Playbook for Enterprises and Large Retailers
How large retailers and enterprises approach agentic commerce: build vs. partner, owned agents, operating model, governance, and B2B agentic buying.
In short: For large retailers, agentic commerce is both a new distribution channel and a threat to owning the customer relationship. The leading pattern as of October 2026 is a dual strategy: participate in third-party assistants through open protocols, while building an owned agent that carries loyalty, membership, and first-party data. On the buy side, procurement teams face the same shift as AI agents begin to mediate B2B purchasing.
What changes for enterprises
Enterprises have more to gain and more to lose than smaller merchants. Scale brings data, loyalty programs, and negotiating leverage with AI platforms. It also brings legacy order management, tax, and fulfillment systems that were not designed to respond to an agent in real time, and governance obligations that make “move fast” harder.
Early experience shows the integration burden is real. In March 2026 OpenAI ended native Instant Checkout in ChatGPT and shifted purchases to retailer sites and retailer apps inside ChatGPT; reporting at the time cited the difficulty of keeping stock status, sales tax, and pricing current through a centralized checkout. ACP continues as the discovery and integration protocol. Retailers including Target, DoorDash, and Instacart had already launched apps in ChatGPT. Walmart has integrated with Google Gemini so members can use their benefits and add items to their live Walmart cart, while continuing to develop its own assistant, Sparky. Access policy is also diverging: Shopify has made agent checkout a default for eligible merchants, while Amazon has been reported as blocking third-party agents such as Meta’s Muse from purchasing on its site.
Build vs. partner
Most enterprises will do some of each. The decision is per capability, not all-or-nothing.
| Capability | Partner when | Build when |
|---|---|---|
| Distribution in third-party assistants | Almost always; reach is controlled by the assistant | Rarely; negotiate merchant apps or native checkout terms |
| Catalog syndication and feeds | Platform or PSP tooling covers your channels | Catalog is large, complex, or highly regulated |
| Protocol endpoints (UCP, ACP) | Your commerce platform or PSP offers adapters | You run a custom stack and need control over checkout logic |
| Owned shopping agent | You lack AI and data teams | You have loyalty, membership, and first-party data to differentiate |
| Fraud and agent identity | Network and PSP programs meet your needs | You have unique risk exposure |
Platform and PSP offerings have converged on a “feed, cart, payments” pattern. Adyen Agentic (announced June 16, 2026, limited availability for US enterprises) explicitly packages an agentic feed, cart orchestration over existing checkout, tax, and order systems, and agentic payments. Stripe, PayPal, Shopify, and others offer comparable building blocks. See commerce platforms and market landscape.
Owned agents
An owned agent is a retailer-run assistant on the retailer’s own app or site, sometimes also distributed inside third-party assistants as a merchant app. Benefits include control of the experience, access to purchase history and membership data, and retention of the customer relationship. Costs include model and infrastructure spend, evaluation and safety work, and the risk of building something customers do not use.
Design questions to settle early:
- Scope. Discovery only, or cart and checkout? Reorders and replenishment are common early use cases because intent is clear.
- Distribution. Owned surfaces only, or also packaged as an app inside ChatGPT or Gemini.
- Interoperability. Exposing the agent through A2A or MCP lets other agents call it, which may matter as personal assistants begin to delegate to retailer agents.
Organizational and operational impacts
- Merchandising shifts from page placement to attribute completeness and feed quality. Category teams need product data ownership and SLAs.
- E-commerce engineering must expose real-time price, inventory, tax, and shipping calculations through APIs suitable for agents, with latency budgets measured in seconds.
- Marketing loses some paid search and display levers and gains new ones: retail media inside owned agents, promotions published through protocols, and agent optimization.
- Customer service must handle orders placed by agents, including disputes where the customer says the agent bought the wrong thing.
- Finance and tax must reconcile orders across platforms with different fee structures and merchant-of-record arrangements.
- Security must distinguish legitimate agents from scrapers; see Web Bot Auth and agent identity and fraud.
A small cross-functional team with a single accountable owner tends to move faster than distributing agentic work across existing channel teams.
Governance
Enterprises should treat agent channels with the same controls as any channel that takes customer orders and money.
- Policy: which assistants and agents you allow, under what terms, and what data you share with each.
- Consent and authorization: how you verify a purchase was authorized by the customer. Protocols such as AP2 produce signed records of user intent; network programs add tokenized credentials bound to agents.
- Liability: who bears the loss when an agent misorders. This is unsettled; see consumer protection and liability.
- Legal posture on third-party agents: the Amazon v. Perplexity litigation shows the stakes. A district court enjoined Perplexity’s Comet agent in March 2026; the Ninth Circuit vacated the injunction on August 4, 2026, holding that Amazon was unlikely to win its federal computer-fraud claim because customers, not Perplexity, were accessing Amazon’s systems. Contract and terms-of-service questions remain open.
- Model risk management: evaluation, monitoring, and incident response for owned agents, aligned with existing AI governance programs.
Procurement and B2B agentic buying
Enterprises are also buyers. Gartner predicted in late 2025 that AI agents will intermediate more than $15 trillion of B2B spending by 2028. Procurement suites are adding agents for intake, sourcing, and bid comparison, and catalog marketplaces connect into those suites.
For B2B sellers, this means:
- Publish machine-readable catalogs with contract pricing, tiered pricing, lead times, and compliance attributes.
- Support punchout and API ordering that agents can drive, with approval workflows preserved.
- Expect agents to compare suppliers on structured terms; vague quote-only pricing may be skipped.
For B2B buyers, this means:
- Set spending limits, approved supplier lists, and approval thresholds that agents must respect.
- Require audit trails linking each agent-initiated order to a human-approved policy or intent.
- Review supplier terms for agent-initiated orders and dispute handling.
Prioritized checklist
Now
- Appoint an accountable owner and cross-functional team for agentic channels.
- Inventory which assistants already surface your products and how.
- Make price, inventory, tax, and shipping available through low-latency APIs.
- Define an agent access policy for your sites and APIs.
Next
- Enable protocol endpoints (UCP, ACP) via platform or PSP, starting with one assistant.
- Pilot an owned agent on a narrow use case such as reorders.
- Extend fraud and dispute processes to agent-originated orders.
- Update supplier and procurement policies for agent purchasing.
Later
- Distribute your owned agent inside third-party assistants and expose it via A2A or MCP.
- Integrate loyalty and membership benefits into agent checkouts.
- Build retail media and promotions inventory inside owned agent experiences.
Common mistakes
- Treating agentic commerce as a single vendor decision rather than a portfolio.
- Building an owned agent before fixing product data and real-time APIs.
- Letting each business unit negotiate separately with AI platforms.
- Ignoring the buy side; procurement agents will reach your own suppliers and approval workflows.
Sources
- OpenAI shifts checkout plans in its agentic commerce strategy — Digital Commerce 360
- Walmart AI shopping strategy: tinkering to transformation — Retail Dive
- Shopify turns on agent checkout by default for a million merchants — Forkast
- Adyen Announces Adyen Agentic — Adyen
- Agentic commerce momentum — Shopify
- Universal Commerce Protocol — ucp.dev
- Ninth Circuit narrows CFAA reach in Perplexity agentic commerce ruling — PYMNTS
- Gartner: AI agents will command $15 trillion in B2B purchases by 2028 — Digital Commerce 360
- Announcing the Agent Payments Protocol (AP2) — Google Cloud