Claim
Gartner forecasts that by 2028, AI-agent "machine customers" will replace 20% of interactions at human-readable digital storefronts, and 33% of enterprise software applications will include agentic AI (up from <1% in 2024). Storefronts and product surfaces designed only for humans will be poorly positioned to capture this segment.
Mechanism
Once consumers and businesses delegate purchase research and execution to LLM agents, the buyer at checkout is no longer human. Storefronts that depend on visual UX, banner messaging, and human-skim copywriting underperform agents that prefer structured data, machine-readable specs, and stable APIs. The transition is gradual but the shape is predictable — storefronts need to expose agent-friendly surfaces or lose the segment.
Conditions
Holds when:
- The category has procurement automation (B2B) or repeated-purchase patterns (commodity B2C).
- LLM platforms continue to expand tool-use capability for purchase actions (currently rate-limited by trust and liability).
Fails when:
- The purchase requires emotional/aesthetic judgment (luxury, fashion, real estate) — agents can shortlist but humans still decide.
- The forecast itself is wrong on timing — Gartner SPAs have a wide error band and 20% by 2028 may slip to 2030+.
Evidence
"Gartner predicts that by 2028, 33% of enterprise software applications will include agentic AI, up from less than 1% in 2024."
"AI agent machine customers will replace 20% of the interactions at human-readable digital storefronts."
— Daniel Sun, Jorge Lopez et al., Gartner Innovate Business Models Using AI Agents (G00824001), 2025-01-16.
Signals
- Storefronts publish machine-readable product specs (schema.org, structured catalogs).
- API/MCP access for purchase actions becomes a vendor differentiator.
- Marketing teams measure agent-mediated conversion as a separate funnel stage.
- B2B procurement workflows automate RFP responses through agents.
Counter-evidence
The 33%-by-2028 number is a strategic planning assumption, not an outcome forecast — Gartner's own SPA error bands are wide. The "20% machine customer" claim assumes agent-payment infrastructure matures faster than 2026 trajectory suggests. Operators should treat as directional, not as a planning target.
Cross-references
- LLM platform is the new OS, agents are the new apps, MCP registries are the new app stores — Mark Petty's complementary platform-layer frame.