AI is already in your customers' shopping journeys: leveraging agentic commerce for enterprise businesses
Agentic commerce, where AI agents research, compare and transact on behalf of consumers, is gaining traction among South African consumers. An estimated 70% of South Africans already use AI chatbots, including for purchase decisions. This article sets out three concrete steps enterprise teams should take now: structuring product data for AI discoverability, building API-first payment architecture and understanding what the businesses leading this shift are already doing differently.
Enterprise conversations about agentic commerce in South Africa often start with a question about timing: is this relevant now, or something to plan for in two or three years? As our 2026 Consumer Payments Report indicates, AI is already in your customers' shopping journeys.
The South African AI adoption picture
70% of South Africans have used AI chatbots, including for use cases that explicitly include research and purchase decisions. Our 2026 How South Africans Shop report found participants describing AI chatbot use for budgeting, financial planning and checking product authenticity before buying. Focus group participants also mentioned using AI to compare product information across retailers before committing to a purchase.
AI use in everyday shopping is a behaviour already present among a substantial share of the South African consumer base, and one that global data suggests will accelerate. According to a recent Shopify study, AI-referred shoppers convert at nearly 50% higher rates and carry 14% higher average order values than shoppers arriving through organic search. As AI agents move from research to comparison to transaction, businesses with structured, machine-readable product and payment data stand to gain a significant and compounding advantage.
What agentic commerce means for enterprise businesses
In traditional e-commerce, the customer navigates a website, evaluates options, selects a payment method and completes checkout and the business optimises for that human browsing and deciding. In agentic commerce, an AI agent does this on the consumer's behalf: researching options autonomously, comparing suppliers, weighing trade-offs across cost, delivery and reliability and executing a transaction once its conditions are met.
This changes three things: how products get discovered, how checkout needs to work and what happens when a payment fails.
On discovery, AI agents can't recommend products they can't parse. If product data isn't structured, metadata isn't clean and the catalogue isn't machine-readable, a business is invisible to the AI layer that is increasingly mediating consumer demand. Structured product data functions as the new SEO, and it needs investment now, before AI-mediated commerce reaches a scale where catching up becomes difficult.
On checkout, when an AI agent transacts on a consumer's behalf, the flow can't rely on redirects, human-facing prompts and manually entered details built for a browser session. It needs to be programmatic, an API-first flow that lets the agent complete the transaction without human intervention at every step. Businesses with manual-only checkout, or payment systems that require human interaction at multiple points, will simply be skipped.
On payment failure, human-driven commerce gives a failed payment room to recover: the customer tries again, switches methods or contacts support. In agentic commerce, a failed payment is more likely to send the agent straight to an alternative supplier. Payment reliability becomes a condition of staying in the consideration set, not just a conversion metric.
Three investment priorities for agentic commerce readiness
- Structured product data. Invest in structured markup, clean product feeds and machine-readable catalogues. This is the infrastructure that makes products discoverable and evaluable by AI agents, and it also improves traditional search performance, so the investment serves current and emerging channels at once. The attributes that matter most for AI discoverability include clear product identifiers such as GTIN and MPN, detailed attribute data covering size, colour, material and compatibility, accurate pricing and availability signals, and structured review data. None of this is new. What's changed is how much it now matters in an AI-mediated discovery environment.
- API-first payment infrastructure. When AI agents transact on a consumer's behalf, the payment layer needs to be as programmable as the discovery layer. That means online payments infrastructure that supports programmatic transaction initiation, variable authorisation models, pre-authorised spending limits agents can operate within, and real-time settlement signals agents can use to confirm completion. This is why Variable Recurring Payments (VRP) and bank-based payment methods are gaining particular relevance here. They let agents act within consumer-set guardrails without re-authenticating on every transaction, which is exactly how agentic flows need to work. The rigid, card-based, per-transaction authentication model creates friction that agentic commerce will simply route around.
- AI-optimised content and agent interfaces. As AI agents mediate more commerce, marketing investment built purely around emotionally resonant human-facing content becomes less sufficient on its own. AI agents evaluate against structured criteria: price, delivery window, reliability signals, compatibility attributes. Businesses that also make their value propositions legible to AI, through structured data, clean API responses and reliable availability data, will be surfaced more often in agentic decision-making. This is still an early-stage discipline, but the direction is clear enough to act on now.
The compounding advantage of early investment
Enterprise leaders often find the mobile commerce parallel useful here. In 2012, most enterprise retailers ran desktop-optimised checkout experiences. Those who invested early in mobile-first checkout saw compounding returns as mobile commerce grew: higher conversion among mobile users, better data on mobile purchase behaviour and a capability gap that took slower movers years to close.
The same dynamic applies to agentic commerce readiness. The volume of AI-mediated transactions is small today, and the investment required to prepare, structured product data, API-first payment architecture and AI-optimised content, is meaningful but not transformative. The advantage of being discoverable and transactable by AI agents once that volume grows will be far harder for late movers to replicate quickly.
Between 31 and 34% of South Africans are already using AI tools in their shopping journeys. The businesses that prepare their product and payment infrastructure now will be the ones capturing that demand as it shifts from research assistance to autonomous transaction.
Stitch is building toward API-first payment infrastructure designed for programmatic payment flows, the foundation agentic commerce requires. To understand how your current payments infrastructure maps to these requirements, talk to our team. For more on what this shift means for payment systems specifically, read our explainer on agentic commerce.
FAQs
What is agentic commerce?
Agentic commerce is a model of digital commerce in which AI agents autonomously research, compare and execute purchases on behalf of consumers, operating within parameters the user has set. Rather than the consumer navigating every step, the agent handles discovery, evaluation and transaction initiation. Read more in our explainer on agentic commerce.
How many South Africans already use AI tools like ChatGPT for shopping?
According to the 2026 Stitch Consumer Payments Report, between 31 and 34% of South Africans are already active ChatGPT users, with use cases including research, financial planning and purchase decision support. Qualitative research found South African consumers actively using AI to compare products and check authenticity before buying.
What does API-first payment infrastructure mean for agentic commerce?
It means a payments system that can be triggered and completed through API calls without manual steps, redirects or human authentication at every stage. This lets AI agents complete transactions within consumer-approved parameters without friction.
Why does structured product data matter for AI-driven shopping?
AI agents can only recommend and transact with products they can parse. Structured product data, including clean feeds, Schema.org markup, accurate attributes and pricing signals, makes products machine-readable and evaluable by AI systems. Without it, products are effectively invisible to AI-mediated discovery, regardless of how well they perform in traditional search.
How do Variable Recurring Payments (VRP) support agentic commerce?
Variable Recurring Payments (VRP) let consumers pre-authorise a merchant to collect variable amounts within agreed limits, without re-authenticating on every transaction. This matches how agentic commerce needs to work, with agents operating inside consumer-set guardrails and executing transactions without friction.
Future-proof your payments with Stitch
