Ecommerce trends for 2025: Many early AI efforts will struggle to deliver results

This is the second article of our four-part series on the ecommerce trends we see on the horizon for 2025.

While Artificial intelligence (AI) is poised to transform everything in ecommerce from customer service to product recommendations, the implementation of these capabilities is just as important as the technology itself. As we look ahead to 2025, it is clear that AI will continue to play a pivotal role in shaping how retailers interact with consumers. But, not all AI implementations are destined for success. Many early efforts will falter due to overhyped expectations, misaligned strategies, and a lack of focus on customer needs. This article explores why these failures occur, goes  into specific examples, and provides actionable strategies to navigate the pitfalls of implementing AI in ecommerce.

The current landscape of AI in ecommerce

AI has quickly become a cornerstone of ecommerce innovation, with retailers racing to deploy tools that promise improved efficiency and customer satisfaction. AI holds a great deal of promise for ecommerce, from providing personalized product recommendations to smoothing out the checkout process to returning more relevant search results. If the technology lives up to this promise, it could have outsized returns for online retailers’ bottom lines.

Two primary approaches have emerged: AI agents and AI assistants.

Agents are autonomous systems designed to execute tasks on behalf of users, such as online shopping or price comparisons. These tools aim to streamline the buying process by handling everything from product searches to final purchases. For example, Perplexity’s AI shopping agent is one of the early entrants in this space, enabling users to purchase products right from their search and recommendation interface. Google and OpenAI aren’t far behind. However, these systems are not without limitations; there are still inconsistencies and limitations in their current iterations. Perplexity, for instance, occasionally fails to fulfill order requests even though it confirms them initially. 

The implications of these systems extend far beyond convenience. Agents have the potential to disrupt traditional ecommerce business models by reducing traffic to online storefronts and limiting data collection opportunities for retailers. Their success will hinge on overcoming challenges in execution and ensuring reliability across a wide range of scenarios, but they seem well-poised to make the leap.

Assistants, on the other hand, are AI tools like Amazon’s Rufus that focus on guiding customers rather than making decisions for them. Assistants are more conversational. Their job is to point you in the direction of products you need based on your “conversation” with them. Rufus demonstrates how conversational AI can enhance the shopping experience by providing personalized recommendations based on user interactions. This technology is already influencing how the company approaches product recommendations with traditionally important factors like “Amazon Choice” badges carrying less weight through Rufus. 

The distinction between agents and assistants lies in their goals: agents aim for independence, while assistants emphasize collaboration. Regardless, both models are driving significant changes in digital presence, pricing strategies, product categorization, and beyond. However, their development remains in its early stages, requiring extensive refinement.

Why many early efforts are destined to fail

While the potential of AI in ecommerce is immense, we believe that ultimate success will hinge upon how ecommerce companies implement this technology. Many early efforts are likely to fail due to several key factors:

The gap between technological hype and realistic application 

The initial excitement surrounding AI often leads to inflated expectations. Companies frequently promise groundbreaking capabilities that their AI systems cannot yet deliver. This creates a disconnect between what consumers anticipate and the actual performance of these tools. Consumers are made to expect that AI will be a magic wand that takes care of all their needs. The reality is that the technology isn’t there yet. Take Rufus, for example. Consumers are already expressing frustrations that the assistant is answering questions incorrectly or pointing them to irrelevant products. For now, the promise of agents that can flawlessly handle complex shopping tasks remains largely unfulfilled, leading to frustration among early adopters.

Overemphasis on technology rather than customer needs 

Most importantly, many companies have been guilty of prioritizing technological capabilities over user experience. Developers often focus on showcasing what their AI can do rather than addressing what customers actually want. This misalignment can result in tools that are technically impressive but fail to resonate with users. This can be seen by the sheer number of ecommerce companies across sectors and target audiences that have launched AI shopping assistants that have virtually the same capabilities - from eBay to Ikea to Michael Kors. This level of uniformity is a hallmark indicator of tech-forward innovation as compared to customer-focused innovation.

AI tools that fail to account for regional shopping habits or niche preferences may alienate significant portions of the target audience. Effective AI systems must bridge the gap between advanced functionality and intuitive usability.

The complexities of ecommerce AI 

Ecommerce presents unique challenges for AI systems, including understanding diverse customer behaviors and managing a vast array of products. At its core, ecommerce is a complicated matching process of products to shoppers. And each shopper is unique. One person, for instance, may prefer cheap clothes but expensive accessories. Or that same person may be shopping for a family member rather than themselves. That person may even be shopping for a one-time event, like a wedding, that differs drastically from their day-to-day shopping behavior. These complexities require AI to be highly adaptable and context-aware, which many early-stage systems struggle to achieve.

Training AI systems to recognize subtle variations in consumer preferences, handle conflicting inputs, and adapt to evolving market trends is a formidable task. These challenges can lead to delays, cost overruns, or outright project abandonment.

The critical role of customer-centric design

To succeed with AI in ecommerce, retailers must shift their focus from technology to the customer and understanding their needs. What do shoppers look for when navigating an online store? How do their preferences vary based on demographics, situations, or product types? Addressing these questions can help retailers design AI tools that align with real-world demands rather than theoretical possibilities.

Consider why a customer may come to your site and the types of products that they may be looking for. If a customer is looking to repurchase a bottle of all-purpose cleaner, a commodity product he’s likely bought before, an AI agent may be the perfect fit. He doesn’t want to spend time comparing 10 different types of cleaners or visiting 5 different sites looking for the absolute best deal. He may be perfectly fine outsourcing all of this work and allowing an agent to do the work for him. He just wants the cleaner to show up on his doorstep as quickly as possible.

During a different session, that same shopper may be looking for the perfect backpack for an upcoming camping trip. This is a differentiated product, one for which he’s more likely to invest time into to ensure that he gets the best option for him. This time, he may be willing to look at a range of options and consider the specific features. It’s a more expensive purchase, so he’s probably looking for the best balance of price and functionality. Here, an AI assistant is probably the right technology for the job. These two examples illustrate how important it is to match the technological capability to the use case.

AI systems that leverage detailed user profiles, historical data, and contextual insights can create meaningful interactions. Understanding the full context of not only the user, but also the specific intent of each interaction is what will unlock these magical shopping experiences. By building out your site with tools that build towards that reality, you will ensure that your money is well-spent with significant customer adoption precisely because that’s who you built it for.

Looking ahead: Balancing innovation with adaptability

Achieving success with AI in ecommerce requires a balanced approach that prioritizes both innovation and adaptability. The ecommerce industry is undergoing constant change. So, new AI systems must be able to evolve alongside changing customer expectations and market trends. Flexibility in design and implementation is essential for long-term success. Retailers should invest in modular AI architectures that can be easily updated and expanded as new technologies emerge.

As we move toward 2025, retailers should focus on building technological foundations rather than pursuing radical transformations. Our view is that ecommerce companies will need flexible, adaptable AI capabilities. Agents and assistants both have a place in ecommerce and there are likely tools that haven’t even been dreamt of yet. The key is to invest in foundations that help you build the AI capabilities of today while maintaining optionality for the future.

There are a number of activities that can support a flexible technological foundation. One is building out augmented, rich product data that will enable semantic search, incorporating additional information such as why a customer should buy one product versus another. Another is investing in systems that collect relevant customer data so that you can better understand your shoppers and provide smooth, personalized experiences. And, lastly, building out a search and discovery platform that marries this rich product information with robust customer data.

The critical part here is that all these systems work together. If your search engine is distinct from your recommendation engine which is distinct from your personalization engine, you will provide disjointed experiences for your customers. In the crowded ecommerce market, anything less than an intuitive, effective shopping experience could drive potential customers away.

Conclusion

The future of AI in ecommerce is both exciting and uncertain. While many early efforts are likely to fail due to technological hype, misaligned priorities, and implementation challenges, retailers that prioritize customer needs and embrace adaptability have a better chance of success. In 2025 and beyond, focusing on incremental improvements, leveraging customer-centric design, and fostering collaboration, will enable companies to unlock the full potential of AI and stay ahead of the ecommerce curve.