Get your free personalized podcast brief

We scan new podcasts and send you the top 5 insights daily.

Isenberg describes a three-rung trust ladder for agents: research, prepare and pay. Most companies tolerate research because it resembles search, but many block all three rungs even though the real financial risk only appears on the pay rung, so they should gate payment at checkout instead.

Related Insights

An ad-based model misaligns an agent, incentivizing it to influence users against their own interests. Instinct is pursuing a "blanket transaction take rate," similar to Apple Pay. The agent remains free for the user, ensuring its actions are solely on their behalf, while merchants pay for the distribution.

To overcome user reluctance to share credit cards and passwords with an AI, Dazzle's agents use their own separate accounts for transactions. They can buy tickets or products and then transfer them to the user, acting as a secure intermediary and sidestepping the need for the user to grant direct account access.

Binary decisions are brittle. For payments that are neither clearly safe nor clearly fraudulent, Stripe uses a "soft block." This triggers a 3DS authentication step, allowing legitimate users to proceed while stopping fraudsters, resolving ambiguity without losing revenue.

Isenberg contrasts human brand loyalty with agent loyalty, saying agents return to whatever worked last time. If your site gets the order and delivery right, the agent remembers; if checkout gets stuck, it quietly goes elsewhere and you may never know the customer was lost.

Giving AI agents buying power isn't a simple payments problem. The core challenge is a full-stack platform that helps agents navigate websites, bypass bot detection like Cloudflare, and securely complete complex checkout forms—a far harder problem than just issuing a virtual card.

Consumers will be slow to trust AI with subjective shopping decisions (e.g., buying clothes). However, they will more quickly adopt AI agents to handle the objective task of payment optimization—choosing the right card or financing—making it the first major frontier for agentic commerce.

Users will restrict an AI agent's access if they lack confidence in its actions, ultimately capping utility. Thibaut Sottiaux emphasizes starting agents with conservative guardrails that request permission for every action, ensuring users maintain control as trust develops.

Consumer AI agents frequently stall because websites and service providers deliberately shut down automated checkouts and bot traffic. Platforms recognize that scriptable, agentic interactions strip away brand equity, eliminate top-of-wallet loyalty, and commoditize underlying infrastructure.

Isenberg points out Amazon earns tens of billions from ads and airlines earn heavily from seat upgrades, bags and checkout credit-card offers. An agent still completes the sale but skips sponsored products and upsells, taking away the highest-margin parts of the purchase journey that finance those businesses.

For agents to buy on users' behalf, merchants need a shared technical language to expose catalogs and process payments securely. Protocols like the one Stripe co-created with OpenAI allow merchants to sell through new AI channels without ceding the customer relationship or control over fraud.