I had a completely different Field Note ready to go this week.
I knew I had a busy week planned at Amazon unBoxed, so my Field Note had been scheduled in advance. But after arriving in San Francisco yesterday and then the opening keynote this morning, I had too many thoughts bouncing around my head to leave it alone.
What a day! Whew.
There were a whole bunch of announcements. New ad products. More AI. More automation. New ways to build campaigns, analyze performance, reach audiences and help people shop.
As a marketer and storyteller, I also love how certain words get positioned at these types of events.
Amazon called the first section of the day “Conversations and Commerce.”
That immediately got stuck in my head. I actually brought it up later with someone at Amazon right after the session. I was telling them how much I liked the wording but couldn’t remember the exact phrase for a second. Then it came back to me.
Conversations and Commerce.
I told them I needed to start practicing saying it until it just naturally came out of my mouth. I liked it that much.
Needless to say, I was on the edge of my seat.
My commitment with this series was to send notes from the field, and today the field is unBoxed. Rather than force everything into a perfectly packaged thesis, I figured I would try a little something different this week.
Here are some meandering thoughts coming out of the unBoxed Day 1 keynote.
Branded Conversations is the big one
If I had to pick my favorite announcement with the biggest implications, it’s Branded Conversations.
Amazon opened the day with a great agentic innovation. Branded Conversations creates an AI-powered shopping conversation for the brand inside the Alexa for Shopping experience.
The way I am initially thinking about this is like a brand-specific conversational experience inside Amazon’s larger agentic ecosystem.
The use cases are obvious. You create more bespoke and contextualized chat experiences with this shopper but the part I found especially interesting was what Amazon told brands to do first.
Get your product content and Brand Store in order. Then understand the questions your shoppers are actually asking. After that, if you want, you can provide even richer product information to help shape the conversation. Alexa for Shopping is then going to combine that brand expertise with product details, reviews and its own shopping signals.
As crazy as this sounds, I have read that story before.
OpenAI is doing something adjacent with shopping, particularly around helping consumers discover and research products. Approved brands can provide structured product feeds so ChatGPT can better understand product catalogs, attributes, prices and availability when helping people shop. Same story, different platform.
We are entering a world where brands don’t just optimize product content for the human who eventually lands on the page. We now have to thread the needle of a PDP that can communicate to both humans and machines at the same time.
Speaking of marketing to those humans, I’ll give this one away for free:
If you want to know what questions your shoppers actually care about, go read your Amazon PDP.
Read the reviews, read the questions and then use that language and answer those questions right in your content. Your shoppers have already written your conversational-commerce research brief.
Then, the most exciting part, Amazon makes the whole thing accountable. They can connect the conversation to the purchase to measure the impact.
Commerce wins again!
There is so much focus right now on GEO and improving share of recommendations. Amazon takes that a step further because it can connect the conversation back to the thing Commerce cares about most: Did the consumer actually buy? What did they ask about before they bought? Could that eventually help us understand which questions, needs and barriers correlate most closely with purchase? The possibilities are endless.
It also left me with a heavier question:
What happens to Sponsored Products?
Not tomorrow. But if product discovery increasingly shifts from search to conversations, how will the advertising model move with the interface?
That one is going to keep me busy for a while.
Conversational creative: show me the behavior
Amazon also showed prompts embedded directly into Responsive eCommerce Creative.
My reaction was immediate. I’ve seen this type of thing before. It looks awesome and demos incredibly well. I can already see it showing up in executive meetings and pitch decks.
But then I put my human-being hat on for a second.
There is a difference between something becoming technically possible and consumers deciding that is actually how they want to interact with ads. I need to see the actual human behavior happen before I get invested in the potential of the creative.
For now, I’m watching this one. Objectively cool. Waiting on the adoption.
Sponsored Services was not on my bingo card
This one pleasantly surprised me.
I mean sure I was expecting lots of AI and automation announcements but Amazon announcing Sponsored Services, which connects people with service providers based on what they’re searching for or purchasing, got me particularly excited.
I think the broader strategic move is smart. Amazon’s advertising opportunity keeps expanding beyond brands that sell physical products on Amazon. Services opens up another huge category of advertisers while letting Amazon use something it has in abundance: signals around what consumers may need next.
But forget Mike the marketer for a second, I absolutely need this in my life!
My house (and life) is currently in the phase where we seem to order something from Amazon and then immediately need a human being to install, repair or replace something related to it.
Simply put, thank you Amazon.
Reviews just became even more interesting
Amazon spent a lot of time talking about Review Requests.
And part of me was sitting there thinking, don’t they know that we know that reviews matter.
Brands have been trying to make review generation easier for years. So much so that there are entire companies built around helping brands generate more reviews within the constraints of the platform.
So my biggest question here is a gating one at this point: How much does it cost?
Amazon’s version reaches verified purchasers natively on Amazon and uses a performance-based model. A number of the existing solutions brands use for this are performance-based too, so the comparison against the alternatives already in market is going to matter a lot for early adoption.
In Amazon’s favor, if the cost is right and all else is relatively equal, doing it natively through Amazon feels pretty darn compelling.
That’s kind of how my mind works. I immediately jump to how would I sell this thing. But after thinking about it for a bit, something started to make this announcement much more interesting this year in particular.
Reviews are becoming AI fuel.
We just heard about how Amazon is using reviews inside AI-powered shopping experiences. Alexa for Shopping highlights customer sentiment, suggested questions and more right on the PDP.
Reviews are becoming one of the richest human signals available to AI shopping experiences. Reviews indicate what shoppers liked, what they hated, how they actually used the product, who they think it’s good for, whether the claims on the PDP actually hold up to consumer perception.
Now go back to Branded Conversations. Someone asks if a product is easy to clean. What do you think Alexa is going to trust more, your PDP bullet saying easy to clean, or hundreds of actual customer reviews saying it was the reason they returned it?
That makes review generation more strategically important than I think a lot of brands realize.
Agentic isn’t a KPI
This was probably my favorite bit of timing from the keynote.
Last week, I wrote a LinkedIn post arguing that I think we’re going to waste a lot of money making perfectly good automations agentic.
My point was that if an API already executes a predictable action reliably and cheaply, putting an AI agent in the middle does not magically make the outcome better.
You have to use agents where the reasoning changes what is possible. Things like complex campaign orchestration, cross-system decisioning, removing significant technical barriers and more. If the objective is to make something agentic, you are likely going to be wasting money. That is a potential risk of MCP integrations in our industry today.
Then Amazon showed an MCP architecture that I actually really like.
Their MCP Server gives agents access to tools across campaign management, reporting, measurement and other workflows. But Amazon also spent time on the fact that simply giving an agent more and more tools creates its own costs and complexity. Their approach includes mechanisms to keep the agent focused, along with more specialized expert tools.
That is a very similar approach that I am used to myself. Enterprise solutions and then craft tools. The AMC SQL Generator is a great example.
I can ask the business question, like I would to my analytics partner if I was looking for some research or a POV. The agent can then delegate the specialized work to the AMC SQL Generator, which turns that request into the query needed to get the answer.
That feels a lot like how I already work with an analytics counterpart today.
I love this because the agent doesn’t need to know everything. It needs to know where the expertise lives. The more specialized craft tools we build underneath it, the broader the questions I can ask without pretending the generalist agent suddenly became the specialist.
Agentic coordination (if that’s a thing) seems critical.
Then my brain went to the future of video
Publisher Insights takes data like what shows people watched, content genres, web-page activities and more into Amazon Marketing Cloud so advertisers can connect that context with downstream outcomes (what content types and partners actually drive sales).
Amazon also showed new Geographic Insights and activation capabilities using shopping, location and third-party signals to identify high-potential geographic opportunities.
I like the announcement, but I had even more fun connecting the dots on this one. To be very clear, this next part is me connecting dots, not something Amazon actually announced. But I think there are some really interesting implications for CTV.
If Amazon can identify geographic pockets where a brand is under-penetrated or has unusually high growth potential using commerce signals AND Amazon has premium video supply through Prime Video, how can those two things eventually connect?
Could I identify a geographic sales opportunity and then disproportionately create demand in that exact market with customized video creative in an automated workflow?
Could commerce intelligence make CTV planning more responsive to where the actual business opportunity exists?
I don’t know yet but that’s sure what I plan on asking about.
Conversations and Commerce
Which brings me back to the phrase that got stuck in my head at the beginning of the day.
Conversations and Commerce.
I just love that because I think there is something deeper there.
I think about my own behavior and what I am seeing on the consumer side of our industry. The shopping journey is increasingly starting with asking questions and getting answers on the way to eventually making a decision.
Amazon’s interface is now moving in a similar direction.
Think about the information hiding inside those conversations. What I need. What I’m confused about. What matters most to me. Where I’m willing to compromise. What stopped me from buying. What finally got me there. Those are incredibly rich commerce signals.
But before I ramble on and start to miss some of the sessions, I will leave you with this:
Language is becoming the new search bar. And when the interface is a conversation, context is what determines whether the answer is actually helpful.
That’s why Conversations and Commerce got stuck in my head so quickly. It didn’t just kick off Day 1 of unBoxed.
I think it’s where the purchase journey is going.