Roadmaps in young categories are wrong in detail and useful in shape. This one organizes what is probable across three horizons, labels what is speculative, and makes no calendar-year predictions beyond the first horizon.
Horizon 1 — Near (next ~12 months)
Status: emerging, mostly probable.
- Stripe Agentic Commerce, Visa Intelligent Commerce and Mastercard Agent Pay exit preview into broader merchant availability.
- Major PSPs offer an "accept agent payments" toggle with standard policy defaults.
- Shopify, WooCommerce and Adobe Commerce ship first-class support for agent-readable catalog endpoints (MCP-flavored).
- Google Merchant Center feed spec adds more policy fields (warranty, age restriction, digital delivery).
- Major answer surfaces (ChatGPT Shopping, Perplexity Shop, Amazon Rufus) publish merchant guidelines and verification programs.
- An independent feed/structured-data linter dedicated to agent-readiness becomes a recognized category.
Horizon 2 — Medium (12–24 months)
Status: probable, partly speculative.
- "Agent-pay acceptance" becomes a normal checkout badge alongside Apple Pay, Shop Pay, PayPal.
- Merchant-side "agent traffic" analytics becomes a standard dashboard; 5–20% of retail traffic is agent-routed.
- A de facto MCP-based commerce server spec emerges (catalog, offer, policy, order).
- Platform-level agent identity begins: verifiable claims attached to agent user-agents.
- Cross-marketplace identifier reconciliation improves (Amazon ASIN ↔ GTIN bridging, for example).
- Insurance and fraud products adapt to agent-attributed transactions.
- Structured returns and warranty claims from agents become routine for top-100 merchants.
Horizon 3 — Long (24–36+ months)
Status: speculative but directionally credible.
- Agent-first discovery overtakes classical SERP for some commerce verticals (electronics, travel, consumables).
- Brand-level agent advocates become a merchandising role — the person responsible for how the brand is retrieved by agents.
- Commission and attribution economics shift: agents as a "channel" with standardized attribution metadata.
- A thin, open "Commerce Object Model" emerges as a candidate W3C or IETF-adjacent effort — converging schema.org, GS1 and MCP commerce flavors.
- Decentralized identity for agents and merchants becomes practical (VC/DID).
- Regional regulators publish first rules on agent-led transactions (consent, evidence, returns).
Risks and unknowns
- Platform concentration. If one closed platform captures the majority of agent-mediated traffic, the "open protocol" vision may fragment.
- Regulatory drag. Data protection and consumer protection may slow or reshape agent-pay rollouts.
- Fraud adaptation. Bad actors will target agent-pay with new attack surfaces. Merchant fraud stacks must adapt.
- Model economics. If answer-surface economics do not work, some of horizon 2 may stall.
What to plan for now
- Budget catalog-quality work this fiscal year. It compounds across all horizons.
- Ask your PSP for their agent-pay roadmap; align your accept-payments code paths.
- Start measuring agent traffic today. You will want the trendline.
- Write a thin internal "agent policy" governing returns, customer service and fraud when the counterparty is an agent.
- Assign a single DRI for agent-readiness — not a committee.
Where to go next
- Re-read the thesis.
- Map the moving parts in standards, schemas and protocols.
- Audit your stack with the readiness checklist.