Enabling customers to simply talk or text with automated systems to make purchases is an opportunity to enhance customer loyalty and increase lifetime value, if done right. But many deployments in conversational self-service have yielded underwhelming bot experiences, clumsy handoffs, and yet another interface layer between customers and what they actually want to do.
Agentic commerce changes that because it's not about talk, it's about action. Artificial intelligence agents can see a goal, navigate underlying systems, and complete the job on the customer's behalf.
Opus Research began tracking this shift when we coined the phrase conversational commerce in 2011 to describe natural language as a new front door to customer experiences. The idea was straightforward in that customers could express intent in their own words. Systems would interpret that intent and orchestrate the purchase. In practice, though, many bots provide unhelpful or inaccurate answers, route inquiries when stumped, while occasionally taking a successful order. They rarely reduce friction, and too often they add to it, forcing customers to repeat information or navigate rigid decision trees.
Today, we are reaching an inflection point as large language models (LLMs) make it possible for agents to hold context, reason across steps, and independently pursue goals. Retailers like Walmart have introduced AI shopping companions, while other companies have launched sophisticated, proactive agents for product discovery and considered purchases. These are not chat widgets bolted onto a website, but rather early examples of a new model in which the primary interface is an intelligent agent that can understand, plan, and act.
The most consequential shift, though, has come from the platforms that control consumer attention. Emerging protocols allow merchants to expose catalog, pricing, and fulfillment logic into third-party agent interfaces. OpenAI turned ChatGPT into a native checkout surface with "Buy It in ChatGPT" and open-sourced the Agentic Commerce Protocol (ACP). Google followed with the Universal Commerce Protocol (UCP) that explicitly covers post-purchase workflows (order status, cancellations, returns, refunds, exception handling, etc.) making it a de facto operating layer for customer service as well.
CX leaders need to pay attention. Customer experience increasingly happens inside someone else's interface, driven by someone else's agent but constrained by your policies, your data quality, and your ability to execute.
Agentic commerce is defined less by any single technology than by a set of capabilities, including the following:
- Multi-step task completion without constant customer input (search, compare, purchase, and track in one continuous flow).
- Persistent context across sessions and channels so the agent remembers preferences, constraints, and history.
- Decision-making within guardrails. Agents can choose products, apply policies, and trigger workflows within parameters you set.
- Deep integration with order, inventory, CRM, payments, and logistics systems, enabling real changes in the physical world.
- Learning from interactions to improve recommendations, routing, and resolution over time.
The practical distinction is simple: Chatbots complete conversations while agentic systems complete transactions. Customers no longer ask "Where is my order?" and receive a link. Now they can say, "This product arrived damaged, and I need a replacement before Friday." The AI agent checks inventory, applies policy, schedules a shipment, and only involves a human if it hits an edge case.
Why This Matters to CX Leaders
Once agents can transact, they can also initiate many of the interactions that historically generated contact volume, such as alerts on delays, proactive offers to reschedule deliveries, subscription renewals, replenishment reminders, and exception handling when something goes wrong. UCP and similar protocols are designed with these flows in mind, which moves pressure directly onto CX-owned assets like knowledge bases, policy engines, and the contact center itself.
For contact centers, this is an existential shift. Traditional performance metrics (average handle time, cost per contact, call deflection) made sense when the job was to process volume. When AI agents can autonomously complete the majority of routine tasks, the role of human agents changes. They become exception specialists, dealing with complex, emotional, or high-value scenarios where judgment and empathy matter more than speed.
That, in turn, transforms the contact center from a cost center into a revenue and relationship engine. Freed from simple returns, human agents can focus on retention saves, cross-sell that genuinely fits customer context, and resolving problems in ways that reinforce trust rather than eroding it. The balance between experience, satisfaction, and profitability becomes more delicate, not less, because ill-designed automation can destroy trust even faster than a bad human interaction.
In an agentic world, the friction point is not what the AI agent can do, it is what happens when it cannot finish the job. If a customer must start over with a human after an agent fails, you have simply moved the frustration upstream. The winners will design seamless agent-to-human handoffs, with full context transfer and clear role boundaries.
Trust becomes the other governing axis. Consumers are already using LLM-powered apps as replacements for search, shopping, and work and as trusted assistants for financial and personal decisions. They willingly share data in exchange for personalization and convenience, creating long-lived context that companies can use for pricing, recommendations, and service. The risk for CX leaders is not that they will under-personalize but likely they overreach, using context in ways that feel manipulative or opaque and thereby erode hard-earned trust.
The organizations that lead customer experience in 2026 will master the interplay between orchestration, trust, and scale. They will meet customers where they are, keep their data honest, and deliver reliably when the customer says "buy." The underlying technology for agentic commerce is ready. The real differentiator now is strategic execution and measuring success by how quietly and competently AI agents deliver better outcomes for customers.
Derek Top is principal analyst and research director at Opus Research.