Shoptalk Debrief: Agentic AI, Loyalty That Doesn't Suck, and the 10% Nobody's Solving For
five things I learned from 72h of vendor booths, brand convos, and way too much coffee
Hi Team!
As much as Vegas is a depraved, soulless city, Shoptalk was still fun as hell.
We spoke to hundreds of brand executives, vendors, and operators trying to figure out what the next two years will look like, and I walked away with a few things rattling around my head that I can’t stop thinking about.
I took a shot every time I saw “agentic” on a booth and was drunk at 9 am. Everyone’s talking about product discovery and shopping via LLMs. Loyalty programs are finally starting to look interesting and experiential.
And weirdly, even though there are a dozen players in the agentic CX space, almost nobody is thinking about what happens to the 10% when AI handles the other 90% of your tickets.
Let’s chat about the big themes.
This week’s newsie is brought to you by Horatio and Mentors CX.
I often get asked how to upskill a CX team without sending everyone to a $3,000 conference or sitting through a webinar that could have been an email.
Horatio and Mentors CX put together something genuinely worth your time.
It’s a 12-module CX masterclass built for operators who actually run these teams. Valued at $950, they made it available for free.
It includes: journey mapping, VoC programs, AI, omnichannel strategy, team building, the whole thing.
Each module comes with real frameworks and downloadable guides so you can apply it immediately, not six months from now when you’ve forgotten everything.
I’ve seen a lot of “resources” come across my inbox, and most of them are lead gen dressed up as education.
This one is different. It’s hosted on the Mentors CX site, fully free to access, and built with the kind of depth that reflects what Horatio actually does day to day running CX operations for some of the best brands out there.
If you’re a CX leader trying to get your team sharper, or you’re building out a function for the first time and want a real framework to work from, this is worth a few hours of your time.
AI Consulting & Implementation Services
Agentic AI isn’t just about cutting costs. It’s about transforming how your business supports customers.
Horatio helps companies deploy and optimize agentic AI tools while integrating high-touch human support where it matters most. The result: more efficient operations, smarter automation, and better customer experiences.
And once AI handles the repetitive 80–90%, what’s left is the 10% that actually defines your brand and drives loyalty: the complex, high-stakes conversations that require real human judgment. That’s where having the right support team in place matters most.
If you’re ready to implement AI, or need a partner to guide you in the right direction, book a call with the Horatio team.
“Agentic” is the new “blockchain”
I lost count of how many vendors told me their product was “agentic.” The word showed up in pitch decks, booth signage, demo scripts, and LinkedIn posts written in real time at the conference. And when I asked what that actually meant for the customer experience, the answer was usually some version of “well, it can handle multi-step workflows autonomously.”
Cool. So can a Zapier flow from 2019.
Sure, some of the tech is real. MCPs, multi-agent orchestration, whatever the vendors are hawking this quarter. It’ll probably matter at some point. But most brands I talk to are still trying to get their basic ticket deflection above 40%. Building for the future is great, but I’d reckon that only 2-5% of your customers would actually use your MCP today.
One brand exec I talked to put it perfectly:
“We implemented the fanciest AI agent setup you’ve ever seen. Customers still email us asking where their package is, and half the time the agent can’t pull the tracking number because it’s in a different system.”
The tooling is racing ahead of what brands actually need, and most of the ones I talked to are drowning in sophisticated solutions to problems they don’t have yet, while the simple stuff still doesn’t work.
Shopping via LLMs is coming and nobody knows what to do about it yet
Perplexity, ChatGPT, and Gemini are all racing to let people shop directly in the interface. Brands know it’s happening, they know they need to show up wherever their customers are, and almost nobody has a plan beyond “I guess we should make sure our product data is clean.”
I talked to a VP of Growth who said their team spent a month trying to figure out how to optimize for LLM-based product discovery, realized they had no idea what that even meant, and gave up. She’s not wrong to be confused because the playbook does not exist yet. SEO took a decade to standardize. This is month six.
The few brands I spoke with that are ahead of the game are obsessing over structured data, clean product feeds, and making sure their brand story doesn’t get mangled when some LLM scrapes it.
A few folks I spoke to are leveraging Shopify Knowledge Base to embed their product catalog so models actually show their stuff when someone searches for the best cruelty-free skincare for dry skin.
Will it pan out? Who knows. But they’re doing something instead of sitting around waiting for a Harvard case study in 2027.
(Shameless plug: Yotpo launched Yotpo Discover to tackle this, and we have a few hundred brands on the waitlist in the first few days. )
Loyalty programs are finally interesting again
I have spent years complaining that loyalty programs are just frequent flyer miles for people who buy deodorant. I still stand by that take for 90% of what I’ve seen. But I walked away from Shoptalk seeing a handful of programs that actually made me think, “Oh, that’s different.”
260 Sample Sale has a program where top-tier members get to skip the line at their physical sales. It sounds small, but if you have ever stood outside a sample sale in New York in February waiting to get in, you know that it is worth more than 500 points toward free shipping.
Fleur de Mal is letting top-tier loyalty members give input on product development, which turns your best customers into collaborators instead of just repeat purchasers.
Points are fine, discounts serve a purpose, but nobody ever felt special because they got 50 points for typing in their email. The brands that actually get it treat them like actual people with names who deserve better than a robotic birthday discount that got queued up at 2 am by some sad automation workflow. 🙃
Everyone’s fighting for AI CX, nobody’s building for the 10% that matters
There are at least 30 vendors right now, all fighting to be the AI-for-CX platform that wins the next five years. And most of them are building the same thing, automating password resets, order tracking, return labels, the stuff that should have been automated in 2018.
They will all do fine, some will win, some will get acquired, but the thing I could not stop thinking about is that almost nobody is building for what happens when that 90% goes away.
When AI handles the repetitive stuff, what is left is the 10% hard conversations.
The customer whose $400 order showed up damaged the morning of an event, or the angry email that is not really about the product, but the fact that someone feels ignored.
Those tickets require judgment, empathy, tone calibration, and the ability to make a call that is not in the playbook. Those edge cases that don’t inherently require a human, but would positively impact LTV and brand if you did get the right human response in the mix pretty quickly.
I talked to exactly zero folks focusing on this.
Everyone is still pitching “we can automate 80% of your tickets.”
The brands that win are going to be the ones who realize that AI is not the endgame, it is the thing that frees up your team to do the work that actually builds loyalty, even if it’s only 10-20% volume-wise.
That’s it for this week!
Any topics you’d like to see me cover in the future?
Just shoot me a DM or an email!
P.S. If you want to figure out how to get your brand to rank high in LLMs and show up in ChatGPT, Gemini, and more… check this out.
The "10% nobody's solving for" point is the most important thing I've read out of any conference debrief this year. Every vendor pitch I've seen is anchored on deflection rate - automate 80%, cut cost-to-serve, ship it. But once you've automated the order tracking and the password resets, what's left isn't just harder, it's higher stakes. The damaged order the morning of an event. The customer who isn't angry about the product but about feeling ignored for six months. Those interactions don't need faster resolution — they need better judgment, more authority, and someone empowered to make a call that's not in the playbook.
The brands that win the next few years won't be the ones with the best deflection rate. They'll be the ones who figured out that AI handles volume, but the 10% builds the relationship - and invested accordingly in the people who handle it.