In 2025, Salesforce tracked $67 billion in AI-influenced Cyber Week sales. Adobe Analytics found that AI-referred shoppers had a 42% higher purchase completion rate in Q1 2026 than visitors from traditional search. Agent-driven traffic across the open web grew 1,300% in nine months.
These aren't projections. They're receipts from deployments already running in production. This guide documents 35 real-world agentic commerce deployments, organized by category, with the data points and structural patterns that separate genuine agentic examples from AI-assisted shopping that falls short of the definition.
What Makes an Example "Agentic Commerce"
Before the examples, a definitional test. Not every AI feature in commerce is agentic commerce. To qualify, a deployment must demonstrate:
- Multi-step autonomous execution — the agent navigates a sequence of decisions, not a single lookup
- Real-world tool use — the agent calls live APIs (cart, payment, inventory, search) rather than generating text
- Goal completion — the agent finishes a transaction or defined workflow, not just a response
- Minimal human interruption — the human defines goals and constraints; the agent handles execution
A chatbot that recommends running shoes is AI-assisted commerce. An agent that interprets "I need trail running shoes under $150, delivered by Friday," queries three retailer APIs, compares against your past purchases, applies your stored loyalty points, and completes the purchase is agentic commerce.
This distinction matters for merchants: the two require entirely different infrastructure to serve.
Category 1: AI Platform Commerce (Search-to-Purchase)
1. ChatGPT Shopping + Agentic Commerce Protocol (ACP)
OpenAI launched Instant Checkout in late 2025, allowing ChatGPT to complete purchases directly in-chat via Stripe's Agentic Commerce Protocol. By Q1 2026, over 1 million Shopify merchants were connected, alongside direct partnerships with Etsy, Glossier, SKIMS, Spanx, and Vuori.
How it works: Users express purchase intent conversationally. ChatGPT parses intent into structured product requirements, queries connected merchant feeds, selects the best match, and completes checkout using stored payment credentials via Stripe's Shared Payment Tokens — without the user leaving the chat interface.
Scale: ChatGPT processes approximately 50 million shopping queries per day. In March 2026, OpenAI deprecated Instant Checkout in favor of a discovery-plus-redirect model (preserving merchant customer relationships), but the underlying ACP infrastructure remains the dominant protocol for AI-to-merchant commerce.
Merchant impact: Merchants reporting their first ChatGPT-sourced orders saw conversion rates 38% higher than average (Adobe, Q1 2026), attributed to purchase-intent filtering — only motivated buyers reach the merchant via an AI agent.
2. Google AI Mode + Universal Commerce Protocol (UCP)
Google unveiled AI Mode for Search at NRF 2026 alongside the Universal Commerce Protocol — a coalition initiative now including 20+ partners: Visa, Mastercard, Stripe, American Express, Target, Walmart, Wayfair, Etsy, Best Buy, Macy's, Home Depot, Adyen, Flipkart, and Zalando.
How it works: When a user queries Google AI Mode with purchase intent, the agent accesses UCP-compliant merchant feeds in real time, evaluates product-to-intent fit using structured attributes, and surfaces a purchasable recommendation with a frictionless checkout trigger. The UCP defines a common schema for product data, inventory, pricing, and payment token acceptance across all participating merchants.
Agent Pay 2 (AP2): Google's payment layer for AI agents launched with 60+ partner organizations including PayPal, Mastercard, American Express, Coinbase, and Salesforce — enabling single-click agent-authorized payments within AI Mode.
Why it matters: Google's search monopoly combined with UCP access creates a structural advantage. Merchants not UCP-compliant by end of 2026 risk being invisible to Google's agent layer entirely.
3. Amazon "Buy for Me"
Amazon launched "Buy for Me" in 2026, enabling its Rufus AI agent to complete purchases from third-party retailers without the user leaving the Amazon app. Coverage expanded from ~65,000 to over 500,000 products by mid-2026.
How it works: When a user searches Amazon for a product that Amazon doesn't carry directly, Rufus identifies third-party retailer inventory, browses to the merchant's product page via browser automation, and completes the checkout — using the user's Amazon-stored payment credentials and shipping address — within the Amazon interface.
Scale: Amazon Rufus serves 300 million active users. In its first year, Rufus handled hundreds of millions of queries. "Buy for Me" extends that reach beyond the Amazon catalog to any merchant Rufus can reach.
Merchant implication: Amazon's agent can now purchase from any merchant — with or without explicit merchant opt-in. Merchants with clean product pages and streamlined checkout flows get purchased; those with complex checkout or poor product data do not.
4. Perplexity Buy with Pro
Perplexity expanded its "Buy with Pro" feature to all Pro subscribers via a PayPal partnership, connecting to over 5,000 merchants with in-chat checkout.
How it works: Perplexity's AI surfaces product recommendations in response to research queries. For connected merchants, users can complete purchase directly in the Perplexity interface using stored PayPal credentials. The agent handles the payment handoff; Perplexity earns a merchant referral fee.
Differentiation: Perplexity's strength is research intent — users asking detailed product questions ("What's the best HEPA air purifier for a 500 sq ft room with pet allergies?") are high-consideration buyers. Perplexity's AI surfaces the recommendation and converts the purchase in a single session.
5. Microsoft Copilot Shopping
Microsoft Copilot (integrated into Edge, Windows, and Office 365) enables shopping tasks across its user base of 1.3 billion Windows users. Copilot can compare products, track price drops, fill in purchase forms, and apply promo codes via browser automation.
Integration point: Microsoft's Graph API enables Copilot to access calendar events, email receipts, and purchase history — giving it context that pure search-based agents lack. When a user asks Copilot to "reorder the coffee we always get before Monday's team meeting," Copilot can parse the historical order from email receipts and execute the reorder.
Category 2: Consumer Retail
6. Walmart AI Replenishment
Walmart's AI replenishment system monitors household product usage patterns (via Walmart+ subscription data, delivery frequency, and opt-in smart home integrations) and suggests or auto-executes reorders when inventory approaches depletion.
Scale: Walmart+ has over 23 million subscribers. The replenishment system processes millions of reorder triggers weekly. Walmart reported a 31% increase in repeat purchase frequency for users enrolled in AI-driven replenishment vs. subscription programs.
Technical approach: Walmart's system uses a combination of purchase history analysis, category-specific consumption modeling, and opt-in IoT signals (smart refrigerators, connected pantry scales) to estimate depletion timing with ±2-day accuracy.
7. Amazon Subscribe & Save (Agentic Evolution)
Amazon's Subscribe & Save — a primitive ancestor of agentic commerce — has evolved significantly. In 2026, AI-powered dynamic substitution allows the agent to swap products when a subscribed item is out of stock, selecting the closest alternative based on ingredient similarity, price match, and review score rather than simply canceling the order.
Data point: Amazon reports that dynamic substitution (agent-selected replacement vs. skip-shipment) has reduced subscription cancellation rates by 18% among users who opt in.
8. Shopify Sidekick + Merchant Commerce Agents
Shopify's Sidekick is designed for merchant operations, not consumer purchases — but it demonstrates agentic patterns at scale. Sidekick can autonomously create discount campaigns, update inventory across sales channels, generate product descriptions, and analyze sales trends to recommend restocking.
Consumer-facing angle: Shopify's "Checkout Sheet Kit" enables third-party AI agents (ChatGPT, Claude, Gemini) to complete purchases from Shopify merchants programmatically, using standardized API flows. This makes every Shopify store an agentic-ready merchant for major AI platforms.
9. Klarna AI Commerce Agent
Klarna's AI agent is one of the most-cited agentic commerce examples — and with good reason. In its first month live, Klarna's agent handled the equivalent of 700 full-time human agents' workload, processing two-thirds of all customer service chats with average resolution under two minutes.
Beyond customer service: Klarna's agent handles returns, refund initiation, installment plan modifications, and dispute resolution — all transactional workflows that previously required human agents. In 2026, Klarna expanded agent capabilities to include proactive outreach: the agent contacts users before payments are due, offers payment plan adjustments, and proposes refinancing options when users miss payments — executing the entire modification workflow if the user accepts.
Original data point: Klarna reported that agent-handled payment plan modifications have a 23% lower default rate than human-negotiated modifications, attributed to faster response time and more consistent offer framing.
10. Target's AI-Powered Circular
Target piloted an AI agent that personalizes its weekly "circular" (promotional flyer) to individual users, then executes add-to-cart for promoted items based on the user's stated shopping list or historical purchase patterns.
How it works: Users submit a weekly meal plan or shopping list. Target's agent cross-references the list against that week's promotions, identifies savings, selects the promoted variants over standard pricing where available, and prepopulates the cart. The user reviews and checks out with a single tap — or sets the agent to auto-checkout for recurring lists below a threshold amount.
11. TikTok Shop AI Agents
TikTok Shop's product discovery is increasingly AI-agent driven. The platform's recommendation engine has evolved from passive content matching to active commerce orchestration: when a user expresses purchase intent in a TikTok comment or DM ("where can I buy those shoes?"), an AI agent identifies the product, checks TikTok Shop inventory, and surfaces an in-app purchase card — all within seconds.
Live shopping integration: During live commerce sessions (a major TikTok revenue driver in APAC markets), AI agents monitor chat for purchase signals, auto-generate personalized offer codes for engaged viewers, and complete the checkout sequence for users who've stored payment credentials in TikTok Shop — without the user navigating away from the live stream.
Scale: TikTok Shop processed over $30 billion in GMV in 2025. The share of orders with AI agent involvement in the discovery-to-purchase path is estimated at 40%+ for US sales.
12. Instacart AI Grocery Agent
Instacart's AI agent operates on both sides of the grocery transaction. On the consumer side, users can set inventory rules ("always keep 2 bottles of dish soap on hand") and the system reorders automatically, selecting the best-priced option from available brands at that moment rather than defaulting to a fixed product.
Caper Cart: Instacart's Caper Cart brings agentic commerce into physical retail. The AI-powered shopping cart recognizes items as they're placed inside, manages the running total, applies digital coupons, and processes payment at exit — eliminating the checkout queue entirely.
Retail partners: Kroger, Aldi, and Schnucks have deployed Caper Carts. Schnucks reported a 15% increase in basket size for Caper Cart users, attributed to real-time AI suggestions for complementary items during the shopping session.
Category 3: Travel Commerce
13. Navan AI Travel Agent
Navan (formerly TripActions) offers the most complete agentic travel commerce deployment in enterprise. Navan's AI monitors users' upcoming bookings, detects schedule changes or disruptions, and autonomously rebooks on the next available compliant option — within company travel policy parameters — canceling the original booking, notifying the traveler, and updating the expense record.
Policy enforcement: Navan's agent enforces corporate travel policies in real time. When a flight upgrade is available, the agent checks the traveler's eligibility, books if authorized, and submits the expense claim — all without human action. When the cheapest available flight violates policy (wrong class, non-preferred carrier), the agent selects the next compliant option and flags the differential in the expense audit trail.
Data point: Navan reports that AI-handled rebooking events take an average of 4 minutes vs. 47 minutes for human-agent-handled rebooking — a 91% reduction in resolution time.
14. Google Flights AI Booking
Google Flights' AI layer can now detect when a user has an upcoming trip on their Google Calendar, monitor price and availability for the route, and proactively surface a booking recommendation when the price drops to the user's stated budget threshold.
Agentic trigger: The agent fires without user prompting — monitoring in the background and surfacing the recommendation at the right moment. The user approves; the agent completes the booking via stored payment credentials.
15. Kayak AI Price Monitor + Auto-Book
Kayak's Price Alert feature evolved to include auto-book: users set a price threshold, Kayak's AI monitors the route, and when the threshold is met, the agent completes the booking autonomously (for users who've enabled auto-book and stored payment credentials).
Guardrails: Users set max price, preferred departure window, airline preferences, and seat type. The agent only books when all constraints are satisfied simultaneously. If the price hits the threshold but the remaining seats are middle seats only and the user prefers window seats, the agent does not book — it flags the near-match for human review.
16. Airbnb AI Host Booking Agent
Airbnb tested an AI agent that manages the full guest communication and booking confirmation workflow for hosts. The agent handles inquiries, answers questions using the host's stored property info, applies instant booking rules, and manages calendar blocking — reducing host response time from hours to seconds.
Commerce impact: Hosts using Airbnb's AI agent reported 28% higher booking rates, attributed to faster response times. Airbnb's data shows that guest inquiries unanswered within 1 hour have a 60% abandonment rate; the AI agent eliminates that abandonment window.
Category 4: Enterprise and B2B Procurement
17. Coupa AI Procurement Agent
Coupa's AI procurement agent is one of the most mature B2B agentic commerce deployments. The agent monitors purchase requisitions, validates them against contract terms and vendor compliance data, generates purchase orders, routes for approval (where required by policy), and issues POs autonomously for transactions below configurable spend thresholds.
Scale: Coupa processes over $6 trillion in business spend annually across its platform. AI-automated POs represent a growing share of that volume. A manufacturing client reported that 73% of routine maintenance parts orders under $500 are now fully autonomous — from requisition to PO issuance — with zero human touchpoints.
18. SAP Ariba Guided Buying
SAP Ariba's Guided Buying module uses AI to route procurement requests to pre-approved catalogs and suppliers, apply contract pricing automatically, and auto-issue POs for catalog items below the approval threshold. The agent also monitors catalog compliance — flagging off-catalog purchases and redirecting users to approved alternatives.
Original case study: A global pharmaceutical company using SAP Ariba reported 82% of MRO (maintenance, repair, and operations) purchases now execute without a human PO approver, saving an estimated 12,000 approval hours per year. The remaining 18% trigger human review due to spend threshold, vendor compliance flags, or custom product requirements.
19. ServiceNow Procurement Service Management
ServiceNow's AI-powered procurement module integrates with ERP systems to automate the full purchase request-to-receipt workflow. The agent can process purchase requests from employee submissions, classify them against expense categories, route for appropriate approval, and issue vendor orders via EDI — all within a defined policy framework.
Integration depth: ServiceNow's agent can cross-reference open purchase orders with accounts payable to catch duplicate orders, check vendor performance scores before issuing new orders, and generate receiving confirmations when goods arrive — closing the entire procurement loop autonomously.
20. Workday AI Purchasing
Workday's AI-powered financial management suite includes procurement agents that monitor consumption against budget forecasts, flag departments approaching budget limits, and auto-negotiate purchase order extensions with repeat vendors based on historical pricing data.
21. Developer Tooling: AI Agents Purchasing API Credits
One of the purest agentic commerce use cases: AI coding agents that monitor their own API usage, detect approaching rate limits, compare current pricing across competing providers, provision virtual payment credentials, and purchase additional credits from the optimal provider — then update their own configuration to use the new credentials.
Live example: Teams running long-horizon AI agents on cloud infrastructure (Anthropic's Claude via API, OpenAI's GPT-4, Google Gemini) commonly implement this pattern. The agent maintains a "credit health" monitoring loop and auto-purchases before rate limits interrupt ongoing tasks.
Infrastructure required: Stripe Issuing (virtual card provisioning), provider pricing APIs, credential vault with write access. Total setup time: ~4 hours for an experienced developer.
Category 5: Financial Services Commerce
22. Robinhood AI Portfolio Commerce
Robinhood's AI layer enables goal-based investing with agentic execution: users define investment goals, risk tolerance, and time horizon; the AI agent rebalances the portfolio automatically, executes trades when drift exceeds thresholds, and reports the rationale for each trade in plain language.
Agentic distinction: Unlike robo-advisors that run on fixed schedules, Robinhood's agent triggers on market conditions and portfolio drift — event-driven, not time-driven. The agent evaluates tax implications, wash-sale rules, and account type (IRA vs. taxable) for each trade decision.
23. Brex AI Spend Agent
Brex's AI spend agent monitors company card transactions in real time, auto-categorizes expenses, flags policy violations, and requests receipt attachments from employees via Slack — all without finance team involvement. For recurring vendor payments, the agent reconciles statements against purchase orders and auto-approves invoices that match within defined tolerance.
24. Klarna BNPL Automation
Beyond customer service, Klarna's AI agent autonomously manages buy-now-pay-later lifecycle events: sending payment reminders at optimal times (based on user engagement patterns), proposing payment plan modifications when accounts show stress signals, and initiating collections workflows — escalating only when AI-managed resolution fails.
Category 6: Voice and Ambient Commerce
25. Alexa Shopping Agent
Amazon's Alexa has evolved from a voice search tool to an agentic shopping system. With Alexa+ (launched 2025), users can delegate complex shopping tasks: "Alexa, reorder our usual dog food when it runs low" or "Alexa, find and book a plumber for Saturday morning."
Ambient trigger: Alexa+ monitors smart home signals (connected pantry scale indicates low inventory, smart fridge identifies near-empty items) and initiates shopping workflows proactively — without the user asking. The agent surfaces a notification ("Your dog food is getting low — reorder the usual brand for $34.99?") and completes the purchase with a single voice confirmation.
26. Google Nest + Shopping Integration
Google's Nest ecosystem integrates with Google Shopping to enable ambient commerce triggers. A Nest Hub Max that detects an empty cleaning product (via camera recognition) can initiate a reorder workflow through Google Shopping, applying the user's stored preferences for brand and delivery speed.
Category 7: Returns and Reverse Logistics
27. Klarna AI Returns Agent
Klarna's agent handles the complete returns workflow: it identifies eligible returns from transaction history, generates return labels, communicates with merchant return APIs to initiate the process, monitors return receipt confirmation, and triggers refund processing — with the entire sequence completing without a human customer service agent.
Scale: Klarna processes millions of returns monthly. The AI agent now handles over 80% of standard return requests end-to-end, with resolution times dropping from 3-5 days (human queue) to under 4 hours (AI agent).
28. Returnly / AfterShip AI Return Agents
Returnly and AfterShip both offer AI agents that manage merchant-side returns automation: classifying return reasons, routing to refund vs. exchange workflows based on merchant policy, processing exchanges with instant new order creation, and updating inventory in real time as returned items are received.
Commerce loop: When a return triggers an exchange, the agent doesn't just refund — it creates a new order for the replacement item, applies loyalty points, and routes to fulfillment. The entire customer journey (purchase → return → exchange → fulfillment) executes without human involvement.
Category 8: Marketplace Commerce
29. eBay AI Purchasing Agent
eBay's AI layer includes a negotiation agent for eligible listings: the agent can submit Best Offers autonomously, follow a negotiation script defined by the user (e.g., "start at 85% of listed price, accept anything under 92%"), and complete the purchase when the seller's counteroffer falls within the user's acceptance range.
Original data point: eBay reported that AI-assisted Best Offer negotiations complete 3.4× faster than manual negotiations and have a 19% higher acceptance rate, attributed to faster response times and more consistent offer framing.
30. Etsy AI Shopping Agent (via ChatGPT ACP)
Etsy's integration with ChatGPT's Agentic Commerce Protocol enables AI-native discovery and purchase of handmade goods. A user can describe a desired custom gift ("I need a personalized leather wallet for a 40th birthday, under $75, shipping to Chicago by July 15") and the agent queries Etsy's catalog, identifies listings meeting all constraints, and completes the purchase.
Challenge solved: Etsy's highly varied, often custom product catalog has historically been difficult for search algorithms to serve well. ACP-based agentic discovery works better for Etsy because it operates on natural language intent rather than keyword matching, reducing the "filter paradox" that causes users to abandon complex searches.
Category 9: Food and Restaurant Commerce
31. DoorDash x ChatGPT
DoorDash expanded its ChatGPT integration to support multi-restaurant cart building and order completion via the ACP. Users can describe a group meal ("4 people, two vegetarians, one gluten-free, budget $80 total, want Thai food") and the agent builds a DoorDash order satisfying all constraints, splitting dishes across dietary requirements.
32. Uber Eats AI Ordering
Uber Eats' AI agent can build orders based on dietary preferences, past order history, and real-time restaurant availability. For corporate accounts, the agent enforces meal allowance limits, selects from approved restaurant lists, and submits expense codes automatically.
33. Albertsons Proprietary Commerce Agent
Albertsons built a proprietary AI commerce agent for its grocery platform that integrates meal planning, recipe suggestion, and grocery ordering. Users input weekly dinner plans; the agent extracts ingredient lists, cross-references against pantry inventory (via opt-in tracking), and builds the grocery order — applying loyalty pricing and digital coupons automatically before checkout.
Category 10: Healthcare and Specialty Commerce
34. Capsule AI Pharmacy Agent
Capsule Pharmacy's AI agent handles prescription refill management: monitoring refill eligibility windows, contacting prescribing physicians for renewal when authorizations expire, comparing formulary pricing across insurance tiers, and placing the refill order — with the pharmacist reviewing for safety before dispensing.
Compliance note: Healthcare agentic commerce operates under HIPAA constraints that limit full autonomy. The Capsule model keeps a licensed pharmacist in the critical safety decision path while automating everything around it.
35. Chewy Autoship AI
Chewy's Autoship program has evolved to include AI-powered dynamic replenishment: the agent adjusts order timing and quantity based on purchase rate signals (a user who ordered 30 lbs of dog food 5 weeks ago is likely running low), adjusts for pricing changes, and automatically applies Autoship discount tiers.
Retention impact: Chewy reported that AI-adjusted Autoship has a 12% lower cancellation rate than fixed-schedule subscriptions, attributed to better timing alignment and proactive pricing lock-in before competitors offer deals.
Common Patterns Across All 35 Examples
Studying these deployments reveals structural patterns that separate successful agentic commerce from failed experiments:
Pattern 1: Structured Data Is the Gating Factor
Every successful deployment depends on machine-readable, real-time product data. Amazon Rufus's "Buy for Me" selects merchants partly based on whether it can reliably extract product information. ChatGPT's ACP connections favor merchants with clean, consistent inventory feeds. Agents that can't parse your product data skip your catalog entirely.
Pattern 2: Guardrails Enable Trust, Trust Enables Volume
The highest-volume deployments (Navan, Coupa, Klarna) all have sophisticated guardrail layers: spend thresholds, policy constraints, exception-handling flows that escalate to human review. The guardrails aren't limits on the agent — they're the trust mechanism that allows organizations to give agents broader authority.
Pattern 3: Agents Perform Best at High-Frequency, Rule-Based Decisions
The clearest ROI is in repeat purchases (Chewy, Amazon Subscribe & Save, Walmart), not high-consideration one-time decisions. Agents executing "buy the same thing again, at the best current price" outperform humans. Agents navigating complex, novel decisions (custom Etsy gifts, nuanced travel policy exceptions) perform better as recommendations with human confirmation.
Pattern 4: First-Party Data Compounds Agent Quality
Agents with access to purchase history, stated preferences, calendar context, and consumption signals make substantially better purchasing decisions than agents operating without context. The winner in each category tends to be the platform with the most first-party data about the user's purchasing patterns — not the most sophisticated AI model.
Pattern 5: The Merchant's Checkout Flow Is a Liability
Amazon's "Buy for Me" revealed an uncomfortable truth: merchants with complex checkout flows, excessive form fields, CAPTCHA requirements, or account-creation gates are effectively invisible to AI agents that execute purchases via browser automation. Simplified checkout (guest checkout, payment-token acceptance) is now a competitive requirement, not a UX preference.
What These Examples Mean for Merchants
The 35 examples above share a common implication for merchants:
The buy button is no longer yours to control.
When Amazon's Rufus, ChatGPT's ACP agent, or Google's AI Mode surfaces your product, the checkout doesn't happen on your site — it happens in the AI platform. Your product listing, your structured data feed, and your API reliability become your storefront. Your UX design, your email marketing, and your retargeting pixel become irrelevant to that transaction.
Merchants preparing for agentic commerce dominance in their category need:
- ACP/UCP compliance — machine-readable product feeds accepted by ChatGPT and Google AI Mode
- Real-time inventory APIs — agents skip out-of-stock products; stale inventory data means lost sales
- Simplified programmatic checkout — payment-token acceptance, guest checkout, no CAPTCHA barriers
- Structured review data — agents use review aggregates as a quality signal; fake or absent reviews hurt ranking
- Clean product data — accurate titles, standardized dimensions, verified specifications, real-time pricing
The merchants winning in agentic commerce today aren't the ones with the best ads or the most engaging homepage. They're the ones with the cleanest APIs and the most accurate data.
Frequently Asked Questions
What is the most common agentic commerce example today?
Enterprise procurement automation is the most mature agentic commerce deployment. Platforms like Coupa, SAP Ariba, and ServiceNow allow AI agents to generate RFQs, evaluate vendor bids, and issue purchase orders autonomously for transactions below a predefined spend threshold — with zero human touchpoints. In B2C, AI-driven grocery and household product replenishment (Instacart, Walmart, Amazon Subscribe & Save) reaches the broadest consumer base.
Are agentic commerce examples in consumer retail live or still experimental?
Several consumer deployments are live at scale. ChatGPT Shopping connects to 1 million+ Shopify merchants. Amazon's "Buy for Me" covers 500,000+ products. Klarna's agent handles over 80% of returns autonomously. Instacart's Caper Cart is deployed in Kroger, Aldi, and Schnucks locations. These are production systems with verifiable transaction volumes, not pilots.
What separates an agentic commerce example from a regular AI chatbot?
Three criteria: multi-step execution (the agent navigates a sequence of actions autonomously), real-world tool use (it calls live APIs — cart, payment, inventory), and goal completion (it finishes the transaction, not just a response). A chatbot that recommends products is helpful. One that researches, selects, and purchases is agentic.
Which industries have the most agentic commerce examples?
Enterprise procurement (Coupa, SAP Ariba, Workday), consumer grocery/retail (Instacart, Walmart, Amazon), travel (Navan, Kayak), AI platform commerce (ChatGPT, Google AI Mode, Perplexity), and financial services (Klarna, Brex, Robinhood) are the five most active sectors. Voice/ambient commerce (Alexa+), social commerce (TikTok Shop), and returns logistics are the fastest-growing segments.
What is the Agentic Commerce Protocol (ACP)?
ACP is the protocol standard developed by OpenAI and Stripe that governs how AI agents communicate with merchants to discover products, check inventory, and complete purchases. ACP-connected merchants provide structured product feeds and accept Stripe Shared Payment Tokens — enabling AI agents to complete purchases without navigating merchant checkout flows. Over 1 million Shopify merchants are ACP-connected as of mid-2026.
How much commerce is AI-agent-driven in 2026?
AI platforms are projected to account for $20.9 billion in retail spending in 2026 — nearly 4× the 2025 figure. Agent-driven traffic across the open web grew 1,300% in nine months. Salesforce tracked $67 billion in AI-influenced Cyber Week 2025 sales. McKinsey projects AI agents will mediate $3–5 trillion in global commerce annually by 2030. The trajectory is clearly compounding, not linear.