❝

“An agent making a purchase should strengthen, not bypass, the trust built into payments.”

Sherri Haymond, EVP, Global Head of Digital Commercialization, Mastercard

Redeployed is a weekly newsletter that breaks down one important AI story at a time for product builders and engineering leaders. Every issue explores what the shift means for technology companies and how leaders can respond.

OpenAI just connected several pieces required to turn ChatGPT into a commercial channel.

Businesses can now claim verified profiles, submit current product information, publish conversational agents, pay for distribution, and inspect how they appear in ChatGPT. Consumers can research purchases through AI and, through the initial rollout of ChatGPT Wallet in dots, complete transactions without exposing their actual card details to the agent or merchant.

These capabilities arrived through two announcements last week: Businesses in ChatGPT on November 10 and ChatGPT Wallet on November 13. Together, they establish the beginnings of a commercial stack inside ChatGPT, covering business representation, discovery, conversation, sponsored placement, and payment execution.

The initial rollout remains limited. Business Agents initially target eligible US ecommerce and online service companies, while Wallet is beginning with adult ChatGPT Pro subscribers in the US using eligible US-issued cards. In the initial Wallet experience, users review purchase details, select a card, and authorize payment through a security check. OpenAI's terms also contemplate advance authorization, making the boundaries of delegated purchasing worth watching as the product evolves.

Even at this early stage, the product implication is significant. A company's customer-facing experience is no longer confined to its website or app. The information and actions exposed to agents now need their own product decisions, reliability standards, and accountable owner.

The Catalog Is Becoming an Interface

A conventional ecommerce catalog feeds product pages, search results, advertising systems, and marketplaces. An agent-readable catalog has a different responsibility: giving a model enough structured, current information to compare an offering, explain its constraints, and determine whether it fits a customer's request.

Through Business Hub, verified companies can supply prices, availability, descriptions, delivery policies, and return policies. Shopify merchants can connect their stores directly, while OpenAI provides aggregated metrics for mentions, website clicks, and the topics through which customers discover a business.

This creates a new form of distribution work. Product teams will need to treat feed freshness, attribute completeness, and policy consistency as customer experience metrics. A stale availability field is no longer just a merchandising problem. It could cause an agent to exclude a product before a customer ever sees it.

There is an important distinction between being represented in ChatGPT and controlling how ChatGPT recommends products. Claiming a profile, publishing a branded Business Agent, advertising, and appearing in an independent ChatGPT answer are separate activities. Submitting a feed does not guarantee that a product will be recommended.

That separation creates a new platform dependency. Businesses can manage their own information and branded experiences, but they do not control the independent recommendations or broader discovery environment through which customers may encounter them.

As Tecla's guide to agentic AI in ecommerce and retail explains, production commerce systems need structured data and human approval boundaries around consequential actions. Those requirements become even more important when the customer interacts through an external agent.

Checkout Compatibility Becomes an Engineering Concern

ChatGPT Wallet is designed to work with existing payment infrastructure. After a user authorizes a purchase, the payment network creates a token containing substitute card credentials. The agent and merchant do not receive the customer's actual card details, while the merchant remains responsible for pricing, fulfillment, returns, and customer service.

But compatibility does not make the integration automatic.

Businesses may need to configure their CDNs and firewalls to recognize signed requests from agents. Payment and fraud systems must distinguish legitimate agent transactions from automated abuse. Merchants may also expose actions such as checking availability, managing carts, or retrieving order information through plugins and WebMCP tools.

This is where agentic commerce becomes an engineering and operations problem rather than simply a new marketing channel.

A successful demo proves that an agent can place an order. It does not prove that the system can recover from an inventory conflict, explain a cancellation, avoid a false fraud decline, or preserve margin after returns and support costs.

The reliability problem behind production agents becomes especially concrete when the agent can spend money. Reliability must extend across the entire transaction, including identity, checkout, inventory, fulfillment, and the external providers the model does not control.

The Metric Is a Healthy Transaction

The first dashboard most companies build will probably count ChatGPT mentions and referral clicks. Those are useful discovery signals, but they reveal little about whether the channel is generating profitable transactions.

Teams should measure the full journey: whether an agent finds accurate information, reaches checkout, obtains customer approval, completes the transaction, and handles exceptions without unnecessary human intervention.

Contribution margin, cancellations, returns, support contacts, and repeat purchases will matter more than agent conversations alone. The same caution applies to sponsored Business Agents. Strong engagement does not necessarily establish profitable incremental demand, a challenge Tecla has also explored in its guide to agentic AI in digital marketing.

This also changes how companies assign responsibility. Growth may manage discovery and paid distribution, while product owns the Business Agent experience. Engineering maintains integrations, payments teams oversee transaction controls, and operations handles problems after checkout.

Each function can report healthy metrics while the customer experiences a broken transaction. Someone needs to own the complete agent-facing customer journey, not just one part of the funnel.

This issue of Redeployed is brought to you by Tecla Labs: As companies prepare their products for agent-mediated discovery and transactions, they need systems that expose accurate data, integrate with existing commerce infrastructure, and operate reliably across every handoff. Tecla Labs helps companies identify high-impact AI opportunities, develop custom AI solutions, and integrate intelligent automation into their operations, turning agent capabilities into practical, production-ready systems.

Build the Second Surface Deliberately

OpenAI reports that ChatGPT has 1.2 billion weekly users. Whether agent-mediated purchasing becomes a major commercial channel will still depend on customer behavior, merchant compatibility, economics, and trust.

DoorDash's disclosure earlier this month offered a useful reminder. Third-party agents could technically place orders, yet the company had received only several hundred such orders. That limited activity does not establish a lack of consumer demand, but it shows why technical availability should not be confused with adoption.

OpenAI is building more of the infrastructure required to make agent-mediated commerce practical. Businesses should prepare for that possibility without assuming the channel will replace their existing customer experience.

The strategic risk extends beyond transaction volume. A merchant may still own fulfillment, returns, and customer support even as an agent platform gains more influence over which products customers discover and choose. Companies could retain the operational responsibility for a sale while losing control over more of the journey that produced it.

That makes controlled experimentation valuable now. Businesses can improve the accuracy of their product information, define which actions agents may perform, establish approval boundaries, and measure transaction economics without rebuilding their entire commercial strategy around an unproven channel.

I would assign one product leader responsibility for the agent-facing customer journey, even while transaction volume remains small. Their job would be to make the company discoverable and operable through agents while preserving visibility into customer relationships, service quality, and profitability.

The ownership question for this week is simple: if an agent tries to discover, evaluate, and buy from your company today, who is responsible for the whole result?

Connect With Other Technology Leaders

If you want to exchange practical ideas with senior technology and product leaders navigating the AI era, check out the upcoming GILD Forums. They bring together experienced operators for peer discussions around the technology and business decisions they are working through right now.

More to come…

– Gino Ferrand, Founder @ Tecla Labs