Retail Pulse Report: The Vibecession Continues, Agentic AI and Workflow, and Amazon’s Latest Physical Retail Attempt
There is nothing new to learn about consumers this week, so let’s think instead about whether AI agents can deliver on the old promise of enterprise workflow.
Adobe Stock associates “shopaholic” with “vibecession” - which may not be so off as it seems at first.
Another week, another set of mixed messages about the economy and how consumers are doing – and how much they’re willing to spend. Different regions are seeing different indicators, both positive and negative, but the result is the same: consumer lack of confidence and a wait-and-see attitude toward the future.
Evolving concepts in how to navigate a GenAI world can give one a lot to think about, and I did my share of thinking this week based on both the concept of latent space, and back to agents in AI.
Let’s dive in!
Retail Economic Indicators
It’s the middle of the month and that means there is a slew of data from various governments and trade organizations on the state of retail around the world.
First up is Germany, where HDE, the trade association, reported that consumer sentiment ticked upward in February after a pretty dismal reading in January. Optimism about the future increased, however this seemed to have minimal impact on willingness to spend. Consumers reported both that they had no plans to increase spend, but also that they had no plans to increase savings either, indicating a wait-and-see approach to the future.
The numbers for the US were very closely watched, to see if consumers did indeed have a spending hangover in January, but this did not appear to be the case. Though you wouldn’t have been able to tell by reading the headlines (sigh).
The January CNBC/NRF Retail Monitor found that sales were up significantly in January – year over year. Folks, this is the only comparison that matters in retail, and the December to January differences are exactly why. If you want a deeper explanation, see Greg Buzek’s post on LinkedIn and for all our sakes, stop reporting on month over month changes. My goodness, people spent a ton of money in the run up to Christmas, pretty much the biggest gift-giving holiday in the US? And then they did NOT spend a lot of money in January, when you don’t even have the Super Bowl to really spur any spending these days? Color me shocked.
Anyway, back to the results. The CNBC/NRF Retail Monitor found sales up 5.44% year over year. Even with inflation ticking back up, this is a great number. And there were several categories that have been struggling that this month, beat the average, including apparel and general merchandise. The US Census Bureau results, while not quite as high, still also beat inflation at 4.2%. Even eating out spending, which some economists noted had fallen off in December, was high – perhaps countering their claims that consumers were holding back on eating out in order to fund gift spending.
Inflation ticking up, labor market staying stronger than expected, consumer confidence falling but consumer spending staying high – it feels like 2024 all over again!
In the UK, an even more complicated picture emerges. The overall economy grew by 0.1% in Q4, which was better than the 0.1% decline that economists had expected. Most of the growth came as a surge in December, and came from services and construction, which offset production declines. Inflation also unexpectedly fell to 2.5% in December. The Bank of England last week subsequently cut interest rates from 4.75% to 4.5%.
Also positive news, UK footfall “surged” 6.6% in January as consumers came out strong for post-holiday sales. This was after a 2.2% year over year fall in December, as tracked by the BRC-Sensormatic figures. And this was despite a lot of snowy weather in January.
Two things to take away from this news across three countries. While it is happy news that some of the retail economic indicators came out positive for January, that’s not necessarily the best thing for retailers’ bottom lines. January sees heavy discounting to clear out the last of holiday inventory in preparation for incoming spring product. Consumers may have come out in force, but they were there to buy some of the lowest-margin inventory that a retailer can offer. March quarterly results will tell us what really happened – did retailers have to discount sooner into December? Did they have to take even bigger hits (unplanned hits) in January?
The second thing to take away is that past results are not good predictors of the future! OK, January was better than expected. But at least for the rest of first quarter 2025, the defining characteristics look to be “uncertainty and chaos”. The UK had some better than expected results – but also, UK retailers are scrambling to figure out how US tariffs are going to impact them – which tariffs, if they are going to go into effect at all, or when will they go into effect if it turns out to be certain (and how long any of that will last once it’s in place).
Next, a UK fashion retailer, is exploring opening a US subsidiary so that any tariffs will be on the cost of goods, not the selling price (which closing the de minimis loophole exposes). Superdry will stop shipping direct to the US from China. And general confusion reigns over whether shipments can even contain mixed origin products, like a product from China and one from Turkey, how such a package might be assessed. That’s even before the on-now-off de minimis order, which had to be paused because the US Postal Service was not prepared to enforce it and packages started piling up.
AI & Retail
There is a lot going on at the headline level in AI, but the more interesting pieces are coming from quieter, more thoughtful places. You’re undoubtedly already sick of hearing about “agentic AI”, but you’ll start hearing another term, if you haven’t already: latent space.
Neil Perkin wrote a great piece defining latent space and what that means to brands. Read it, but for the sake of discussion, my ham-handed explanation is that LLMs create a lot of connections and associations behind the scenes to help them identify the patterns that result in the recommendations or answers they give. It’s this pattern recognition of things that humans miss that helps to make LLM’s powerful and useful. As those connections are not typically directly exposed, they live in “latent space”.
Brands would naturally find those under-the-waterline associations that LLMs are making about their own brand extremely valuable. Since we can (still) say that the vast majority of data that LLMs are trained on is human-generated data, when they make behind the scenes associations about a brand, they’re reflecting associations that people make, and these days that is brand gold.
That’s not to say that a brand can do a whole lot to change those associations – the days of dictating a brand’s positioning are long gone if they ever existed – but knowledge is power. You can lean into the associations that you want to make stronger, and do things to de-emphasize or drown out the associations that you don’t want out there. This seems like a better use of brand resources than trying to figure out how to market to chatbots.
One question I touched on last week related to whether agentic AI will replace APIs (since they will very likely be able to navigate UI’s), and whether you’ll end up with some kind of coordinating network of agents, some specializing in specific tasks and some specializing in marshaling a set of task agents into a larger, more complicated effort.
Continuing down that vein, I remember when the promise of enterprise workflow first burst out on the scene. The idea was basically that you have a bunch of services or microservices that all have a specific job, and you should be able to string together whichever services you want into whatever workflow you need. I know some companies had some success in investing in enterprise workflow, but my impression is that it turned out to be far more difficult and expensive and took far longer than they expected. One problem that always seemed to catch people up was figuring out exactly how to expose that workflow to end users – you need a UI, and only in the last few years has it been possible to even think of generating UI as you go.
But, with agentic AI, do you need a UI? Maybe not. Or maybe all you need is a chat interface, and the agent can do the rest? Still mulling over that one.
Retail Winners and Losers and Store Innovations
I’m combining two topics into one here because this week the news was all about only one company: Amazon. Amazon reported on the state of its Prime Membership. The emphasis was on shipping speed and how much faster Amazon has become. I reiterate, I really don’t want something delivered at 4am in the morning because it sets off our dogs and this is a very unpleasant wake-up call. And despite my attempts at restricting my hours, saying I want the later delivery slot, what always happens? Ignored. 4am delivery. Faster is not always better.
But the statistic that really stood out to me (and makes me reassess my own purchasing habits): Prime members in the US placed an average of nearly 100 orders in 2024, or nearly two orders every week. How anyone else is going to be able to break into that level of addiction – and afford it – I have no idea.
And Amazon is back to attempting stores. More than just grocery stores, that is. And any time Amazon attempts it, it’s worth paying attention. This time, Amazon is focusing on beauty and Italy, which is seen as a foothold towards expanding into Europe. The store has experience zones, including digital skin analysis and pharmacists on staff who can recommend (non-prescription) products.
Beauty is a very competitive space, though it’s one where brands are as active as multi-brand retailers, so Amazon’s entry may be seen more as a new channel than a competitor. And Italy, in a European market that overall is shakier than other regions, is an interesting choice of location. It’s not quite as difficult to set up a store in Italy as it is in, say, France, but it’s not easy. But then again, Amazon has historically been unphased about implementing and then wrapping up store experiments, even when the store count suggested they had moved well beyond the experimentation phase. We’ll see.
What Did We Learn This Week?
On retail economic health, we didn’t really learn anything new. The positives maybe are coming from different places, but they’re offset by new negatives in just such a way as to prevent any sense of consumers’ definitive future spending intent. The vibecession continues.
I should note that there were a few consumer surveys that came out last week that I did not cover, mostly because they were either “meh” conclusions – when eggs get ridiculously expensive, did you know that consumers will buy less of them and will complain bitterly about it? Or they were clearly surveys designed to promote a solution provider’s point of view. Sometimes those are still useful, but not this week!
In the meantime, I’m continuing to ponder retail’s AI future. We’re still in the first mile of the marathon, I think that’s worth emphasizing as I get more questions from the press that start out with sentiments like “now that AI is done, what’s next”. And while others are tackling much bigger questions than me – you know, things like what does AI mean for the future of human capacity for independent thinking (not good, by the way) – it will be impossible to consider retail’s AI future independent of retail companies’ tech investments and tech future.
It’s tempting to think that you don’t really need to worry about the tech side (there’s a chatbot for that), but if we’re talking about coordinating agents which have been refined to complete a specific task – that still ultimately needs to be enabled by software – we’re still firmly in the realm of tech investments. Retailers sitting on technology older than my children (who are in their 20’s, to be specific) are going to have a much harder time adapting than those with their solutions sitting in the cloud.
Oh yes! And one more thing to think about: in my poll on whether you would go on a first date to eat meatballs in bed in an IKEA, the headline is not that 80% of you said “hell no!”, it was that 20% of you said yes. But it was really not a statistically significant response pool, even of my current subscriber base so make of that what you will. Perhaps there is a dating app profile technique to try in there somewhere, I would not know.
Until next week!
- Nikki