How Brands Can Adapt When AI Agents Do the Shopping

HBR arrives at what the AXD Institute has argued since its founding: that agentic commerce demands a trust layer, not merely a technology layer. The article identifies five failure modes - agent misunderstanding, authority overreach, data liability, brand misrepresentation, and recovery failure - that map directly onto the AXD practice frameworks of delegation design, consent architecture, agent observability, and failure architecture. The authors' call for 'generative engine optimisation' (GEO) over traditional SEO confirms the shift from persuasion to performance that defines the machine customer era. Most significantly, the article draws the analogy to early e-commerce trust infrastructure (SSL, PCI standards), reinforcing the AXD position that trust is not a brand attribute but structural material.

What does HBR say about the future of brand strategy in agentic commerce?

Harvard Business Review's analysis arrives at a conclusion the AXD Institute has held since September 2024: that agentic commerce is not a technology upgrade but a structural transformation requiring new design disciplines. The article identifies five distinct failure modes that organisations face when AI agents begin acting as customers on behalf of humans.

These failure modes - agent misunderstanding, authority overreach, data liability, brand misrepresentation, and recovery failure - are not speculative risks. They are the operational consequences of deploying autonomous systems without the trust architecture to govern them. Each maps directly onto an AXD practice framework.

How do the five failure modes map to AXD frameworks?

Agent misunderstanding is a delegation design problem: the human's intent was not translated accurately into the agent's action parameters. Authority overreach is a consent horizon failure: the agent exceeded the boundaries of its delegated authority. Data liability is an agent observability gap: the organisation cannot trace what data the agent accessed, processed, or shared.

Brand misrepresentation is a trust architecture failure: the agent presented the brand in ways the organisation did not authorise. Recovery failure is a failure architecture gap: when something goes wrong, there is no designed pathway back to a trustworthy state.

• Agent misunderstanding maps to Delegation Design - intent was not accurately translated

• Authority overreach maps to Consent Horizon - agent exceeded delegated boundaries

• Data liability maps to Agent Observability - no trace of data access or sharing

• Brand misrepresentation maps to Trust Architecture - unauthorised brand presentation

• Recovery failure maps to Failure Architecture - no designed pathway to recovery

Why does the shift from SEO to GEO matter?

The authors' call for generative engine optimisation (GEO) over traditional SEO confirms a shift the AXD Institute has been mapping since its founding. Traditional SEO optimises for human attention: rankings, click-through rates, and time on page. GEO optimises for machine evaluation: structured data, factual accuracy, and protocol-level discoverability.

This is not a minor adjustment to marketing strategy. It is a fundamental reorientation of how commercial organisations present themselves. The machine customer does not browse, scroll, or respond to emotional appeals. It evaluates structured data against delegated criteria and transacts accordingly.

What is the trust infrastructure analogy and why does it matter?

HBR's most significant contribution is the analogy to early e-commerce trust infrastructure. When online shopping emerged in the late 1990s, the technology existed before the trust layer. SSL certificates, PCI compliance standards, and payment card security frameworks had to be built before consumers would transact online at scale.

Agentic commerce faces the same structural requirement. The AI capability exists. The protocols are emerging. But the trust infrastructure - the standards, verification mechanisms, and governance frameworks that enable humans to delegate commercial authority to autonomous systems with confidence - is still being designed. This is the work of Agentic Experience Design.