time pricing and inventory. One inaccurate data point and the agent moves on. Hidden fees and stale inventory numbers are conversion killers — agents have no mechanism to tolerate them.
Agent-friendly checkout infrastructure and measurable trust signals. CAPTCHA walls, JavaScript-heavy cart interactions, and forced account creation block agent purchases entirely. Reviews, ratings, and certifications become confidence inputs agents verify before recommending.
A 90-Day Starting Point
Businesses paralysed by the scale of the shift can start with a focused pilot. The most successful early implementations — retail and B2B — follow the same staged approach.
Weeks 1-2: Identify one high-frequency, rule-based workflow. For retail, replenishment of top-selling SKUs or automated reordering for a loyalty program. For B2B, MRO reorders, office supplies, or routine component purchasing. Define guardrails: spend caps, approved vendors, escalation triggers.
Weeks 3-6: Clean the data required for that workflow. Ensure product catalogues have structured metadata, pricing is API-accessible, and contract terms are machine-readable. Expose secure APIs. This is the least glamorous phase and the one that determines whether everything downstream works.
Weeks 7-10: Deploy in "assist mode." The agent drafts orders and recommends actions, but a human approves every transaction. Track cycle time, error rates, and exception frequency.
Weeks 11-13: Expand to limited autonomy under defined thresholds. Orders below a spend cap proceed automatically; everything else routes for approval. This is not a technology moonshot — it is a process improvement that happens to use AI agents as the execution layer.
Why AI Visibility Underpins Everything
There is a trap in thinking about agentic checkout and autonomous procurement as separate problems from AI search visibility. They are the same problem viewed from different angles. If AI agents cannot find and cite your brand, they will never navigate to your checkout page. If a procurement agent never encounters your supplier profile, it will never send a quote request to your API.
Discovery and transaction are collapsing into a single connected flow. This is why thinking in terms of context rather than funnels matters. The traditional funnel assumed discovery, consideration, and purchase were separate stages. In agentic commerce they are a single interaction: the AI discovers, evaluates, and purchases in one continuous process. If your brand is not present in the discovery layer, the checkout layer is irrelevant. For B2B suppliers, being invisible to an AI procurement agent managing a $10 million annual spend category is not a branding problem — it is a revenue problem.
Frequently Asked Questions
What is agentic commerce?
The use of AI agents to autonomously discover, evaluate, negotiate, and complete purchases — in both consumer retail and B2B procurement. In retail, a consumer's personal agent finds products and completes checkout without the buyer clicking a button. In B2B, agents execute procurement, supplier evaluation, negotiation, PO creation, and payment reconciliation — governed by procurement policies, budget limits, and compliance frameworks.
Is agentic commerce already happening?
Yes. Google launched agentic checkout across Search (AI Mode) and Gemini. Amazon's "Buy for Me" shops competitor websites. OpenAI's Operator completes multi-step purchases. Google's Project Mariner builds similar capabilities into Chrome. In B2B, buyer and seller agents are negotiating through API flows today, and Gartner projects AI agents will command $15 trillion in B2B purchases by 2028.
How does agentic checkout differ from a traditional shopping cart?
Traditional checkout requires a human to add items, review the cart, enter shipping, choose payment, and click "Place Order." Roughly 70% of carts are abandoned in that sequence. Agentic checkout lets an agent complete the purchase autonomously — it already has payment details, shipping address, and preferences. The cart does not disappear overnight, but it becomes optional for an entire class of transactions.
How does agentic replenishment differ from a subscription?
Subscriptions deliver on a fixed schedule regardless of consumption. Agentic replenishment adapts — the agent monitors real consumption and adjusts order timing, and optimises across retailers for each cycle. Loyalty is earned per transaction, not assumed per subscription.
What data do AI agents need from retailers and suppliers?
Structured, machine-readable product data (specifications, pricing, availability), accurate real-time inventory, machine-readable terms and policies, API-accessible catalogues, transparent pricing with no hidden fees, agent-friendly checkout infrastructure, and measurable trust signals (reviews, ratings, certifications). B2B suppliers additionally need compliance and audit capabilities — agents in regulated industries must log every decision and demonstrate policy compliance.
Do AI agents negotiate prices?
Yes. When contract terms are machine-readable and negotiation boundaries are explicit, buyer and seller agents negotiate autonomously through API flows. In retail, a consumer's agent negotiates within guardrails the consumer sets. In B2B, enterprises orchestrate hundreds of supplier negotiations simultaneously, with human intervention only for exceptions. Fixed product-page pricing becomes one input among many.
Which industries are adopting agentic commerce fastest?
In consumer retail, platforms with structured product feeds — grocery, fashion, electronics, beauty — are seeing the earliest agent traffic. In B2B, technology firms lead, services follow, and goods and manufacturing trail noticeably behind. That adoption gap creates a significant early-mover window for suppliers in industrial, manufacturing, and construction sectors willing to invest in agent-ready infrastructure now.
Businesses that treat agentic commerce as a future consideration are already behind. The infrastructure is live, the agents are shopping and procuring, and those not visible to them are losing transactions they will never know about. You can see a preview of how AI-ready your website is with a free AI scan — 30 seconds, no signup. For the complete picture, SwingIntel's AI Readiness Audit delivers expert research across 9 AI platforms.
