#71: A Year in Review - Relentlessly Curious
Excerpts that resonated the most
In February, I decided to try my hand at writing (pun intended) as a creative pursuit. I told myself, “Commit to writing for three months. That’s it. Just try it out.” Before I knew it, three months turned into ten, and ten months turned into 70 editions.
The growth of Relentlessly Curious has been both humbling and energizing, a reminder that there’s real appetite for clear thinking amid the noise and for curiosity that’s practical, not performative.
In 2026, topics will continue to explore the human interaction between business, media, and technology. And yes, expect plenty of AI along the way.
I’m hitting the pause button for the next two weeks. In place of a typical Relentlessly Curious piece, I’ve pulled passages from some of the highest-engagement articles this year.
Thank you to everyone who read Relentlessly Curious this year. I’m grateful for your continued readership. See you back on Tuesday, January 6th, for the first Relentlessly Curious edition of 2026!
Evaluating AI’s Impact on the Customer Service Industry
Publication Date: November 18th
Tackling the consumer side (B2C) of customer service automation is tougher as there is a theoretically infinite number of inquiries for businesses of all types and sizes. It’s difficult for a company without a swath of data and global distribution to start training an AI agent that can represent consumers. And that’s not even to mention the sheer capital needed to turn all this data into insights. In summary, B2C AI customer service can be best supported by the foundational layer (possess the resources and expertise), while B2B AI customer service (niche, proprietary data sets) is a better fit for the application layer.
Yet, you know who has a tremendous amount of data, expertise in AI, and global distribution? Google. In the grand scheme of things, the acquisition of hyperlocal data is an ancillary benefit. I believe Google introduced “Let Google call” as a first step at redefining the consumer side of customer service.
What makes the “Let Google call” product launch even more impressive is how it bridges AI agents across the online and offline worlds. Plenty of businesses barely have a website, so why not have AI do the legwork and reach out to these businesses in their preferred communication method? Over the phone.
Date: November 11th
Whether brands realize it or not, AI doesn’t just support the brand: it becomes part of it. Most brands are not technically qualified to develop or represent AI tools. When customers receive poor recommendations, trust erodes, and they may switch to competitors. It’s not natural for a consumer to verify a brand’s guidance, nor should it be their responsibility.
If ChatGPT recommends a serum that doesn’t work, you might shrug it off, thinking, “It’s ChatGPT, it sometimes gets it wrong.” Tools get more slack than brands when it comes to wrong answers because it’s their business to move quickly and build innovative technology. Some screws are bound to fall off in the process. When AI tools fail, we shrug. When brands fail, we are likely to switch.
The Case Against Instant Checkout
Publication Date: October 14th
Can OpenAI get 35M people per week to spend $50 and how long will it take them to get there? But most importantly, where is this sales volume (gross merchandise value, GMV) going to come from?
OpenAI currently has an Etsy integration in place and Shopify is soon on the way. Etsy as a marketplace puts up ~$12B in GMV per year, and Shopify reports ~$300B in annual GMV. By my estimates (which could be wildly optimistic, please comment below what you think), Instant Checkout will receive about one third of these massive platforms volume. Will there be a pure share shift of demand from shopping on brand websites to shopping in a prompt format? Or will new demand be created?
In terms of demand generation, I’m dubious as it’s not like people just suddenly have more disposable income because the user experience is cleaner. Remember, 10% of US consumers make up 50% of overall spending. I don’t know how much more we can squeeze out of the top 10%.
Why GenAI Isn’t Supercharging Businesses
Publication Date: August 22nd
Furthermore, it’s critical to define what “GenAI” means for each business. Are executives viewing it as, “let’s get everyone a ChatGPT account?” or are they seriously thinking about how each function can become more operationally efficient in a silo, as well as a cohesive machine? The former is helpful but unlikely to lead to a material revenue stream as there isn’t an enterprise strategy of how automation leads to top-line growth. The latter is strategic yet requires patience, time, and an attitude shift towards innovation.
There’s another reason why companies aren’t seeing huge revenue growth thanks to AI. And it’s rather simple. AI isn’t good enough yet.
We discussed in Are We Ready for Everything to be Automated? that the technology isn’t at a point where businesses can consistently get 100% accurate answers. Particularly in the way of LLMs like OpenAI and Perplexity. They are looking to create broad-sweeping solutions that appeal to everyone, not to your very specific use case. LLMs have raised enormous amounts of capital, so they need to tackle major industries at large like advertising, commerce, and software engineering. With that said, they’re more focused on providing a solution that is “good enough” for your specific use case. The problem is, 80% correct doesn’t cut it when you need 100% correct.
The vertical layer of AI presents a tremendous opportunity for value creation because of the need for companies to go from the 80% that LLMs will provide to 100% correct. That extra 20% is incredibly valuable and companies will pay dearly for it.
Why AI Companies are Licensing Publisher Data
Publication Date: July 15
I’m curious if we’ll see a wave of front-end (chatbot) licensing deals emerge to strengthen recommendation systems. Think of how helpful all of those “My Top 10 Skincare Products” or “What Type of Grill Should I Buy for Summer BBQs?” articles will be in matching shopper intent to products. Ask Rufus for a breathable golf polo, and it’ll likely pull from a “Top 10 Golf Shirts in 2025” article.
Not only are these articles relevant to the transactional intent of someone using Rufus, but they also have clean metadata for AI companies to train on. Publishers should lean less on how many people read their site, and more on how many algorithms crawl it. Sure, this clashes with the ethos of journalism. But times are changing as publishers need to realize that AI is a growing share of their customer base.
Target, Walmart, and other retailers will likely launch their own chatbots to guide purchases, and they’ll need publisher partnerships to train them. The business case for integrating media (publishers) with commerce (retailers) is so clear and one that the publishers will need to hold on as tight as possible for the sake of survival. Commerce media is thriving, and monetization models and deal structures are evolving fast.
Why Brands and Consumers Love Erewhon
Publication Date: June 10th
A boutique market shopper is sophisticated, prioritizes what goes into their body, and has the disposable income to choose healthier, cleaner products. They view the extra cost (compared to Wegmans or Trader Joe’s) as an investment in their health, so of course they would pay extra.
Taking a trip to a boutique market is a more emotional grocery shopping experience than strolling into Trader Joe’s. You’re likely to find brands and products you aren’t going to find anywhere else, all emphasizing health and wellness. These markets make you feel healthier when you walk out, even if all you do is gawk at a 1.5-ounce vitamin C shot for $8.
Publication Date: May 13
Say you want to make lasagna for dinner tonight but don’t know which ingredients you’ll need. It’ll soon become instinct to ask ChatGPT (or Claude, or Perplexity) what you need from the grocery store to make lasagna. The LLM output will give you a specific grocery list and a recommendation for picking up the ingredients at nearby stores. Or if you want to really automate the process, ChatGPT will communicate with your Instacart account to add the ingredients to cart and then make the grocery purchase for you (of course, using your credit card details).
Boom, lasagna ingredients delivered to your door 30 minutes later. Type in another prompt and you’ll have a detailed recipe walkthrough too.
The lasagna example is just one way AI makes commerce more efficient. No endless browsing or guessing what you may need. One clear list of ingredients and you didn’t need to jump from website to website or tab to tab on your phone or desktop. The experience happened all within your ChatGPT app.
How Credit Card Companies Manipulate You at the Airport
Publication Date: March 21
That’s how they prime you. Airlines and credit card companies want you to have a positive association with travel as the data shows you spend more when you are on the go. So, their goal is to get as top of mind as possible as you are starting off your trip, which may lead to you using that same credit card throughout your travels. Fancy dinner? Museum tour? Train from London to Paris? Throw it on my Amex. I’ll get the extra membership rewards points so I can book my next trip to Mexico.
And in some ways, they physically prime you for your card to be top of wallet as well. Ever realize that the lounge front desk asks you to see your credit card? They need to verify you but that’s likely going to be the last time you touch a card before you land at your destination. And you may have put that credit card at the top of your wallet. This nudge is thought out way before you stepped into the airport.
In a lot of ways, airlines are marketing companies that happen to have planes as assets on their balance sheet. They market all the ways you can spend your money on travel and have revenue share agreements with credit card companies to make a buck on every part of that cobranded spend.