Agentic commerce is turning shopping into delegation 

The old idea that needs retiring, retail assumes the human is always the shopper 

For most of modern e-commerce, the customer journey has been designed around a single assumption, a human will browse, compare, decide, then click “buy”. Even when we modernised the interface, mobile-first design, one-click checkout, personalised recommendations, we kept the same mental model. The human remained the decision-maker, the website or app remained the arena, and “conversion” was the moment the person completed the last step. That model still works, but it is no longer the whole game. The hottest topic in retail and e-commerce right now is the emergence of agentic commerce, shopping where an AI agent acts on behalf of the customer, moving from recommending to executing.  

A new way to thin: The customer is becoming a delegator, not a browser 

Here is the educator twist that changes everything, what if the new “customer” is not a person clicking through pages, but an agent interpreting a person’s intent. The human does not want 200 search results and a filter panel, they want an outcome, “find me the best-value option that meets these constraints, and buy it when the price is right”. Agentic commerce is the shift from shopping as interaction to shopping as delegation. It is not simply a better chatbot. It is a new layer in the economy where intent, trust, policy, and permission become the core design problem. If you accept that, you stop asking superficial questions like “should we add AI to our site”, and start asking structural ones like “how does an agent understand our catalogue, our pricing, our inventory, our delivery promises, and our returns logic”. 

The biggest platforms are already fighting over shopping agents 

This is not theoretical. The global market is already producing legal and commercial friction because the stakes are obvious. Reuters reported in March 2026 on a court decision temporarily allowing Perplexity AI’s shopping agents to operate on Amazon while an appeal proceeds, in a dispute that explicitly revolves around agentic tools and platform access. This kind of fight is a signal, not noise. When distribution shifts, incumbents defend the gates. The same week, Reuters described Alibaba’s increasing focus on AI agents across its ecosystem, positioning agents as a central part of digital consumption and enterprise automation.  

The lesson the market is learning in real time: Agents must shop like humans actually shop 

The first generation of agentic shopping experiments exposed a blunt truth, shopping is rarely a single-item transaction. Wired reported that Walmart and OpenAI reworked their approach after disappointing performance from an “Instant Checkout” style experience, in part because it did not reflect how customers bundle and buy in real life, and Walmart’s assistant approach aims to integrate with carts rather than forcing unnatural item-by-item behaviour. The educator takeaway is that agents will not replace retail journeys, they will compress and reshape them. That reshaping rewards retailers that have clear, structured product information and coherent basket logic, substitutions, bundles, delivery thresholds, loyalty benefits, and returns. If your retail experience depends on humans patiently navigating ambiguity, agents will expose that weakness fast. 

Why this is the hottest topic: It shifts power from “being found” to “being chosen by machines” 

For the past decade, many brands treated visibility as the main battle, win search, win social, win marketplaces, win the algorithm. Agentic commerce escalates that. The question becomes, will the agent recommend you, trust you, and complete a purchase with you. That is a more demanding test because agents prioritise clarity, reliability, and low risk. A product page full of vague claims can still sell to humans. An agent will mark it as uncertain. A returns policy written like a trap can still convert a first-time buyer. An agent will downgrade you because the downside risk is too high. This is why governance and structured truth matter more than aesthetics. Deloitte’s 2026 retail outlook explicitly frames AI-driven commerce and data-driven insight as central dynamics reshaping competition.  

Agentic commerce fits the continent’s real buying behaviours 

Now narrow the lens. Africa is often framed as “catching up” to global e-commerce, but that framing is not only tired, it is strategically misleading. Many African markets are mobile-first, conversational by habit, and deeply shaped by trust networks. In that environment, the agentic model, “tell the system what you need, let it do the work”, is not alien. It is aligned with how commerce already happens, through dialogue and recommendation, just digitised and scaled. Adoption still depends on infrastructure, connectivity, smartphone affordability, and reliable delivery, but the interface itself maps naturally to the culture of buying. 

The constraint that becomes an advantage: Trust and rails are the real product 

Agentic commerce in Africa will be won less by flashy AI demos and more by trust infrastructure. Payments are not just a checkout step, they are inclusion. Regulatory harmonisation is not just policy, it is the ability for digital trade to scale safely across borders. That is why Africa’s digital trade direction matters. The AfCFTA Digital Trade Protocol was adopted in February 2024, aiming to establish harmonised rules and common principles for digital trade, including areas tied to business and consumer trust, data governance, and cybersecurity.  

Africa will not delegate to agents unless the agent can actually deliver 

Delegation only works if outcomes are reliable. If an agent orders the right product but delivery is unpredictable, the customer will not delegate again. If returns are difficult, the customer will avoid agent-driven purchases for anything risky, fashion sizing, electronics compatibility, high-ticket items. This is why logistics models like pickup points and flexible fulfilment will matter enormously, and why the “agentic” future in Africa will often be hybrid, mixing delivery, pickup, and store-based collection. The continent’s opportunity is to build a model where agents are not just shopping assistants, they are routing systems that choose the fulfilment path that is most reliable in a given geography, at a given time, for a given customer preference. 

South Africa is the test bed: The market has scale and competitive pressure now 

Now bring the funnel to South Africa, where the conditions for an agentic commerce inflection are already visible. Reuters reported in September 2025 that a World Wide Worx study forecast South Africa’s online retail turnover would surpass R130 billion in 2025, reaching about 10% of total retail, after a 35% increase in 2024 to R96 billion and about 8% of retail. The same reporting highlighted that on-demand grocery and improved fashion platforms are key drivers, and that competition is intensifying, including global entrants and cross-border platforms. Mastercard’s newsroom release on the same study adds detail on category momentum, including rapid growth in on-demand grocery and strong growth in online fashion and home categories.  

Why agents matter specifically in South Africa: They will punish friction and reward reliability 

South African e-commerce is reaching the point where marginal gains are won through operational credibility, not just marketing. Agents will accelerate that shift. They will steer customers toward retailers with accurate stock visibility, dependable delivery windows, clear substitution rules for grocery, and returns processes that do not require a fight. This is where agentic commerce becomes a board-level topic, not because it is trendy, but because it changes how customers are allocated across competitors. If an agent can see that Retailer A has ambiguous delivery promises and Retailer B has transparent delivery and simple returns, Retailer B becomes the safe choice. Over time, safe choices become default choices. That is how market share quietly moves. 

Stop treating AI as a feature, treat it as a distribution regime change 

Here is the narrative I want to challenge. Too many retailers still discuss AI as a feature roadmap, a chatbot here, a product copy tool there, a recommendation widget somewhere else. That mindset is the same mistake retailers made in the early mobile era, when they treated mobile as a smaller website rather than a new behaviour. Agentic commerce is not a feature. It is a regime change in distribution. It will reorganise how discovery works, how trust is earned, and how purchasing decisions are executed. The retailers who treat it like a toy will get routed around by the retailers who treat it like infrastructure. 

The practical playbook for 2024 to 2028: Build a retail system that agents can trust 

The winning moves are not mysterious, they are just demanding. First, make your product truth structured and consistent, attributes, compatibility, sizing logic, warranties, and clear imagery. Second, unify inventory truth, so your availability claims are reliable across channels. Third, make policies explainable, returns, refunds, delivery thresholds, substitutions, and service escalation paths should be clear enough that an agent can summarise them without guessing. Fourth, invest in fulfilment resilience, especially hybrid fulfilment, because agents will choose the option most likely to keep the promise. Fifth, treat security and trust as part of customer experience, because delegation collapses if the customer feels unsafe. These are the same fundamentals Deloitte highlights, customer centricity, operational excellence, and data-driven insight, but in an agentic world they become existential rather than aspirational.  

The future retailer is not the loudest, it is the most legible 

Agentic commerce forces retail to become legible. Legible to machines, and therefore dependable to humans. The hottest topic in retail is not “AI” in the abstract, it is the specific shift where software begins to buy, and that turns retail into a new kind of competition. Those who build clean data, clear promises, and reliable fulfilment will be chosen more often, by agents, and by the people who delegate to them. The window from 2024 to 2028 is when those habits will form. If you wait until the agents are mainstream, you will already be negotiating from behind.