AI Agents: Revolutionizing Commerce with Trust and Context

Agentic AI: A New Era for Commerce

Imagine being able to tell a digital agent: “Use my points and book a family trip to Italy. Stay within budget, choose hotels we’ve liked before, and manage the details.” Unlike traditional assistants that merely provide a list of links, this agent assembles a complete itinerary and executes the purchase. This shift from assistance to execution is what distinguishes agentic AI. It also changes the speed at which commerce operates. Payment transactions are already clear in milliseconds, but the new acceleration concerns everything that precedes payment: discovery, comparison, decision-making, authorization, and tracking across multiple systems. As humans step away from routine decisions, “sufficient” data ceases to be acceptable. In an agent-driven economy, the constraint is not speed; it’s trust at the speed and scale of machines.

Data Management: The Key to Trust

Automated marketplaces already function because identity, authority, and accountability are embedded. As agents conduct transactions between businesses, this same clarity is required. Master Data Management (MDM) — the discipline of creating a single master record — becomes the exchange layer: tracking who an agent represents, what it can do, and where accountability lies when value moves. Marketplaces do not fail due to automation; they fail due to ambiguous ownership. MDM transforms autonomous action into legitimate, scalable trust.

The Agent: A New Participant

Digital commerce has long been built on two main sides: buyers and suppliers/merchants. Agentic commerce adds a third participant that must be treated as a primary entity: the agent acting on behalf of the buyer.

This seems straightforward until you pose the questions every business will face:

• Who is the individual, across channels and devices, with enough certainty for automation?
• Who is the agent, and what permissions and limits define what it can do?
• Who is the merchant or supplier, and are we sure we are designating the right one?
• Who holds responsibility if the agent acts with permission but against the user’s intent?

The practical risk is confusion. Humans, for example, can deduce that “Delta” means the airline when booking a flight, not the plumbing company. An agent needs deterministic signals. If the system gets it wrong, it breaks trust or forces a human confirmation step that contradicts the promise of speed.

Why “Sufficient” Data Fails at Machine Speed

Most organizations have learned to live with imperfect data. Duplicate customer records are tolerable. Incomplete product attributes are annoying. Merchant identities can be reconciled later.

Agentic workflows change this tolerance. When an agent takes actions without a human verifying the output, it needs nearly perfect data, as it cannot reliably notice when data is ambiguous or incorrect like a person can.

Failure modes are predictable, and they appear in the areas that matter most:

• Product Truth: If the catalog is inconsistent, an agent’s choices will seem arbitrary (“the wrong t-shirt,” “the wrong size,” “the wrong material”), and trust collapses quickly.

• Beneficiary Truth: Agentic commerce extends beyond cards to account-to-account experiences and open banking, broadening the universe of beneficiaries and the need to recognize them accurately in real time.

• Identity Truth: People operate in multiple contexts (work versus personal). Devices change. A system that cannot distinguish these contexts will either block legitimate activity or approve risky activity, harming adoption.

This is why unified enterprise data and entity resolution shift from a “nice to have” to an “operational requirement.” The more autonomy you desire, the more you must invest in modern data foundations that ensure its safety.

Contextual Intelligence: The Missing Layer

When leaders talk about agentic AI, they often focus on the capabilities of models: planning, tool usage, and reasoning. These elements are necessary, but they are not sufficient.

Agentic commerce also requires a layer that provides authoritative context in real time. Think of it as a real-time context system capable of responding instantly and coherently:

• Is this the right person?
• Is this the right agent, acting within the right permissions?
• Is this the right merchant or beneficiary?
• What constraints currently apply (budget, policy, risk, loyalty rules, preferred suppliers)?

Two design principles are important.

First, entity truth must be sufficiently deterministic for automation. Large language models are probabilistic by nature. This is useful for creating writing and drawing options. It is risky for deciding where money goes, especially in B2B and financial workflows, where “probably correct” is not acceptable.

Second, context must flow at the speed of interaction and remain portable across the entire value chain of the connected network. Mastercard's experience in optimizing payment flows is instructive: the more services you overlay on a transaction, the more you risk slowing it down. The evolving model pre-resolves, curates, and packages the signal so that execution is lightweight.

This is also where tokenization is headed. Initiatives like Agent Pay and Verifiable Intent from Mastercard signal a future where consumer identifiers, agent identities, permissions, and provable user intent are encoded as cryptographically secure artifacts — allowing merchants, issuers, and platforms to verify authorization and execution deterministically at machine speed.

What Leaders Should Do in the Next 12 to 24 Months

Adoption will not be uniform. Early adherence will often depend less on industry and more on the sophistication of an organization’s systems and data discipline.

This makes the next two years a window for practical preparation. Five actions stand out.

• Treat agents as regulated identities, not as features. Define how agents are onboarded, authenticated, authorized, monitored, and offboarded.

• Prioritize entity resolution where the cost of error is highest. Start with beneficiaries, suppliers, employee versus personal identity, and high-volume product categories.

• Build a reusable context service that every workflow and agent can call. Do not force every system to rebuild identity and relationships from scratch.

• Pre-calculate and compress signals. Resolve and curate context upstream so that real-time decision-making remains fast and predictable.

• Expand autonomy only as trust is earned. Establish a governance framework to handle disputes, keep humans involved for high-risk actions, measure accuracy, and expand automation as outcomes prove reliable.

A Tsunami Effect Across Industries

Agentic AI will not be limited to shopping carts. It will impact purchasing, travel, claims, customer service, and financial operations. It will compress decision cycles and eliminate manual steps, but only for organizations capable of providing agents with a clear identity, accurate entity truth, and reliable context.

Winners will treat entity truth and context as essential infrastructure for automation, not as a back-office cleanup project. In machine-speed commerce, trust is not a brand attribute; it is an architectural decision coded into identity, context, and control.