Advanced Commerce Experiences: AI Agency Across The Retail Customer Journey

[Note: This article was originally published on Forbes on January 7th, 2025.]

In my last article, I looked at why organizations should consider giving artificial intelligence (AI) more agency and autonomy to positively impact customer support and overall customer experience (CX).

I want to take things one step further and examine how autonomous AI agents can result in exceptional CX in the e-commerce journey.

With enough agency, AI agents have the power to replicate the dynamic environment of a physical store within the digital world. Instead of being faced with static web pages, customers could engage with intelligent assistants that cater to their needs and preferences, much like a human salesperson in a brick-and-mortar store.

Let’s look at the four distinct types of AI agents and their potential impact in this segment:

Agent #1: Generic Information Provider

AI tools already excel at providing generic information. Traditionally employing machine learning (ML), these solutions can inform customers about common scenarios and answer queries about particular needs based on pre-loaded and aggregated data.

Internally, they help to efficiently segment customer categories, inform the most common product basket compositions, forecast sales, predict when a machine will fail and much more. A generic information provider can, for example, analyze historical open rates and suggest the best times of day to send a marketing email to a specific demographic.

Most organizations that have not started using generic information providers will benefit from the current optimism and bullish outlook on AI, which could help them gather support for using it and roll it out.

Agent #2: Specific Information Provider

For online commerce customers, specific information providers can act as virtual shopping companions, offering detailed product descriptions and personalized recommendations. For example, pairing tailored suggestions with explanations of why they’re a good fit in real time is an impactful experience and can greatly influence a customer's decision.

This type of agent can also analyze a customer’s search query in detail to bring up products that match very specific wants, making the shopping process more satisfying. Plus, they are capable of performing post-sale follow-ups with customers to gather reviews or keep the brand top of mind.

Internally, a specific, targeted, contextualized information provider may inform the organization that a particular user is likely to buy a specific product if they receive a direct marketing email.

Many brands are starting to mix traditional AI with GenAI to make shopping smoother and more enjoyable in these ways. Still, until these agents have agency, they depend on the customer to take action.

Agent #3: Actionable And Transactional Agents

These agents are the first level of agency that I see playing a key role in redesigned commerce journeys. They can act upon insights from the first two types of agents with an intent to drive toward specific outcomes. Here’s a use-case example:

  • First, a generic information provider might suggest that a marketing email sent to a certain demographic on Monday afternoon is best.

  • Next, a specific information provider will share that a particular user is likely to buy a product if receiving a marketing email.

  • The actionable agent can then plan and perform an experiment based on this information, initiating a targeted, personalized email campaign, tracking its success and tweaking the strategy as needed.

Actionable agents could also modify prices, promotions, how products are showcased and more based on information provided by other AI agents. Still, organizations will need to give them permission to do so and have the guardrails to do it safely.

Agent #4: Exceptional Agents

The first three categories are based on known and pre-planned moments and actions of the commerce journey. In the future, agents will be able to step in to solve exceptional requests and find novel paths or actions to optimize objectives.

Still, this is currently a distant goal that would only be acceptable once results from the previous three types bring great value to organizations. The big takeaway here is that exceptional AI agents will be able to imagine, create and offer experiences that no human would have ever thought of, and in that lies the excitement and uncertainty of this technology.

The Road Ahead: Overcoming Challenges

As more organizations consider giving AI agents the autonomy to act throughout the commerce customer journey, naturally, they’ll face a few challenges.

Perhaps the most significant hurdle is shifting the user experience (UX) mindset. The goal must be to move away from static, pre-defined online experiences and encourage customers to embrace personalized, automated, unique interactions.

Of course, full personalization comes with privacy and compliance concerns—just because we have access to private data doesn’t mean we (or an AI agent) should always use that data to take action. To align and comply, organizations will need to improve their data access management and ensure that AI agents can only tap into the information they need to decide on an action.

Establishing effective autonomy guardrails plays a huge part in this process. Granting AI agents the ability to execute transactions on behalf of customers or the organization requires safeguards to prevent unintended consequences. As autonomous agents become more common, organizations will need to create policies for situations that aren't currently covered in their guidelines. For instance, training AI on scenarios where humans already know what to do or not do without being told, such as unwritten rules or common sense.

Another major challenge is finding the right balance between letting autonomous agents handle unique situations on their own and managing the risks and liabilities that come with it. An organization or team can't just pass off responsibility for mistakes to a machine. It will be crucial to include risk management strategies in how these agents make decisions to keep their actions within the organization's acceptable risk limits.

The shift toward AI agency and autonomy is more about organizational adaptation than technological innovation. We know that the future of commerce is undeniably personalized, dynamic and intelligent; it’s only a matter of embracing AI as the driving force behind the transformation.