Agentic Commerce Isn’t Monolithic. And That’s the Point
The fintech world is gripped by a new anxiety: agentic commerce. Overall, trust is questioned because objectives are not aligned.
Stories of AI bots autonomously making purchases, navigating checkouts, and completing transactions have triggered a familiar refrain. While merchants are wary and payment processors are experimenting, regulators are watching.
Agentic commerce is being discussed as if it’s a single, homogeneous topic. And it isn’t. That distinction is precisely what should guide regulatory and commercial thinking.
Start with what’s already normal: merchants and platforms have deployed AI for nearly a decade to shape consumer behavior. Dynamic pricing adjusts rates based on behavioral signals. Recommendation algorithms are engineered to maximize engagement and conversion for merchant benefit, not always consumer benefit. UI/UX is optimized through A/B testing and behavioral psychology, often to entice rather than inform. Dark patterns are deployed across e-commerce: artificial scarcity, urgency nudges, hidden charges buried in checkout flows.
The AI Asymmetry That Nobody Questions
This is AI working often working against consumer interest. It’s commerce and from a consumer perspective it often requires calm, concentration and a lot of patience.
Now introduce a consumer tool: an AI agent that evaluates options against consumer criteria like price, quality, size, technical specs, and autonomously executes a purchase that aligns with what the consumer actually wants. Suddenly, the conversation shifts. There’s talk of fraud, trust deficits, agent empowerment, and regulatory risk.
The asymmetry is uncomfortable: AI that serves merchant interests is innovation; AI that serves consumer interests is risk.
A Precedent Most People Miss
The travel industry already solved this. Kayak, Skyscanner, and Google Flights are agents. They don’t just recommend holidays or flights, but they actively filter, rank, and execute consumer searches against criteria the consumer sets. These tools have been operating at scale for 15 years. Nobody questions their legitimacy. Merchants (airlines, hotels) don’t reject them as untrustworthy. Regulators don’t treat them as systemic threats.
Why? Because the principal-agent alignment is transparent. The agent works for the consumer. The consumer sets the criteria. The agent executes.
Extend this logic forward: if a consumer can authorise an agent to filter and select among holiday packages, why not among other products via e-commerce channels? Why not at payment and checkout? The technology is already proven. The trust model is already understood.
The difference is friction. Flight comparison sites don’t threaten at scale how airlines manipulate consumer choice at the point of sale. Agentic commerce does.
Agentic commerce in the hands of hyperscalers (Google, Amazon, Meta etc), who already store consumer payment methods and already have entrenched themselves into consumer daily routines with AI assistants are actively pushing the boundaries of agentic commerce.
What Merchants Actually Fear
Merchant resistance isn’t fundamentally about fraud or security. It’s about losing control, especially because they also rely on the same hyperscalers to showcase and sell their products.
When consumers can programmatically filter by genuine criteria like lowest total cost, fastest shipping, best reviewed, merchants can no longer rely on:
- Opaque pricing: Hidden fees buried in checkout
- Dark patterns: Artificial scarcity, pre-ticked options, misdirecting button placement
- Psychological manipulation: FOMO, urgency, color psychology, visual hierarchy
- Information asymmetry: Knowing consumers won’t read terms, won’t compare alternatives, won’t check reviews carefully
An AI agent, configured by and for the consumer, strips away these levers. It doesn’t care about visual hierarchy. It reads terms of service. It compares prices across merchants automatically. It evaluates reviews at scale.
Merchants who compete on actual product quality, fair pricing, and genuine service innovation have nothing to fear. Merchants whose margins depend on heavy sponsorship, consumer confusion, behavioral capture, or hidden terms will feel threatened.
Who Builds Agentic Commerce Matters
Here’s where nuance enters: agentic commerce is a category that can be built by multiple types of players, each with different incentives.
1. Merchants themselves
A retailer builds an agent that helps consumers configure and purchase products based on dimensions, budget, and aesthetic preference. The agent works within the merchant’s own catalog and ecosystem. It’s personalized customer service at scale.
2. Marketplaces
Amazon, eBay, or vertical marketplaces build agents that help consumers navigate their massive inventory based on flexible criteria, not just a confusing and buggy list of static filters. The agent is native to the platform. It drives efficiency.
3. Payment Service Providers
A PSP builds an agent layer that performs the checkout on behalf of the consumer selecting the payment method that best matches the criteria of the consumer (BNPL with best conditions vs card vs digital wallet).
4. Third parties (comparison sites, AI natives)
A standalone comparison site or AI-native entrant builds an agent that aggregates across merchants and helps consumers make cross-merchant decisions. It’s the Skyscanner model. The agent represents the consumer, not any merchant.
Each model has different incentive structures and regulatory implications:
- Merchant-built agents are aligned with merchant success but operate within merchant control. Low regulatory friction, but merchant incentives may not be fully consumer-aligned.
- Marketplace agents optimize for platform health and consumer experience on that platform. Because the marketplace serves both merchants and consumers, it is appropriate to question whose interests it serves most (e.g. what happens to the “sponsored listing” model so common on Amazon or eBay?)
- PSP agents sit at the trust boundary. They’re positioned to protect consumer interest at the point most vulnerable: checkout. They are under strict regulatory scrutiny.
- Third-party agents are consumer-aligned by design but lack merchant relationships. They’re the closest to the travel comparison precedent, but they face the most merchant resistance.
This is why I believe the strategic conversation should be specific to agent type, not blanket.
Guardrails Built Around Consumer Interest
The framework is straightforward and no different from the general principles under which e-commerce has been operating: guardrails should protect the consumer’s interest and ensure transparency about the agent’s decision-making, regardless of who built it.
What does that look like?
- Spending authorisation: Agents operate within consumer-set limits (per transaction, per day, per merchant category)
- Transparent decision criteria: Consumers know what factors the agent is optimizing for and how it weights them
- Merchant approval lists: Consumers can restrict agents to approved merchants if they wish
- Recourse mechanisms: Clear paths to reverse or dispute agent-initiated transactions, but I see them as an expansion of existing mechanisms for online transactions.
- Explainability: When an agent makes a decision, the consumer can understand why
These guardrails don’t disable agentic commerce. They make it trustworthy and compliant by design. They also apply regardless of who builds the agent, merchant, platform, PSP, or third party.
Agentic commerce guardrails should protect consumers and ensure fair competitive dynamics.
The Rebalancing Ahead
Agentic commerce isn’t a threat to commerce. It’s a threat to margins built on information asymmetry, behavioral capture, and merchant-side AI manipulation.
The merchant ecosystem will adapt as it did with the advent of price comparison sites, review aggregators, and travel searches. Merchants will compete on genuine product value, fair pricing, and service quality rather than psychological tricks.
The winners won’t be those who resist consumer AI. They’ll be those who build trustworthy, efficient relationships with consumers by earning their patronage through genuine value, not behavioral capture.
The regulatory path is clear: protect consumer interest, require transparency, and enforce the same rigor that applies in e-commerce.
Agentic commerce is coming with different implications for competition, trust, and consumer welfare. It is not just a technical exercise or experiment.
That’s the conversation worth having.