Rishi Katdare, Senior Leader in Networking and Edge for Global Financial Services at Amazon Web Services.
A customer authorizes an AI agent to buy from your business, but you decline the agent’s access. You still want the sale. What you may not accept is a transaction that bypasses the advertising and direct interaction on which your business also depends.
That conflict sits at the center of agentic commerce, where software compares offers and takes commercial action on someone else’s behalf. The buyer’s instruction establishes what the agent should accomplish. It does not settle the terms on which sellers will participate or how the agent’s operator will earn its return.
As customers delegate more of the buying decision, agent operators could gain influence over which businesses get considered. A company might continue serving the customer while becoming dependent on someone else for the opportunity to do so.
Delegating the comparison changes the decision.
Intermediaries have always influenced distribution. An AI agent extends that influence into judgment the customer previously exercised. Instead of examining results, the customer can ask the agent to resolve trade-offs and return with a recommendation or completed purchase.
An agent could recognize that flexible terms matter more than a small discount or ask a question that changes the buyer’s priorities. A business with a better offering could benefit from that richer comparison, particularly when its advantages are difficult to convey through a price ranking.
But the customer delegates precisely because conducting the comparison takes effort. They may be less likely to notice that an offer was excluded, a seller was inaccessible or a purchase could only be completed outside the agent’s service.
An operator that narrows the comparison also narrows the opportunity to compete. The excluded business cannot explain its value to a customer who never learns it was an option. Improving its product or lowering its price may accomplish little if neither change affects whether the agent considers it.
This is where convenience can alter the balance of power. The customer saves time by accepting the agent’s judgment, while the operator gains influence over which businesses receive that customer’s attention.
The seller has a reason to negotiate.
Sellers have legitimate reasons to set conditions. Advertising may support access to information, and direct interaction can encourage repeat business. They also need confidence that agents will accurately present their terms and use a supported transaction process.
Agreements addressing those concerns can make automated commerce more dependable. The balance in those negotiations will depend partly on which side the customer is more willing to replace.
A seller with a distinctive offering may persuade customers to complete purchases directly. An agent representing substantial demand may persuade sellers to accept its conditions instead. Neither position is permanent. As customers become accustomed to delegating decisions, a seller’s refusal to participate could increasingly mean exclusion rather than a visit to its own website.
Consider a business that initially joins an agent’s network to reach new buyers. Over time, existing customers also begin using that agent for purchases they previously made directly. Sales through the channel grow, but part of that growth reflects a change in route rather than new demand.
If participation fees then rise, withdrawing could put both groups of customers at risk. The company may find itself paying an intermediary to reach buyers whose loyalty it helped establish.
A channel can therefore look increasingly successful while the seller’s negotiating position weakens. That possibility deserves attention before growing volume makes dependence look like validation.
Whose interests determine the recommendation?
Agent operators need viable business models. Seller-funded services can reduce what customers pay directly, and a commission does not automatically make a recommendation unsuitable. A paid integration may support reliable inventory and better service when something goes wrong.
The conflict becomes clear when compensation overrides the buyer’s priorities. Suppose a customer wants the lowest total cost and accepts completing checkout elsewhere. The agent finds an offer that meets those requirements but recommends a more expensive partner because its operator earns a fee there. Even a flawlessly completed purchase would reflect a decision that failed the customer’s stated objective.
If that customer instead accepts a stated premium for automated completion, the partner could reasonably be the better choice. The distinction is whose preference determined the trade-off and whether the customer understood it.
That distinction connects trust to competition. Sellers should be able to improve their prospects by improving what customers value. If undisclosed commercial incentives override those priorities, sellers may face pressure to redirect investment from the offering toward the intermediary.
Over time, that could make access more valuable to the seller than differentiation. The buyer who expected a more discerning comparison could receive a narrower one.
Negotiate before dependence builds.
Technology and commercial leaders should evaluate an agent partnership together. The integration establishes that orders can flow. The agreement determines whether access remains worthwhile as the channel grows.
Separate eligibility from paid promotion. Establish whether declining preferred placement affects basic consideration, and seek visibility into whether offers are excluded or evaluated and rejected. Without that distinction, teams may cut prices to solve a distribution problem.
Track changes in the customer relationship alongside channel revenue. Where evidence permits, distinguish newly acquired business from purchases migrating out of direct channels. Review what remains possible if fees increase or the partnership ends. An exit clause offers limited protection if the company no longer has a practical route to those customers.
Operators should make the scope of their recommendations understandable when the buyer can still act. If a relevant offer requires an extra step outside the service, explain that trade-off rather than removing it. Customers should not have to repeat the entire search to discover what their agent left out.
Agentic commerce could make purchasing easier while changing who sets the terms of market participation. Sellers need alternatives before access becomes dependence. Buyers need confidence that delegated decisions still reflect their priorities.
An agent earns its place by making those priorities easier to act on, including when the customer’s best option is one its operator earns nothing from.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?