Why AI Agents Still Won't Pay for Anything — Jan Curn, Apify
Summary
The video analyzes the nascent field of agentic payments, arguing that despite the emergence of numerous payment protocols (e.g., x402, AP2, ACP, MPP), real-world adoption remains low. The core challenge lies in the agent payment funnel, where friction points exist at every stage: establishing the initial need, setting up payment mechanisms, discovering trusted services, and executing variable-cost transactions. The speaker posits that API keys currently dominate because they offer superior convenience compared to complex crypto setups, and that the transition requires improvements in LLM-driven agent autonomy and the implementation of robust, pay-per-use billing models like batch settlement.
Key takeaways
-
API Keys vs. Agent Payments
Currently, API keys are the dominant method for service access because they are more convenient than setting up autonomous agent payments. Agentic payments are competing with the established ease of use provided by fixed API key access.
-
The Agent Payment Funnel Friction
13:56
The adoption of agent payments is hampered by friction at four stages: 1) Need (agents need more than fixed services); 2) Setup (crypto wallets are still opaque and difficult to set up); 3) Discovery (agents struggle to find and trust services); and 4) Buying (variable-cost billing is complex).
-
Crypto's Advantage in Payments
Crypto is uniquely positioned for agentic payments because it lacks a single owner, is cheaper than traditional methods for microtransactions, and, critically, does not support chargebacks, which are impossible to manage with an infinite number of autonomous agents.
-
The Need for Variable Billing
Traditional fixed-price per API call models are insufficient for complex agent jobs. The industry needs variable pricing and mechanisms like batch settlement to handle pay-per-use billing for compute jobs and LLM usage.
Technical details
-
Agent Payment Protocols
1040s
A wide array of standards are emerging, including L42 (Lightning Labs), Mastercard Agent Pay, x42 (Coinbase), KYA pay (Visa), AP2 (Google), ACP (OpenAI/Stripe), TAP (Visa), UCP (Google/Shopify), ACTP (Alipe), MPPP (Stripe), MMP (Alipe), and APOP (Union Pay). This indicates a heated battleground for future agent e-commerce infrastructure.
-
Payment Settlement Mechanisms
The evolution of payment schemes shows a progression: X42 initially required a signature and subsequent settlement to the blockchain. While 'up to' payment schemes allowed variable cost charging, they failed for pay-per-use because the agent could spend the funds twice before final blockchain settlement. The 'batch settlement' scheme was introduced to solve this by using crypto magic to escrow funds on the blockchain, enabling reliable microtransactions.
-
Agent Tooling and Marketplaces
100s
Platforms like Apify operate as two-sided marketplaces, providing 'actors' (tools for automation) that are sold to users. The integration of agent payments is seen as a critical expansion vector for these tool ecosystems.
Mentioned resources
- Apify
- X42
- MPPP
Channel & topics
Watch on YouTube · Back to latest
This independent, AI-assisted summary is provided for commentary and informational purposes. It may contain errors or omit important context. Please watch the original video for the creator's complete presentation. Video, thumbnail, and related copyrights belong to their respective owners.