Strategy, Experience, Technology
Intent-Based Commerce: How It Works and Where It Will Be Used
The interface may look like one sentence. The trust will come from everything it refuses to assume.
Mert Barbaros ·
Commerce still makes the customer do a surprising amount of clerical work.
Imagine six friends are coming for dinner. Two are vegetarian. One cannot eat nuts. The food should arrive by 7:30 PM, the budget is $120, and nobody wants another pile of side dishes pretending to be a meal.
Today, the customer converts that dinner into keywords, compares menus, checks allergens, estimates quantities, watches delivery times, and rebuilds the basket when something disappears. The software displays products. The person does the planning.
A search query is not an intent.
Search commerce begins with nouns: running shoes, hotel, coffee machine, airport transfer. Intent begins with a result: shoes for a first half-marathon, a quiet place near tomorrow's meeting, equipment that fits a small kitchen, or a ride that still works if a flight is late.
What an actionable intent records
The outcome
What the customer wants to achieve.
Hard limits
Budget, timing, allergies, accessibility, and other non-negotiables.
Trade-offs and alternatives
Preferences that can change and acceptable substitutes when the first choice fails.
Delegation and approval
What the system may do independently and what still needs a human decision.
Search and intent begin with different questions
A comparison of the starting point, output, customer role, merchant role, and success measure in search commerce and intent-based commerce.
| Product decision | Search commerce | Intent-based commerce |
|---|---|---|
| Starting point | Product or category | Desired outcome and context |
| Main output | Ranked options | A journey shaped around intent |
| Customer role | Search, compare, coordinate, checkout | Clarify boundaries and approve important decisions |
| Merchant role | Publish products and pages | Expose accurate product, checkout, and fulfillment capabilities |
| Success | Relevant result or completed payment | Purchase remains consistent with expressed intent |
Interest in research is much higher than comfort with payment
The immediate job is to help people decide, not to assume they want an agent to transact.
A horizontal bar chart showing 73 percent interest in using AI to research or compare products and 24 percent comfort with AI completing a purchase.
| Activity | Respondents |
|---|---|
| Research or compare products | 73 |
| Comfortable completing a purchase | 24 |
The public infrastructure is starting to exist.
AI could recommend products long before 2025. The newer development is that commerce and payment companies are giving agents structured ways to interact with merchants. Universal Commerce Protocol, Agentic Commerce Protocol, and Agent Payments Protocol define building blocks for discovery, checkout, authorization, payment, and fulfillment without requiring one company to own the entire journey.
A commerce capability flow
Human request
Capture the outcome and boundaries.
Agent interface
Interpret ambiguity and clarify risk-sensitive details.
Merchant discovery
Read current capabilities, inventory, rules, and fulfillment promises.
Checkout and fulfillment
Present the cart for consent; keep the merchant responsible for acceptance, delivery, returns, and support.
The useful cases begin where the path is harder than the outcome.
A Buy again button is faster than a long conversation when the customer knows the exact item. Intent-based commerce becomes valuable when the desired result is clear but achieving it requires comparison, coordination, or several constraints.
The outcome is clearer than the path
Five applications where intent-based commerce can reduce coordination work.
| Application | Example intent | Information that matters |
|---|---|---|
| Grocery | Feed six people under dietary and budget limits | Products, substitutions, inventory, delivery |
| Travel | Arrange a short trip under one total budget | Timing, location, cancellation, availability |
| Local services | Find a suitable service before a deadline | Service scope, location, availability, price |
| Procurement | Source a compliant input before production | Specification, supplier policy, delivery, approvals |
| Recurring commerce | Rebuy within delegated limits | Trigger, price ceiling, expiry, confirmation rule |
Trust is the product boundary.
The attraction is obvious: people can describe a result and avoid the administrative work between wanting and buying. The risk is equally clear. Human intent is incomplete, merchant information becomes stale, ranking systems have incentives, and a valid payment does not prove that a purchase was wise.
Security is the first stated barrier to agent-led shopping
Authority, data scope, and correction paths must be understandable before a system asks for trust.
A bar chart of consumer concerns: data security 50 percent, privacy 44 percent, purchase accuracy 42 percent, reliability 40 percent, and control 36 percent.
| Concern | Respondents |
|---|---|
| Data security | 50 |
| Privacy | 44 |
| Purchase accuracy | 42 |
| Reliability | 40 |
| Control | 36 |
Five boundaries deserve explicit design
Show the interpretation
Make the understood outcome and important limits visible before action.
Match approval to risk
Higher cost, lower reversibility, and more uncertainty should produce stronger confirmation.
Maintain merchant truth
False inventory, outdated pricing, and unavailable capacity cannot be corrected by better reasoning.
Keep data sharing narrow
Merchants need transaction-relevant information, not a customer's complete conversation.
Make responsibility visible
Customers should know who discovered, accepted, fulfilled, and supports the order.
For merchants, discoverability becomes operational.
An agent needs structured, current answers: what is available now, which conditions apply, what can be changed through the integration, which payment methods are supported, and who handles fulfillment, returns, and support.
Intent will not replace search.
Search remains useful when people want to explore, learn, compare, and keep control of the journey. Intent-based commerce is better suited to moments when a person knows the outcome but does not want to become the temporary project manager required to achieve it.
Sources and research notes.
Bain consumer research · Visa trust research · Universal Commerce Protocol · Agent Payments Protocol · Agentic Commerce Protocol