For retailers, the next phase of digital commerce is unlikely to be about building another channel. Most likely, it will be about making sure their existing infrastructure can transact with machines.
That is the emerging challenge as artificial intelligence (AI) moves from helping consumers research products towards potentially making purchasing decisions on their behalf.
Hayley Fisher, MD, Australia and New Zealand at Adyen, says retailers should not wait for consumers to become fully comfortable with agentic commerce before preparing for it. The foundations are more familiar than the terminology might suggest: Accurate product data, connected systems, real-time inventory and pricing, and payment infrastructure capable of supporting new ways of transacting.
“Retailers need to make their products machine-readable” is the core of Fisher’s advice. That means standardising product feeds and ensuring information such as pricing, availability and policies can be accessed and interpreted accurately by AI agents.
The same principle applies to the transaction itself. Retailers will need APIs that support automated checkout processes, including session creation, order validation, and fulfilment.
“The shift matters because AI-assisted shopping changes where and how the customer interacts with a retailer.”
Payments become the connective tissue
In a conventional online journey, the website provides much of the continuity between discovery, browsing, login and checkout. As shopping increasingly moves to conversational AI interfaces, that digital storefront may become less visible.
Fisher sees payments becoming an important thread connecting the customer to the retailer in this environment.
“Tokenisation and delegated payment credentials could allow transactions to take place without requiring consumers to repeatedly enter payment details or navigate a conventional checkout. At the same time, retailers will need new approaches to authorisation and fraud prevention.
“That creates a particularly important challenge: Distinguishing between an automated agent acting with a customer’s permission and an unauthorised bot.”
Payment security therefore needs to evolve alongside the shopping experience, with risk systems that can assess machine-led transactions at speed rather than treating automated activity as inherently suspicious.
The challenge is already significant. Adyen’s research found that 88 per cent of Australian merchants were targeted by or fell victim to payment fraud in the past 12 months, while 24 per cent identified payment fraud and chargebacks as an ongoing business challenge.
Agentic commerce adds another layer of complexity to a problem retailers are already managing.
AI exposes the weaknesses of fragmented commerce
Here unified commerce becomes more than an omnichannel aspiration.
An AI agent needs an accurate answer to basic questions: Is the product available? At what price? Can it be delivered? Can it be collected from a store? What loyalty benefits or customer history are relevant?
If those answers sit in disconnected systems, the retailer risks giving the AI agent inaccurate information or failing to seal a sale at the point of purchase.
Fisher argues that unified commerce provides the underlyingntory, pricing, customer identity, loyalty and order history across channels allows both humans and machines to interact with the same retail operation
“The alternative is increasingly problematic. Fragmented systems can result in mismatched inventory, failed transactions, duplicated orders or delayed responses – precisely the friction an AI-led shopping journey is intended to remove.”
Adyen’s research found that 97 per cent of merchants had experienced payment performance issues over the previous 12 months. For consumers, the consequences are tangible: 62 per cent said payment errors negatively affected their perception of a business, while 24 per cent said they abandoned a purchase altogether.
Trust will determine how quickly agentic commerce develops
For now, agentic AI technology is advancing faster than consumer confidence.
Adyen found that 64 per cent of Australians have used AI assistants to help with shopping, but the same proportion are uncomfortable allowing AI to complete a purchase on their behalf. Only 26 per cent said they were comfortable with it.
For Fisher, this gap makes control and transparency critical.
“Consumers need to understand what an AI agent is authorised to do, how much it can spend and when human approval is required,” she explains. “Retailers will also need clear processes for resolving transactions when an agent misunderstands an instruction.”
That could mean spending limits, merchant restrictions, explicit approval triggers and authentication mechanisms that make it clear when an AI agent is acting on a customer’s behalf.
The issue extends beyond payments. More than half of consumers surveyed – 56 per cent – said they find purchase or browsing tracking invasive. As AI becomes more involved in shopping, retailers will need to balance personalisation with clear boundaries around how customer data and permissions are used.
Preparing for a fragmented future
Fisher advises retailers not to wait for one dominant AI platform or protocol to emerge.
The more practical approach is to build flexible infrastructure that can accommodate multiple interfaces and standards.
An agentic-ready retailer, she suggests, will have a composable technology stack, unified commerce capabilities and APIs that can serve both human customers and machine agents. “It will also need governance mechanisms capable of authenticating authorised agents, enforcing spending mandates and managing risk.”
“That does not necessarily mean rebuilding the entire technology stack; rather, it means examining the existing architecture to ensure the retailer’s data, commerce and payment layers are sufficiently connected and adaptable to support new ways of discovering and buying products.”
That matters because AI adoption is already widespread among merchants. Adyen’s research found that all surveyed merchants were using AI in some capacity, most commonly for customer experience, product discovery and customer support.
The next step is moving from AI that assists the retailer or shopper to AI that can participate in the transaction itself.
Fisher says retail leaders need to avoid the misconception that agentic commerce is primarily a technology-platform decision. “The more fundamental question is whether the business can provide accurate information, connected commerce and trusted payments wherever the customer – or their AI agent – chooses to shop.
“Retailers best placed for that future will be the ones that make their own systems ready for whichever platform comes next,” she concludes.
