Musing on Advertising

After spending a week or so with Meta’s new personal AI agent, Muse, I’m starting to see a clearer picture of the future that many in Silicon Valley probably fully understood much sooner. Our personal lives will be coordinated and assisted by semi-autonomous agents.

Software developers have been directing small battalions of coding agents to complete tasks with tools like Claude Code and ChatGPT Codex for over a year now, but no tool has cracked the ordinary consumer market until Meta Muse rocketed to the most-downloaded app on the Apple and Google Play stores.

Going from Answering to Acting

So what is it about Muse that’s driving the hype?

Maybe the easiest way to explain is to cite a specific example from my experience as someone who uses ChatGPT as my daily AI driver with a dash of Claude and Gemini.

Last week, my wife suggested we should go to the Broncos Sunday Night Football game against the Rams for my birthday. I jumped at that opportunity quickly, but as anyone who has been to an NFL game for any team that consistently sells out knows, these games aren’t cheap.

So I set Muse on the task. I told it to find me the cheapest tickets given my parameters (lower bowl, rows 20-35, Broncos sideline, etc.), and it went to work. It scoured Ticketmaster and resale sites and returned the results. It continued to monitor pricing updates every four hours for a few days. Once it found tickets I wanted, I took over and completed the transaction. (I’m not completely OK handing over any login or financial info to Meta yet).

Now, this isn’t revolutionary, as ChatGPT could browse the web and do this, but the more magical part happened after I bought the tickets.

The real moment Muse clicked for me was a few days later, after I granted Muse read permission to my Gmail (scary, I know). Muse knew I bought the tickets because it saw the email. I got a little notification on my phone with my various transportation options (parking sold out, light rail and rideshare pickup location info), and it told me which stadium gate I should enter—the one closest to my seat.

The proactive nature of Muse is what other personal agentic tools were missing for me. It was a tiny interaction in the grand scheme and something I’m sure ChatGPT or Claude could do with the right setup, but Muse did it gracefully without me really thinking about it. This is the agentic future imagined by AI companies: an always-on, proactive, semi-autonomous helper of sorts that makes your life a little easier.

Now that I can see a future in which normal people offload their shopping to an agent, my mind turns to the question: What does this mean for advertising?

Amazon’s Dilemma: Protect Ad Revenue or Welcome Agents

Another big story from last week illustrates a very real fear many companies should have right now. Amazon quickly blocked Muse from browsing Amazon.com, which puts a real dent in Meta’s agentic shopping ambitions. Muse’s utility as a shopping agent is significantly diminished without access to the most popular retailer, which often offers the lowest prices.

But Amazon has the customer relationship to protect, alongside a 2026 forecast of ~$60 billion/year in retail media business. Amazon went through this last year when it started blocking AI bots from its platform, likely due to fear of losing the direct customer relationship with someone accessing their platform via a third-party tool and the opportunity to present sponsored listings to users.

Blocking agents was a relatively easy defensive move when only extreme early adopters were using agents to shop. But as more consumers delegate purchases, Amazon will have to weigh protecting its customer relationships and sponsored listings against the risk of losing sales to retailers that let agents shop on their users’ behalf.

Amazon has huge bargaining power here. Consumers may choose their agent based on whether it can shop on Amazon, rather than abandon Amazon for their preferred agent. An Amazon partnership could make one app the shopping agent of choice. The question is how much business agents need to control before blocking them costs Amazon more than letting them in.

Could Amazon open its catalog to agents in exchange for user shopping data? That could give Amazon more visibility into off-platform discovery. But that still does not protect its retail media business or its opportunity to advertise to our meat computers (human brains), not agents.

There has been talk of advertising to agents rather than consumers, but I’m skeptical of any model where someone can pay to determine what my agent recommends. If it steers me toward a product because someone paid for placement rather than because it meets my criteria, that runs against the reason I asked it to shop for me.

Retail Media’s Business Model Comes Under Pressure

So does this mean retail media will collapse? Probably less collapse and more evolve. If agents handle more of the shopping, retailers will have fewer opportunities to put sponsored products in front of human shoppers. They may seek to make money through fees for facilitating agent-driven transactions, while some advertising could appear in the agent’s interface through clearly labeled offers that users evaluate themselves, like OpenAI is currently experimenting with in its chat interface.

But there’s a distinction to be made: An agent showing me a sponsored discount and letting me decide is different from quietly favoring a product because someone paid it to. I’m skeptical of both, but the latter directly undermines why I delegated the shopping in the first place.

Ultimately, who gets paid may depend on whether AI companies need retailers more than retailers need AI companies. Retailers could charge AI companies for access to their catalogs and checkout systems to offset lost advertising revenue. Or AI companies could demand referral fees for bringing retailers customers. Either way, growing sales wouldn’t necessarily mean a growing retail media business.

And this doesn’t stop at e-commerce. The trend is that human eyes will spend less and less time on external applications outside their AI or agentic application of choice. Much like social media and search aggregated user behavior to drive more time spent on closed properties, we will see that trend play out with even greater gusto as the biggest tech companies vie for their agents to be the entry point to a user’s entire digital existence.

The unfortunate reality is that as advertising surface shrinks, the opportunities for retail media, individual ad tech companies, and publishers shrink with it. It’s hard not to be pessimistic as human web traffic declines and the biggest tech companies increasingly own agentic entry points. There could be more opportunity for companies to focus on increasingly scarce, potentially more valuable advertising inventory.

The Agent Race Will Be Won on Access and Trust

One optimistic angle is that this looks like a much more competitive race than the era of search, which Google dominated. Meta’s recent resurgence with Muse and varying user allegiances to ChatGPT, Claude, or Gemini show that this is very much an ongoing race of who will be the AI of choice for mass consumer adoption or, as Zuckerberg puts it, “personal superintelligence” for all.

Assuming all these solutions eventually reach parity, the agentic winner may be the one that threads the needle on ease of use, access to user data, partnerships, and trust.

Meta nailed ease of use, as Muse feels like the first agentic app that even my mom could find useful without complicated setup.

Access to relevant personal data could be another advantage. Connecting Gmail already makes Muse more useful to me, but that usefulness depends in part on access controlled by a competitor.

Meta could build on its relationships across Instagram and WhatsApp, while Apple could draw on personal context across the iPhone and iCloud, subject to what users permit its assistant to access. But having that information available and convincing people to grant access are different challenges. Apple could win here by continuing to make privacy and trust central to its pitch.

Partnerships can play a huge role as we see AI companies cutting licensing deals with publishers to keep their models fresh. Partnering with Amazon to facilitate agentic shopping could ultimately become a killer feature for one of these companies.

Trust is the last and arguably most important factor to consider. Muse has already been useful enough that I’ve let it read my email. I’m still not comfortable with it sending messages or making purchases for me. That gap between what an agent can do and what I’m willing to let it do may matter as much as the technology itself.

AI continues to introduce existential questions around advertising. AI chat and search introduced a reckoning for publishers, and now agentic workflows and shopping may herald a similar reckoning for retail media.

Retailers will face difficult partnership decisions if we experience the profound shift in consumer behavior that Meta and others are looking to capitalize on. But the agent companies will have to decide what they can sell without losing the trust that got them access to our lives in the first place.

The room every marketer wants to be in is returning to New York City.

On March 9-10, 2027, Marketecture Live will be back at The Glasshouse NYC for a two-day event that will convene leading executives across brands, agencies, publishers, and adtech.

Wondering what the hype is all about? Marketecture Live takes a different approach: expertise first, no panels, no filler, no fluff. Across more than 40 sessions and breakouts, every conversation is designed to be sharp, relevant, and worth your time.

The event has consistently sold out, so don’t wait to lock in your ticket at the best price you’ll find.