Publicis Sapient and Salesforce outline roadmap as agentic commerce accelerates
A new whitepaper from Publicis Sapient and Salesforce warns that commerce orchestrated by AI agents is moving from experiment to operating model fast, and that organisations which move early will have a structural advantage. The joint report, titled “Agentic Commerce: Unlocking the Emerging Shopping Experience,” projects that AI agent-orchestrated transactions could generate between $3 trillion and $5 trillion in global revenue by 2030 and sets out an architectural blueprint for businesses, especially across EMEA, to become agent-ready.
Data points underline rapid shift in discovery and demand
The report aggregates market signals that point to sharp behaviour change: adoption of AI-powered search is rising, platform agents are fielding hundreds of millions of shopping queries weekly, and referral traffic from generative AI tools to retail sites spiked dramatically in mid-2025. Publicis Sapient and Salesforce argue these indicators show the agent-mediated commerce era is already underway rather than a distant prospect.
That pace matters because, unlike prior platform shifts that required new devices or mass consumer hardware changes, AI agents plug into existing digital infrastructure — APIs, payment rails and fulfilment systems that enterprises already run. That makes the barriers to entry lower for agent builders and raises the cost of inaction for retailers that lack agent-addressable capabilities.
Framework for implementation: A.C.E. architecture
To translate opportunity into operational plans, the paper introduces the A.C.E. Framework as a three-layer architecture intended to make products and services discoverable and actionable by AI agents.
Agentic Experience Interface focuses on exposing product and service semantics so external agents can find, evaluate and act on offers. That includes structured metadata, intent signals and transactional hooks that agents need.
Composable Micro-Apps advocates breaking core commerce functions into modular services — pricing, inventory, promotions, recommendations and checkout — that can be invoked programmatically by agents, partners or third-party platforms.
Enterprise Context Orchestration covers governance and context management across the agent ecosystem, with emphasis on data control, security, compliance and the systems needed to provide agents with the right contextual signals while protecting consumer privacy.
Three agent types and the practical questions they raise
The report categorises the agent landscape into platform agents, brand-owned agents and personal consumer agents. Platform agents include general-purpose models and assistants such as ChatGPT and Google Gemini that can act as intermediaries between users and merchants. Brand-owned agents are proprietary assistants or APIs that operate within an enterprise’s ecosystem. Personal consumer agents are agent instances that act on behalf of individual shoppers, retaining preferences and memory.
Each agent type creates different operational priorities: platform agents drive discoverability and standards compliance; brand agents require strong identity and data integration; personal agents create requirements around long-term memory, preference management and trusted recommendation. The paper stresses that optimisation needs to happen not just for humans but increasingly for these software intermediaries.
Risk, failure modes and the hard work of delivery
The whitepaper does not frame agentic commerce as an automatic win. It highlights research — cited from an academic source — that suggests a large share of generative AI initiatives fail to produce measurable business impact. Causes include poor workflow fit, absence of memory architectures and unclear service-level targets. The authors recommend disciplined, phased rollouts and provide a suggested blueprint intended to move capable projects from pilot to production within 90 days, a substantial acceleration compared with current market averages.
Standards and market formation
The report highlights the emergence of the Universal Commerce Protocol, launched in January 2026 with backing from major payments and retail players. That protocol aims to provide open standards for agent-mediated transactions. Early integration with such standards could position brands as trusted data sources within agent knowledge graphs and influence future recommendation flows as market consolidation occurs.
Implications for EMEA retailers and enterprise leaders
For retailers in EMEA, the takeaway is strategic: agentic commerce is not only a technology project but a change to how demand is discovered and fulfilled. Companies will need to consider new metadata strategies, composable service architectures, partner and protocol integrations, and tighter controls over customer context and consent. Organisations that move quickly could secure durable advantages in attribution and integration, according to Salesforce’s regional leadership quoted in the report.
Operationally, teams should prioritise: (1) auditing current APIs and data readiness for external agent consumption; (2) piloting composable micro-services for key commerce functions; and (3) establishing governance for how agents access and use enterprise context. Those steps will determine whether businesses can capture agent-originated demand or risk being bypassed as recommendation flows consolidate around a smaller set of agent-friendly suppliers.
As enterprises weigh investment, the new whitepaper provides a practical starting point rather than definitive prescriptions. Its value is in combining market-scale projections with an implementation-oriented framework that maps directly onto engineering and product decisions. For EMEA retail and commerce leaders, the question now is whether to treat agentic commerce as a strategic priority and begin the work of becoming addressable to the software intermediaries that increasingly shape customer journeys.
Read the full report from Publicis Sapient and Salesforce for more detail on the A.C.E. Framework and implementation guidance.
