Overview
McKinsey projects $1 trillion in agent-mediated retail sales by 2030, but most businesses remain invisible to AI agents because their systems aren’t agent-readable. The companies that survive the next three years will be those that restructure their data architectures to make their entire business readable and writable by AI agents, moving from 20 years of anti-bot defenses to pro-agent infrastructure.
Key Takeaways
- Build agent-first data architecture now - The work of cleaning and restructuring your data for agent readability takes months or quarters, and waiting means becoming invisible to the growing share of agent-mediated commerce
- Move beyond surface-level API wrappers - Simply wrapping existing APIs in MCP servers only covers a few percentage points of the use case; true agent readability requires clean data all the way down the stack
- Capture tribal knowledge in structured data - 80% of product meaning exists in marketing copy rather than data structures, but agents need this context in readable formats to match human intent with products
- Agent discovery works differently than search - Unlike humans browsing ranked lists, agents evaluate structured data against explicit constraints, so clean schemas and low-friction data access matter more than ad budgets
- Start with competitive benchmarking - Test how far you can get transacting with your top three competitors using Claude or ChatGPT, then compare to your own systems to identify where you can lead or need to catch up
Topics Covered
- 0:00 - The Agent-Readable Infrastructure Crisis: Companies built 20 years of anti-bot defenses that now block valuable AI agent customers, requiring massive ecosystem changes for agents to function
- 2:30 - Data Quality Determines Agent Effectiveness: Lessons from Prime Video on how clean underlying data is essential for good customer experiences, now critical at scale for agent interactions
- 4:00 - The Battle Between Walled Gardens and Agent Access: Big tech companies like Google and Apple resist agent-readable systems while startups embrace them, creating competitive opportunities
- 6:00 - McKinsey’s $1 Trillion Projection and Current Reality: Agent commerce is already happening with specific examples of shopping queries, but success depends on clean schemas and clear product data
- 8:00 - Why Surface-Level Solutions Don’t Work: Simply adding API wrappers isn’t enough; true agent readability requires fundamental changes to internal data architecture
- 12:30 - Stripe vs SAP: The Implementation Spectrum: Real examples of how even forward-thinking companies like Stripe face complex challenges, while legacy systems like SAP lag far behind
- 16:30 - Four Critical Misconceptions About Agent Commerce: Common wrong assumptions about agent discovery, product complexity, customer trust, and the ‘wait and see’ approach
- 21:30 - The Tribal Knowledge Problem: Most product meaning exists in marketing copy rather than structured data, requiring massive effort to make this information agent-readable
- 26:00 - Practical Steps for Getting Started: Competitive benchmarking exercise to test agent interactions with competitors and identify opportunities for leadership or catch-up