Mental Models for the AI EconomyAI Economics
Agentic Checkout: Is Your Store Buyable by AI?
Being recommended is no longer the finish line. AI agents need reliable product state, policy, authority, and checkout rails to complete the transaction.
The next customer may never see your product page. Their AI agent will inspect the product, compare constraints, confirm a mandate, and attempt the purchase. A store can be visible to AI and still be impossible for AI to buy from.
Recommendation was only phase one
AEO and GEO focus on being named and cited. Agentic commerce extends the funnel: discovery, comparison, negotiation of constraints, checkout, and post-purchase actions can happen inside an agent experience. Google’s Agent Payments Protocol describes roles for agents, merchants, credential providers, and payment processors. OpenAI and Stripe’s Agentic Commerce Protocol publishes checkout and payment interfaces. Both are evolving.
Can an AI agent move from recommendation to transaction?
The six layers of agent-buyable commerce
- Identity: stable product IDs and variants.
- State: current price and inventory, not cached copy.
- Constraints: shipping, returns, eligibility, subscriptions.
- Intent: proof of what the customer authorized.
- Transaction: a checkout path an agent can complete.
- Audit: a record humans can inspect and dispute.
In agentic commerce, product content becomes executable infrastructure.
What to do now
Do not rebuild checkout around one beta protocol. Clean the underlying product and policy data first. Make availability reliable, returns explicit, variants stable, and total cost calculable. Those investments improve today’s feeds and marketplaces while preserving options for tomorrow’s agent protocols.
Sources: Google Cloud on AP2, AP2 specification, and Agentic Commerce Protocol.
Get the next essay the day it publishes.
A few a month, written from real builds. Unsubscribe in one click.
Performance marketing and search leader who builds — 15+ years in digital growth and product (AT&T/DirecTV, Microsoft, Employbridge, Site Impact), now running his own properties with AI. These essays are field notes from that loop: what AI actually changes about work, product, and the economics underneath. Based in South Florida.
- AI EconomicsYour AI Can Finally Stop Talking and Start Doing
For two years, AI has been the world's most brilliant intern locked in a room with no laptop. Model Context Protocol just handed it a badge, a login, and actual work to do.
5 min read - AI EconomicsThe Next KPI Every AI Company Will Track: The LLM Token Expenditure Index
We've spent two years asking which AI is smartest. The better question — the one that will actually show up on executive dashboards — is which AI gets the job done using the fewest resources.
6 min read - AI EconomicsYour Business Probably Doesn't Need GPT-6
The fastest AI isn't always the best AI. Choosing a model should look like hiring an employee — matched to the job — not chasing the highest IQ score on a leaderboard.
5 min read - AI EconomicsWhen Code Is Cheap, What Still Protects a SaaS?
AI lets anyone clone a feature over a weekend. The moats that remain are data, workflow, distribution and trust—and you can audit yours in two minutes.
8 min read · interactive - Web & AI SearchAnswer-First Writing: The Inverted Pyramid for Machines
Put the answer in sentence one under a question-shaped heading. The four rules that make a paragraph quotable by snippets, voice and AI assistants — with an interactive before/after.
7 min read · interactive - AI DevelopmentThe Practice: Spec-Driven, Maker-Checker, Governed Scope
Two credible trials on AI coding point in opposite directions — 55.8% faster in one, 19% slower in the other. The difference isn't the model. It's the practice. Three patterns, with interactive math and cited data.
9 min read · interactive