Even online shopping can be painstaking, given the manifold selection of products and e-commerce stores. If the mere thought of browsing through myriads of online shops’ inventory items, comparing prices and logistic options leaves you feeling exhausted – agentic AI is on its way to the rescue mission.
AI agents are popping up more and more frequently in the news headlines. And yet, most businesses are still cautious about or even unaware of their capabilities and potential. However, some companies dealing with online sales are already experimenting with agentic AI. Can these early explorations crystallise into something tangible? Let’s put on our thinking caps.
What Is Agentic AI?
While artificial intelligence per se is not a novelty anymore, the concept of agentic AI is often vague for people in the street. It refers to artificial intelligence systems designed to act as autonomous agents, needing minimal human oversight. Unlike traditional AI tools (like those for language processing or image recognition), AI agents don’t just process inputs and produce outputs, operating under user guidance with a fixed set of instructions. Instead, they can set goals, make decisions, and act independently in dynamic environments. Having an ultimate mission in mind, agentic AI tools can self-direct and initiate required actions, changing course if needed in response to the altering circumstances.
AI Agents in E-Commerce: Potential vs Existing Applications
Agentic AI is especially advantageous for tasks requiring adaptability. When it comes to e-commerce, AI agents can particularly excel in scenarios that require proactive decision-making, such as
- dynamic pricing optimisation in response to market conditions;
- creating highly targeted promotions and marketing campaigns;
- managing inventory and supply chains;
- refined and adaptive product recommendations;
- risk management and fraud prevention;
- ordering of miscellaneous items for a specific event/purpose;
- personalised shopping assistance;
- intelligent search with tailored results;
- alerts about deals occurring in real-time and purchases at the best price;
- proactive customer support.
However, few of these use cases are currently being explored in real-world settings. While ubiquitous artificial intelligence tools are expansively utilised in many scenarios mentioned above, most current e-commerce applications don’t fully meet the criterion of full autonomy, self-generated objectives, and adapting to new and unforeseen situations. Today’s systems are reactive, efficiently automating well-understood tasks rather than operating as independent agents.
Nevertheless, we can already see that retail giants are eager to investigate the next level of innovation and convenience AI agents bring. They try the agentic AI waters in baby steps.
Amazon Is Testing Buy For Me Agent
Amazon is one of the most advanced companies in terms of AI use. It leverages and offers its customers proprietary AI assistants, AI-powered logistic and supply chain solutions, checkout and payment AI tools, personalised AI-driven recommendations, image and store listing generation, AI-powered security tools, AI-generated review analysis, and more.
It’s no wonder that Amazon is also one of the earliest AI agent adopters in the retail space. The e-commerce giant has recently started testing the “Buy for me” service powered by agentic AI. It leverages two different AI models to navigate external websites autonomously, select requested products that are absent in Amazon stores, and complete purchases from supported third-party resources on behalf of customers. Although the exploration still happens on a limited scale, upon the initial feedback from users, the feature might expand its reach.
Understandably, Amazon has chosen a limited beta testing model for such a service. The new shopping mode requires not only technical but also ethical fine-tuning. After all, some earlier studies have shown controversial customer sentiment towards AI use for payment purposes.
While the product discovery part might generate great enthusiasm, some shoppers may feel less confident about automating the final transaction step. They might question whether the AI tool has chosen the best option or considered all essential factors related to delivery and payment. While some customers will feel alleviated, others may want to gain more control over the purchase process. Only time and customer feedback will tell how to strike the right balance between facilitation and empowerment.
Amazon’s Autonomous Fulfillment Robots Are Getting More Independent
Besides outright testing of agentic AI tools for end customers, Amazon has continuously enhanced its internal logistics processes, using robots that have gradually evolved into semi-agentic smart systems.
The firm utilises a variety of warehouse robots, mainly developed by its own robotics arm, to move goods, streamline picking, and optimise inventory flow. Some of them already display certain features that are characteristic of what we describe as AI agents. For example, Amazon’s robot systems can make real-time decisions about navigating busy warehouse environments or where to store items based on product demand forecasts and item interrelatedness rather than fixed locations. These warehouse assistants can adapt their routes based on traffic, obstacles, or task priorities. They also coordinate actions with other robots and human workers using decentralised communication systems. Besides, they can prioritise tasks dynamically to some extent (e.g., speeding up urgent item delivery, reassigning themselves if a path becomes blocked, etc.).
Alibaba Agentifies Its Virtual Assistant
Alibaba’s recent upgrade to its Quark app illustrates advanced functionalities that allow the system to autonomously perform tasks on behalf of users, bringing it one step closer to agentic AI tools. In its latest update, Quark moved from a simple search engine to a smart system able to engage in a dialogue with users to understand and fulfill their needs better, autonomously perform actions such as booking travel arrangements or generating images based on user prompts, and providing thoughtful, context-aware responses to complex queries. This proactive and adaptive behaviour aligns with the principles of agentic AI. So far, the tool can enhance the shopping experience only when it comes to travel planning. However, media reports suggest that later this month a new version of Alibaba’s flagship artificial intelligence model that powers the Quark app – Qwen 3 – will reach the market. Who knows what new opportunities it may bring!
Startups Focused on Agentic AI in E-Commerce
While large e-commerce corporations attempt to incorporate agentic AI tools into their existing complex workflows, a range of startups emerge, making AI agents the backbone of their offerings.
Fermat Commerce Develops Influencer and AI-Driven Shoppable Content
Founded in 2021, Fermat has recently raised $17 million in seed funding to create embedded commerce experiences powered by agentic AI. The startup enables shoppable content placement directly inside blog posts, videos, and social media without redirecting users to traditional product pages. Its marketing technology provides context-aware product recommendations, autonomously deciding which products to showcase, prioritising those based on intent, engagement, and content themes. Moreover, the recommendations can update in real-time as trends shift without any manual changes.
Next to the shoppable content placement, Fermat tools can act as a mini shopping concierge. Its AI agent handles product selection, pricing, display, and checkout options when a customer hits the “Buy” button. It can even proactively customise the flow based on whether the client is engaging with content on mobile, iOS, logged into Shopify, etc.
The firm’s clients significantly improve their conversion rates, average order value, and return on attention, while reducing cost per acquisition or action.
Giftpack.ai Transforms Global Corporate Gifting and Rewards Systems
Operating since 2020, Giftpack.ai offers a personal gifting agent that automates the selection and sending of personalised gifts in B2B (corporate gifting, client retention, etc.) settings. Customers should just give it a few inputs, like recipient info, preferences, relationship type, occasion, budget, and the agent will take care of the rest: gift selection, customisation, purchase, and even delivery. Giftpack automatically places the order, handles customisation details (card, message, wrapping, etc.), tracks delivery, and may provide a follow-up with satisfaction surveys or gifting metrics, if needed.
When integrated with HR platforms, Giftpack’s AI agent gains automated access to employee data, enabling the tool to register key moments, create gifting occasions without HR initiation, and finally select and send gifts automatically, using employee preferences.
Apart from this, the AI giftgiver can be embedded into the corporate reward points system, offering a points-to-gift redemption portal for employees and assisting managers with auto-suggestions or distribution of curated rewards. The tool can optimise rewards for cost, availability, and personalisation even within a fixed budget.
Perplexity’s Shopping Assistant Shows Agentic Traits
Perplexity AI, a startup building search engine powered by artificial intelligence, has recently launched a virtual shopping assistant that illustrates certain agentic behaviour. The “Buy with Pro” agent goes further than simply showing users search results. The AI tool autonomously drives the purchase process from discovery to transaction.
It filters available offerings based on users’ goals, compares, and recommends most suitable products without a user hand-picking them from a list. Furthermore, with a user’s approval, the Pro agent can make the purchase itself using the customer’s saved billing/shipping info. Although the assistant may lack some adaptability characteristics and still require manual customer approval of the selected items instead of possessing full decision-making autonomy, “Buy with Pro” service certainly is more than a typical chatbot or search engine.
Challenges for Agentic AI Implementation in E-Commerce
Agentic AI systems offer plenty of new opportunities for marketplaces and merchants. However, their implementation requires some technical upgrades as well as solving some ethical dilemmas.
To begin with, who should decide on the level of autonomy for AI agents? There are no regulations yet that oblige providers to take responsibility for possible financial risks associated with AI use. At the same time, if an AI system misinterprets a customer’s preferences or makes an erroneous decision based on incomplete data, the business would be the one to incur financial losses or reputational damage.
Bearing that in mind, many business sectors are already establishing certain guidelines for AI use, oversight and transparency. E-commerce companies should probably do the same to make sure their customers are protected.
Determining the appropriate level of human oversight for agentic AI systems in online shopping requires delicate balancing. While routine product recommendations might need minimal oversight, high-stakes tasks like automatic purchasing or high-value transactions may require closer human supervision. Perhaps, in some cases, customers can opt in or out of fully autonomous experiences to adapt to various customer’s risk-tolerance and trust levels. For instance, an auto-buy feature might be accompanied with an additional “confirmation” toggle if a purchase exceeds a certain amount. To maintain public trust, critical decisions should still have a human review or verification layer, while AI rationale should be accessible and transparent. Finally, users might be empowered with options to override or customise some agentic AI actions to feel in control rather than helpless.
Summary
Agentic AI in e-commerce is all about making shopping smarter and easier, letting automated systems act even more independently. Instead of just crunching data or answering simple queries, these AI agents set their own course based on user or corporate goals and make decisions on the fly. Think of it like having a personal shopping buddy that not only recommends products, but also deals with dynamic pricing, seeks personalised promotions, and even handles checkout. Some early examples include Amazon’s “Buy for Me” service or Perplexity’s “Buy With Pro” feature. While most current tools rely on preset rules, the new wave of agentic AI adapts to changing market conditions and user preferences with minimal human oversight, pushing e-commerce from being reactive to proactive, creating more efficient, curated, and convenient shopping experiences.
