What if the biggest customer of your business was someone else's AI?

Online commerce already runs on extensive machine-readable information. Retailers expose product catalogs, inventory systems, delivery options and payment services to software. Price-comparison tools are familiar. The new development is the possibility of agents that can interpret a person's goal and coordinate several steps while remaining within authorized limits.

Google's 2025 Agent Payments Protocol announcement described a framework intended to support authorized transactions by software agents across participating organizations. Mastercard introduced Agent Pay and expanded work on transaction credentials and verification. In 2026 the company also announced work on machine-to-machine payments. These announcements indicate serious commercial investment, but standards and early products are not evidence that the entire buying experience is already dependable or widely adopted.

Three hard problems remain. First, how does a merchant know the agent truly has permission to buy? Second, how does a customer know the agent is acting for them rather than promoting a preferred supplier? Third, who is responsible when a purchase is wrong? The value of an automated buyer disappears quickly if it orders the wrong appliance, misses a return deadline or invents a product feature.

By 2031, imagine a household using an agent for selected routine purchases. The person sets a monthly budget, preferred retailers, quality standards and rules requiring approval for larger commitments. The agent replenishes an item only after checking whether it is genuinely needed, whether a substitute is suitable and whether the final transaction fits the authorized limits.

Now consider a small manufacturer selling replacement components to businesses. Customers' purchasing agents request specifications, delivery guarantees, environmental documentation and maintenance history. A company with accurate, structured information might win orders from buyers who never visit its marketing pages. A company that relies on vague claims and difficult ordering procedures could become invisible to these systems.

The potential benefit is less wasted time and greater competition based on measurable performance. The risk is market concentration. If a few platforms control which products agents can evaluate, those platforms could become powerful gatekeepers. Automated price negotiation might also favor organizations with the most sophisticated software rather than the best products.

By 2036, imagine a travel business, parts supplier or consumer brand treating automated buyers as a major customer category. Products are described in ways that both humans and machines can understand. Pricing systems distinguish public offers from private negotiated terms; authorization systems preserve evidence of what the customer approved; and service systems handle disputes arising from purchases initiated by agents.

Brands might still matter enormously. A person may instruct their agent never to buy from companies they don't trust. But trust could shift from memorable slogans toward product durability, transparent policies and dependable customer outcomes. The machine might become a ruthless filter for unsupported claims.

Another possibility is a new type of product design. If automated agents routinely compare lifetime repair costs, energy requirements and replacement parts, manufacturers may have stronger incentives to make those attributes visible and competitive. A store might distinguish itself by what its goods genuinely deliver rather than how cleverly it attracts clicks.

This doesn't mean human shopping disappears. People enjoy exploring, trying products and making emotional choices. A wedding dress, a special gift or a work of art is different from replenishing printer paper. The important transformation is that a significant portion of practical buying might migrate toward delegated decisions, and businesses will have to earn a place in that process.