What agentic commerce means for retailers when AI agents start doing the shopping

In the not-so-distant future, consumers may no longer need to tap “Buy Now” at all. Instead, autonomous AI agents will make purchasing decisions, execute payments, and complete checkout processes. This shift from human-initiated shopping to agent-led commerce may mark a fundamental transformation in the way retail operates.

In this interview, Ravi Kumar shares his thoughts on how agentic AI will influence the customer journey, the challenges and tasks retailers will face, and what they are already doing to address them.

Ravi, how does agentic AI differ from traditional AI and ML solutions? In what ways might it be considered even more transformative?

Ravi Kumar: Traditional AI and ML models are more focused on forecasting based on their prior learning. Usually, we train ML models on historical data and then perform inference to predict future values.

With agentic AI, it's about having a virtual entity making decisions based on its knowledge and reasoning capabilities. Agentic AI solutions employ very sophisticated reasoning and can break down complex goals into sub-tasks, evaluate multiple options, and make goal-driven decisions. Agentic AI solutions can learn from their interactions and outcomes, continuously improving their decision-making over time, often leveraging Large Language Models (LLMs) as orchestrators for this process. It's about a continuous cycle of perception, reasoning, planning, action, and learning, often orchestrated by an LLM. And then there are tools like LangChain that allow LLMs to interact with external apps and data sources, transforming them from just text generators into proactive decision-making agents.

Agentic AI is very transformative compared to traditional AI, as it drives accelerated innovation and research. It can run simulations, analyze vast patterns, and make very quick business decisions.

How could agent-based AI used by retailers change the customer journey?

Ravi Kumar: Agentic AI can make shopping experiences very personalized. It can make customers feel like a real person is helping them purchase things. Instead of waiting for a user query, an agentic AI system might proactively suggest products based on prior purchases and life events. Agentic AI can dynamically offer personalized discounts or bundles based on a customer’s real-time behavior and likely conversion point. It can also provide post-purchase support, including reordering, handling product issues, suggesting accessories, and assisting with product setup.

If AI agents begin making autonomous purchasing decisions on behalf of consumers, what impact will this have on retailers?

Ravi Kumar: With the use of AI agents in purchasing decisions, many things will change for retailers. AI agents are very good at finding the best deals, which will drive even more intense price competition.

Retailers will have to embrace AI agent marketing – meaning they’ll need to adapt their strategies to make their products more appealing and accessible to AI agents acting on behalf of consumers. They will have to rely heavily on improving the quality of their products and product reviews. Retailers need to maintain high-quality inventory data and make it accessible to agents for better recommendations.

Alongside these challenges, there are also opportunities. This will open new customer segments and level the playing field for both large and small retailers. With AI agents, retailers will gain unprecedented insights into actual customer preferences and decision-making logic, leading to better product offerings.

Are there already concrete initiatives or pilots in the retail sector exploring agentic commerce?

Ravi Kumar: Yes, there are many retail organizations running concrete initiatives and pilot programs to explore agentic commerce. I read an article a few days ago about Walmart’s initiatives, and it seems like they are working on developing various agents. Many other retailers are actively experimenting and investing, recognizing the immense potential of this evolving landscape.

Are you developing agentic AI at Dollar General?

Ravi Kumar: Yes, Dollar General is a leading retail company, and we are also exploring agentic AI for various solutions such as conversational AI, knowledge AI, and supply chain management bots. The vision is to build separate agents for each domain and then consolidate them at the organizational level to have one complete agentic AI solution under one umbrella.

What new business models might emerge around agentic commerce?

Ravi Kumar: As per my understanding and knowledge, it will surely lead to the emergence of new business models. Subscription-based shopping agents are a highly probable and already emerging business model. At the same time, agentic AI will also lead to significant shifts in existing business models.

What are the most pressing data and cybersecurity risks associated with agentic AI that do financial transactions autonomously?

Ravi Kumar: The autonomous nature of agents presents significant risks when it comes to financial transactions. Many AI models are black boxes; they can make decisions and perform fraudulent transactions in large numbers. In a highly interconnected financial system where multiple agentic AIs interact autonomously, an error or malicious act by one agent could trigger a domino effect, leading to widespread market instability or significant financial disruptions. Therefore, security- and privacy-oriented agentic design and compliance procedures need to be followed before agents can be used in any financial system.

What safeguards or technical requirements are necessary in this new model of agentic commerce?

Ravi Kumar: This is a very important point. Ensuring trust, compliance, and safety in agentic commerce, where AI agents autonomously handle financial transactions, requires various safeguards. If I had to mention the top five, I would list the following: 

1. Robust authentication methods
2. Data encryption
3. Real-time threat detection
4. Adherence to compliance and regulations
5. Privacy-oriented agentic design

Is agentic AI ready for widespread deployment today?

Ravi Kumar: If we talk about the retail sector, there are many domains where agentic AI solutions are widespread and currently being used effectively. Domains like supply chain, IT operations, customer support, and business reporting are using agentic AI solutions very aggressively.

At the same time, domains like financial transactions are still not ready for fully autonomous agents. I know one use case where Amazon's “buy-for-me” feature allows autonomous purchasing on third-party sites, but it still requires explicit user confirmation to complete the final transaction.

What milestones do you see in the coming years on the path to adopting agentic AI?

Ravi Kumar: There is a huge scope of work that needs to happen in the coming years.

From a technology perspective, we need the development of more mature agentic AI solutions with real-time auditing, monitoring, and governance of AI tool fleets to streamline the use of agentic AI. This will enable businesses to deploy agentic AI safely and compliantly.

From the consumer acceptance side, the adoption of agent-based retail solutions will be an important milestone – where consumers fully rely on agents to make purchases.

Work to make AI shopping a reality is already well underway. With the next few years focused on solidifying trust, security, and interoperability, we can easily shape the path toward true mainstream adoption.