Agentic commerce: Designing for an economy where software does the shopping

More than half of consumers (58 percent) are already comfortable using genAI-powered search and delegating discovery to large language models (LLMs). In the first three months of 2026, retailers saw AI-driven traffic to US retail sites grow 393 percent year-on-year.

This surge is not slowing. According to Adobe’s latest AI traffic trends report, AI-driven retail traffic grew a further 138 percent year-on-year as of May 2026, signalling a sustained and structural shift in how customers discover brands.

More importantly, this traffic is higher quality: Adobe reports that AI-referred visitors are more engaged, spend more time on site, and are less likely to bounce than traditional channels.

From discovery to decision making, the agentic commerce marketplace creates a further paradigm shift in how people buy. Today, machine-readability and fulfilment matters more to agents than storytelling and interface, demonstrating how brand is moving from a differentiator to a signal.

So how do retailers retain strong brand differentiation and redesign for a marketplace where it’s software, not humans, doing the shopping?

Retailers must build architectures and frameworks through disciplined integration where data, pricing, fulfilment, and governance are reliable and clear to AI agents. It’s about a system shift where trust needs to be earned through what works, not just what’s said.

Build an agent-ready data foundation

Being discoverable in today’s marketplace means building an agent-ready infrastructure based on a data foundation which is clean, unified, and machine-readable.

Retailers need to apply transparency and standardisation across product master data to standards that agents expect. This means exposing data through APIs so agents can access live, trusted data on availability, delivery, and pricing in real time. Enriching the quality of the data should also be prioritised with machine-readable structured attributes, including materials, compatibility, and dimensions. This will avoid agents selecting certain products, the most expensive for example, just because the data has been correctly structured for machines.

Multiple protocols are being developed to ensure data structures are defined by industry standards. These include Model Context Protocol (MCP) for the data and API layer, Agentic Commerce Protocol (ACP), Agent Payment Protocol 2 (AP2) for application-to-payment system, and increasingly Universal Commerce Protocol (UCP) for application-to-application interacting with commerce systems. Even as these protocols evolve, retailers need to understand where they best sit, for example UCP for top-of-funnel discovery intent, versus ACP for deeper-funnel conversational intent. Retailers should then align this view to overall business strategy to ensure products are showing up in the right way.

Prioritise data to influence generative engine optimisation

Consistent product identifiers and standardised master data are a critical single source of truth not just for the ‘what’ of product data, but the ‘why’. Consumers using answer engines are changing their behaviour from product-based queries to broader goal-based ones, supported by agents that now conduct ‘missions’ across multiple retailers and platforms – bundling, substituting, optimising, and executing decisions.

Retailers should start by understanding which segments and product categories they want to influence through generative engine optimisation (GEO). An audit using any of the GEO ranking tools will help retailers understand how their current products show up in queries around key use cases that customers would select their product for. And a further audit of a retailer’s web properties will help retailers see where the content on their products aligns to the queries being filled in. Consistent tagging of content and metadata across all of a retailer’s web properties is crucial here as LLM’s don’t work well if retailers refer to the same product in different ways across websites.

In a world of finite resources, it’s not about perfect data, but about a ruthless focus on the data that moves decisions (leaving the rest behind). Agents don’t evaluate everything equally, and neither should retailers. The real advantage lies in identifying the handful of signals that drive choice for the most valuable customers, and systematically outperforming competitors on those dimensions.

Redefine conversion in zero-click journeys

As AI agents increasingly manage the shopping journey end-to-end and drive choice, conversion can occur without a customer ever interacting with a retailer’s owned channels. This fundamentally reshapes what ‘conversion’ means.

Adobe’s analysis shows that AI-driven traffic is not just discovering products, but converting more effectively – with AI-referred retail visitors converting at a rate 54 percent higher than non-AI traffic.

These visits are also more valuable, generating 53 percent more revenue per visit than non-AI traffic – a complete reversal of the dynamic seen just 12 months ago.

Retailers are no longer optimising for clicks or sessions, but for selection within an algorithmic decision process.

Higher conversion efficiency isn’t a trade-off to manage. Instead, it signals that the rules of influence have changed. As agents streamline the shopping journey, they distil decision-making into a defined set of signals. So the opportunity for retailers is to engineer those signals deliberately shaping how products are evaluated, surfaced, and prioritised. Here we see brand shift from something experienced in a journey to something embedded in the criteria that drives choice.

This shift also extends into how transactions are executed. Payment and fulfilment infrastructure is evolving quickly, from agent-held wallets to autonomous settlement. Early moves such as OpenAI’s Instant Checkout highlight how the market is maturing where initial attempts to capture value within closed ecosystems are giving way to a recognition that trust, interoperability, and scale require coordinated infrastructure. The ecosystem is now beginning to align around shared protocols and capabilities across payments, identity, and fulfilment. So, for retailers, advantage will come from aligning with, and helping shape, the infrastructure that agents rely on. Those who ensure their checkout, payment, and fulfilment logic are interoperable with emerging standards will not only remain visible at the point of decision, but influence how trust and seamlessness are built into agentic commerce.

This mindset shift in conversion redefines measurement too. Rather than focusing solely on fragmented attribution, retailers must understand how they are evaluated within agent decision processes. Signals such as recommendation visibility, price competitiveness, and fulfilment reliability become a new layer of performance data – shifting measurement from tracking journeys to strengthening how brands are chosen.

How retailers can win in an era of agentic commerce

Success for retailers in the delegation economy is in making their brand the one that agents, as well as humans, trust. This will come about through integration of data, pricing logic, fulfilment operations, payments, and governance.

Experimentation and a continuous beta-mindset is key to success as agentic commerce grows at pace. While the infrastructure is still being built by the payment providers to make agentic commerce happen more seamlessly and standards still need to mature, retailers should be treating experimentation as a core operating discipline as customer behaviours and technology adjust in real time.

Retailers should look at their data infrastructure and understand what they need to engineer into their brand interfaces to allow agents to do their job. Ultimately, it’s about a refresh of targeted data governance rather than a rip-and-replace, building on what they have currently, and optimising it for a new delegation economy reality.