Agentic AI in ecommerce: How merchants can stay relevant

For merchants in the retail landscape, the rise of AI agents will fundamentally reshape how they operate and serve consumers. Sameer Verna from PaymentGenes walks through the latest developments in agentic commerce from the perspective of merchants and the strategies they can embrace to stay relevant.

The rise of AI agents marks the next major inflection point in commerce as we move towards an even more sophisticated selling environment. AI agents interact on behalf of shoppers, engaging with thousands of stores and catalogues in a matter of seconds to find the best deal, bringing unprecedented speed and convenience to the consumer.

On the other hand, merchants face the risk of losing brand loyalty, proximity to customers, or worse, seeing their products and services reduced to commodities in the sea of AI-driven recommendations.

Agentic Commerce User Journey

Whether Agentic commerce will help unlock a new and powerful acquisition channel for merchants boosting sales or end up cannibalizing them relies on intricate technical underpinnings, including API endpoints, data pipelines, and cross-platform connectors that determine how agents access and transact with merchant inventories.

Nonetheless, it is certain that the customer journeys will be radically compressed compared to traditional e-commerce. There are some key differences between human and machine behaviors that will impact buying journeys and consequently conversions:

Information absorption & processing
Machines can consume and compare vast amounts of product data. Detailed comparisons between brands, product quality, warranties etc. can be made at the click of a button.

Speed and Time
Machines can read at the speed of light and be active in hundreds of funnels simultaneously 24/7/365. Example: when booking flights, agents can simultaneously test full checkout flows across multiple airlines to surface the true total cost (including add-ons like baggage and seat selection).

Insensitive to intangibles
Human shopping behavior is directed by feelings around a brand, the quality of the product imagery or the website experience. These are all poor proxies for product quality. Agents are likely to ignore these intangible proxies and more likely to leverage processing power to assess the actual product quality.

The Promise and the Challenges

According to McKinsey & Company research, US B2C retail market alone could see up to $1 trillion in revenue from agentic commerce by 2030, with global projections estimated between $3 to $5 trillion. The impact of agentic commerce is expected to be similar to the web and mobile commerce revolutions, but there are a range of challenges pertaining to its enablement and merchant inclusion given the current e-commerce infrastructure.

Some of the key challenges yet to be navigated to ensure a smooth agentic commerce ecosystem:

Merchants cannot differentiate between AI agents and bots
The current merchant infrastructure and fraud setups will block out a large share of legitimate AI agents, identifying them as bots.

Merchants risk losing consumer proximity and data
As AI agents increasingly intermediate purchases, they consume detailed product and pricing data while owning the customer interaction, reducing merchants’ direct visibility into consumer needs and threatening long-term brand loyalty and differentiation.

Fragmented payment landscape
Payment methods and providers are fragmented, standardization is needed to access alternative payment methods (especially relevant in Europe).

Marketplaces risk disintermediation
Aggregation platforms risk losing their relevance as AI shopping agents develop the intelligence to independently discover, compare, and recommend products and sellers.

Long tail at the risk of lockout
Tens of thousands of merchants operating with HTML only storefronts with no API connectors, catalog feeds or checkout endpoints face the risk of being locked out of major agentic marketplaces.

Compliance & risk
Merchants will have to respond to new fraud signals while allowing genuine transactions. Agentic commerce could potentially expose merchants to raw payment data leading to PCI risk.

Operating Models and Use-cases

Agentic commerce remains in its nascent state with various operating models emerging. Currently most protocols enable a live product search and real-time user authentication for payments in an ‘assisted checkout’ mode. Future use cases will allow for fully delegated and autonomous transaction authentication with verified user consent.

In a fully autonomous payment ecosystem, the shopping agent executes transactions under a digitally signed authorization that specifies parameters such as budget, timing, and payment methods enabling secure, human-out-of-loop commerce.

Embedded payment functionality in AI chatbots is transforming digital transactions. This is most visible through seamlessly integrating secure payment flows into AI chatbots. Consumers can leverage AI agents for a range of use-cases. The most common one use-case today is product and deal discovery by aggregating pricing data across multiple websites or within a single merchant’s catalog, pre-populate shopping carts and then await manual authorization to complete checkout.

But agentic commerce has a diverse and wider list of use cases. Some of the most emblematic use-cases include concierge services for shoppers, complex multi-city travel bookings, or acting as a targeted purchase assistant with pre-defined parameters.

In corporate environments, procurement agents may enforce spending policies by using virtual cards with predefined limits. They then automatically trigger purchase orders once inventory thresholds are reached, processing invoice payments instantly against approved budgets.

Enabling Agentic Commerce

Various ecosystem players including big tech, AI firms, PSPs and card schemes are rushing to create protocols and adapt the current infrastructure to cater to agentic commerce facilitation. The different protocols may cause confusion amongst merchants regarding the use case and relevance but most of them are solving different problems along the agentic commerce value chain.

While Mastercard’s Agent Pay and Stripe’s Agentic Commerce Protocols create programmable tokens to facilitate transactions, Google’s Agent Protocol 2 (AP2) creates cryptographic digital signatures to help merchants verify user intent. Visa’s Trusted Agent Protocol addresses a critical problem of distinguishing between a legitimate AI agent versus malicious bots using cryptographic signatures.

An overview of some of the recently announced protocols:

Common security approaches include:

• All protocols help increase visibility and verify user intent typically using cryptographic signatures or programmable tokens for transaction authentication
• Multiple authentication layers (biometric, Passkeys, 2FA) across traditional network protocols 
• Non-repudiable audit trails for dispute resolution and compliance 

At a high level, there are 3 main implementation models that merchants will adopt to enable agentic commerce. The first, and the most common one for merchants that have limited operational capacity and resource constraints would be to use light adapters and hook into current gateways.

The other 2 options are more complex and require a dedicated agentic commerce strategy, either using a hybrid approach separating the agent traffic and protecting the existing human traffic. The last one entails fully offloading the agentic traffic by creating new catalogues, agent specific landing pods to optimized engagement and conversions.

Creating Differential Value in Agentic Commerce

While enabling agentic commerce may be a defensive move for most merchants, the ones that can facilitate this transition quicker and position themselves as not just enablers, but leaders of the change stand to gain a competitive advantage. There is a range of strategies merchants can leverage based on their size, risk appetite and current e-commerce infrastructure:

Personalized offers based on demand tracking
Agentic commerce will enable merchants to collect a much wider range of data on user demand. Package deals and personalized offers based on user profile and browsing trends will boost loyalty and conversion.

Smart inventory and pricing
Richer SKU-level analytics can enable merchants to track precise product demands and fine tune their inventory and pricing based on market trends and demand signals.

Concierge service
Larger e-commerce merchants could integrate an in-site concierge to help consumers find precise product offers and nurture clients to increase retention in both pre and post sales stage.

Agentic Landing Pads
An optimized space for agents to interact with the merchant website as well as creating specific mandates to ease in hand-offs from agent to human, providing context of their search and the possible next steps.