A Finer-Grained Search – With Actionability
Free Article
Agentic commerce is evolving at a remarkable pace. Blink, and you’re behind. Having just wrapped up a survey of top merchants, I wanted to share some observations—and offer a few predictions worth tracking as we look ahead to the next two years.
If you’re not already tracking agentic systems, Citi’s Agentic AI: Finance & the ‘Do It For Me’ Economy is essential reading. The report outlines how autonomous agents—software bots trained in specialist tasks—are shifting from assistants to doers. Think personal AI that not only helps you shop, but completes the transaction. In commerce, these agents are redefining what it means to search.

Source: Citi GPS Report, Jan 2025
From Search to Execution
OpenAI’s Operator and Perplexity are pioneering a new type of interactive commerce. Imagine asking: “Buy an Easter dress, shoes, and hat for my 6-year-old daughter.” That single query kicks off a series of agentic workflows—spanning merchants, comparing styles, factoring past purchases—and delivers bundled, relevant results. It’s search, orchestrated and executed.
We surveyed top retailers actively preparing for this shift. Their responses revealed a consistent stance:
- They plan to leverage AI and their proprietary data to enhance in-domain customer experience.
- They’ll resist intermediaries that threaten their customer relationships or reduce them to price-based competition.
- They are explicitly pushing back on becoming just another node in someone else’s mega-platform—they won’t support the creation of a “new Google.”
- Shopping isn’t just transactional. Consumers love the experience. It’s a form of recreation and expression. One retailer noted: “People will send their kids to daycare just to shop.”
Within financial services, banks have the same resistance. Imagine a financial agent that will automatically open new accounts for you and switch your card if better rates can be found.
Predictions for Agentic Commerce: 2025–2027
1. Agentic Experiences – Starts with Who Consumers Trust Today
Agentic Commerce in the next 2 yrs will see progress in customer experience and acquisition. Consumers love to shop and changing shopping behavior is hard.. there is loyalty to merchants, brands and Google/Amazon. Its not just about price, but also about delivery, reputation, and many other factors. If you think of agentic as a better search, existing consumer endpoints have a big advantage. My example would be Amazon Alexa becoming more conversational as users describe their plans or needs, and Amazon makes recommendations. For example “I’d like to cook Indian food tonight, can you give me a good starter recipe and key ingredients that I may not have at home”?
Assuming Agentic drives significant improvement in CX, expect retail sales volume to accelerate movement to eCommerce/OmniChannel at 5-10% above baseline growth. From a TMT perspective, Agentic should drive significant change in digital advertising campaign spend (subject for future blog).
2. A Finer-Grained Search – With Actionability
Where Google built an empire on keywords, agentic platforms like Perplexity will broker deeper, intent-based partnerships. Agentic commerce is a much finer grained intent signal. This signal may be sourced by a single trusted merchant that then will collaborate with with others to meet the need. For example Kohler currently partners with Sherwin Williams and the Tileshop to share intents for bathroom remodeling. While Google or Perpexity may source the intent and seek an affiliate/lead fee, the sources could be much less concentrated than Google’s search today. Google’s ad network operates across every merchant and publisher, an extension of their current model into agentic make the most sense.
3. Ad Networks Will Adapt to Power Agent Transactions
The Agentic revenue model will begin with ad networks. Agent operation and collaboration may be led by specialists (ie Anthropic and the MCP Protocol) but Google, Amazon and other Ad Platforms will adapt their advertising networks to handle the financial exchange (and measurement of effectiveness). Measurement is particularly important as feedback to model effectiveness and also for pricing (of this “enhanced search”).
4. Wallets Will Become the New Gatekeepers
Wallets will handle permissioned access to personal data and actioning authority for agents. Consumer data, authorization, and trusted agents. My good friend Dave Birch and I agree on this topic. Identity and authorization will be key to this next phase of commerce. I see Wallets as the gatekeeper or “controller” that permissions data among trusted agents as well as permissioning “action”. BTW, none of the wallets can make this happen without Visa and Mastercard networks. Who owns the risk in an agent transaction? It is the merchant.. And they will have a significant say in how transactions are authorized within the new experiences they build.
Final Thoughts
Agentic commerce won’t replace shopping—but it will rewire how value is delivered and who captures it. Merchants, with their deep consumer data, remain in a position of power—but only if they act quickly. Open agents will thrive in high-friction categories like travel. For everything else, agentic commerce will be a hybrid dance between shopper, agent, and merchant. Will the agent sit closer to the merchant or closer to the consumer? There will be both consumer agents and merchant agents.. but in payments (and in advertising) merchants pay for everything.. so we will first see success in merchant driven efforts. This is where Google has the clear lead.. with merchant relationships, all the consumer data, google cloud, and a world class set of data scientists.
And yes—AI friends will beat me up for this post..
Great post Tom and at the risk of sounding like an echo chamber, yes of course I agree with you. I’ve just been writing some more on this for Forbes and I’ll put it out tomorrow. I guess the point I want to make is that while some shopping is fun, of course, most of it isn’t, so I think people will readily hand it over to bots, especially for things that are boring (like savings accounts).
I would only further qualify.. boring where there is no loyalty or convenience. For example Amazon handles most of my boring.. but they know everything.. the sizes of my replacement filters and the models of my cars (for car items). I also am extremely loyal to hotels and airlines. For anything above $100-$200 community reputation is important (which bots could automate .. or fake).
Thus I would see the first successful agentic commerce categories as those products which are boring without loyalty or convenience (costs). This could also include complex bundles (build your own custom bicycle). This is not a high margin space.. nor are ad $$ high. The real $$ will be in helping brands construct better customer experiences that they own and control. IMHO