Market Forecasts & Studies
Analysts agree: Agentic Commerce will fundamentally transform digital commerce. The forecasts differ on the pace and volume — from conservative to transformative.
McKinsey: $3–5 Trillion by 2030
McKinsey published the most ambitious forecast to date in February 2026:
- US B2C: Up to $1 trillion in orchestrated revenue by 2030 through Agentic Commerce
- Global: $3–5 trillion aggregate market volume
- Key statement: "Not an incremental increase, but a reconfiguration of the entire commerce ecosystem"
McKinsey's analysis distinguishes between "orchestrated revenue" (purchases initiated or influenced by agents) and "direct agent revenue" (purchases that run fully autonomously). The $1 trillion refers to orchestrated revenue — a significantly broader term. The Microsoft-McKinsey partnership through which these figures were published underscores both companies' commitment.
Gartner: 33% with Agentic AI by 2028
Gartner forecasts that 33% of enterprises will deploy Agentic AI by 2028 — a dramatic increase from under 1% today. Adoption is driven by three factors:
- Maturation of LLM technology and tool-use capabilities
- Availability of open protocols (ACP, UCP)
- Competitive pressure: those who do not automate will lose market share
Futurum Group: Enterprise Perspective
The Futurum Group surveyed 830 enterprise decision-makers in the first half of 2026. The results:
- 39% expect to use GenAI primarily through agents (not through chat interfaces)
- 38% rank Agentic AI as a top-2 purchasing criterion for software procurement
- 56% cite talent shortage as the biggest hurdle for Agentic AI adoption
- 66% prefer a platform-first approach (extending existing platforms with agent capabilities)
- 71% are open to a vendor switch in 2025–2028 if their current provider does not offer agentic capabilities
Bain & Company: Consumer Adoption
Bain & Company provides the consumer perspective: 30–45% of US consumers already use generative AI for product research. This means: the demand side is ready — consumers already trust AI systems for purchase preparation. The step from "AI for research" to "AI for purchasing" is technically possible (via ACP/UCP) and culturally not far off.
PYMNTS/Visa: Acquirer Perspective
A PYMNTS study in collaboration with Visa (March 2026) examines the payment infrastructure side:
- 90%+ of acquirers cite governance, permissions and consent as top priorities for agent transactions
- Agentic Commerce is classified as a "systems-and-trust-layer problem" — not purely a technology problem
- The payment industry sees the greatest need in standardized authorization frameworks (→ AP2, Mandates)
Putting the Numbers in Context
What the numbers mean — and what they do not:
- The direction is clear: All major consultancies, tech companies and financial institutions see Agentic Commerce as a transformative trend
- The ranges are wide: McKinsey's $3–5 trillion reflects different adoption scenarios. The optimistic end assumes rapid protocol standardization and global payment infrastructure
- Definitions vary: "Agentic Commerce" is defined differently. McKinsey's "orchestrated revenue" also includes purchases influenced (not fully executed) by agents
- Europe lags behind: Most figures refer to the US. Europe will adopt more slowly due to regulatory requirements (GDPR, PSD2, AI Act)
Frequently Asked Questions
How reliable are these forecasts?
Market forecasts for new technologies inherently carry high uncertainty. The ranges ($3–5T) reflect different adoption scenarios. The direction is clear (significant growth), the exact volume uncertain.
When will Agentic Commerce go mainstream?
Most analysts expect the tipping point in 2027–2028. First implementations have been live since 2025, broad adoption will follow once standards are established and payment infrastructure is globally available.
Will Agentic Commerce replace traditional e-commerce?
No. Agentic Commerce will be an additional channel — just as mobile commerce complemented desktop e-commerce rather than replacing it. McKinsey speaks of "reconfiguration", not replacement.