What is it?

Agentic commerce refers to AI agents or assistants acting on a user’s behalf to carry out online shopping tasks, such as researching or buying products or services.

Agentic commerce can include AI agents searching, comparing, recommending, applying discounts and completing purchases.

Where is it happening?

For consumers, agentic commerce can take place within the conversational interfaces of generative AI applications such as ChatGPT or Gemini. Retailers are also integrating it in their websites or e-commerce platforms, while third-party aggregators are using it to allow shoppers to browse, compare and buy products across multiple brands.

In B2B contexts, businesses can use agentic commerce to automate parts of the procurement process, including supplier selection and purchasing.

Not to be confused with: chatbots

E-commerce chatbots are reactive models that respond to a user’s questions and provide information based on pre-set rules, whereas agentic commerce is autonomous and can make decisions. Agentic commerce also has a wider scope, as it can carry out purchases from start to finish.

What does this mean for market research?

Technology that automates purchasing decisions will change the purchase journey and businesses will need to understand its implications for buyers in their sector.

A shift to agentic commerce also has the potential to change the focus for marketers and researchers. Felicia Hu, managing director, Assembled, said: “Agentic commerce will shift the basic question market research has always asked. For decades, we studied what people choose. In the agentic world, it will be what they let an AI choose for them. 

“When an agent recommends a serum, a travel package, an insurance plan or a B2B vendor, the choice barely matters. What matters is the moment after: does the buyer accept, check, push back or override? That hesitation is the new market research data. We expect new terminology such as the approval journey, trust cues, social proof and risk tolerance to arise in our field.”

• This series has also looked at agentic AI, digital twins, synthetic personas and synthetic panels.