Spend a little time in a retail analyst’s shoes and you’ll hear a LOT from the vendor community about artificial intelligence. Do it for as long as we have, and you’ll likely allocate most messages – quickly – into one of several buckets. The first buckets gather up the over-the-top presentations, whereby AI will either improve our lives to points unrecognizable in no time flat or directly orchestrate our species’ demise. The next buckets are filled with the sentiments of the non-believers, a group of folks who seem intent on dismissing AI because their first interaction with ChatGPT didn’t meet their exacting expectations to the letter. We hear examples of both far too frequently. 

However, there is a third categorization: practical applications for these algorithmic-based predictive tools that retailers can use right now – and in the very near-term future – to make life easier (or faster, or both) for the folks they’ve hired to help meet increasingly-fickle customer demand. Whenever we opine on artificial intelligence and machine learning (as in our latest benchmark on the topic) we do our very best to take this pragmatic approach. But I recently found someone else taking this tack, and I must admit, I wasn’t quite expecting it.

I’ve had a while to think about what I saw in Chicago at Salesforce’s Connections event last month. And what’s really stuck with me most is that this was a show that was a) refreshingly practical in its demonstrations of current Artificial Intelligence applications and b) hosted completely in reverse. Allow me to start with the latter.

At vendor-led events, mainstage content can sometimes be a bit “fluffy”, to put it kindly. In Salesforce’s case (with the constant host of humans-in-animal-costumes milling about both stage and crowd) the term can be quite literal. But keynote presentations are important. They afford an opportunity to experience – in person – the vibe of a company. And hearing first-hand from leadership about the direction they are taking said company in the following months and years is important. But the details  that fill in the bigger picture often require real investigative work for an attendee to uncover. Typically this happens later in the day, in afternoon sessions with troublesome demos in smaller rooms well away from the concert-lit stages and the slick, video-laden powerpoints.

This was simply not the case.

Salesforce managed to use the Connections show (which addresses multiple industries from a marketing and commerce perspective) as a chance to highlight the company’s right now AI vision in a very real way right from the opening notes of President and Chief Marketing Officer Sarah Franklin’s keynote address. The company’s message going-forward can be positioned as “AI is the new UI” – and much of what happened in the first few hours didn’t just render deeper dives later in the agenda redundant, but actually eclipsed their content. It was genuinely a show in reverse.

“We believe 90% of online marketing content will be generated by AI by 2025,” began the CMO. She and her colleagues then proceeded to give demonstration after demonstration of how brands can – right now – use generative AI tools (powered by Salesforce’s Einstein GPT technology, of course) to get better content into both products and communications with shoppers before the buy button. Consider the following demo presented by Leah McGowen-Hare, SVP of Salesforce’s Trailblazer Community: 

• Rossignol, the legendary French ski company, is facing a very real problem. As climate change shortens winters and drastically affects snowfall, ski seasons are getting shorter and less predictable. The company is now racing to fill its product catalog with sports equipment for 3 more seasons of the year.
• Bicycles are a great way to fill this void (often times with the very same customer), and Rossignol has pushed hard into bikes. But as a latecomer to a market that is already saturated with brands, it needs real help standing out.
• The demo showed the process of introducing a brand new Rossignol bike model to the eCommerce site.
• Using Salesforce’s Commerce GPT, a merchandise planner clicks one button – “generate description” – and all of the new product’s details are automatically populated. This isn’t just spec-based attributes like frame size and component set – this is detailed information about what type of rider this bike is intended for, how it handles in the mud vs. loam vs. sand, how it compares to other bikes out there: the type of hard won, real-world expertise that typically had to come from a seasoned rider/salesperson in an upscale bike store. It happens instantaneously.
• The planner then simply does what Salesforce calls “vet it, edit, and take the credit.”

The time savings component to this demonstration was immediately evident. (Granted, proofreading someone else’s work – even if that someone is a machine – can sometimes take longer than writing it yourself). But from what was shown, content folks are still very much involved in the creative process of launching new products, but can now “spend their time” on more engaging tasks (more on that in a moment).

I got a little more hooked when the predictive tool was then asked to translate this particular bike’s text description into a foreign language. Commerce GPT had included the phrase, “This bike is like love at first sight” in the auto-generated blurb. As it turns out, that is a uniquely English-language phrase, and when translated into French, would have read, “This bike is like being struck by lightning.” Because the amount of norms and rules pre-learned by Commerce GPT, it knew the workaround for every language in play. Immediately, of course.

However, it wasn’t until the last phase of the demo that I really sat up: in the demo, a customer obviously elects to purchase a very nice bike from Rossignol online. Once they do, an Einstein chat window pops up to tell them (based on their shipping information) where the best local trails are for them to ride. If those trails are currently open or closed. What the weather is. And because they are still engaging, makes a suggestion for the latest model of backpack they might want to bring (buy). All while “payment now” is available as a single click in the chat window.

If you’re a cyclist, this is dangerously good software. If you’re a brand, it’s dangerously good software to shrug off.

Throughout my time at Salesforce Connections, I saw myriad practical applications of AI that’s ready for primetime right now. How Yeti uses similar tech to size coolers to people’s space at home and in their vehicles. How FI Racing is using the fact that 99% of its fans will never attend a race (over a third of which are net-new due to the Netflix series following driver’s every move) – but that these folks (which also happen to be largely female) engage with the app while watching the race on their TV – and love to buy merchandise. How each of us will soon have an AI-bot that buys all our clothing in ways we would never think to, and takes outfit risks that we would never take – but because it knows us better than we know ourselves, all with positive result. But none were as powerful as the first. Each had a disclaimer attached to it: some iteration of “we’re not replacing (insert title here – merchandise planner, customer service agent – any title with highly repetitive tasks) – we’re helping them. And that’s the part that requires all of us to pay more mindshare as we go forward. As with each iteration of “people replacing” technology (the steam shovel, the automobile, the personal computer), it would be a fool’s errand to beat back the ceaseless march of progress here. But whereas those revolutions required people to morph, to re-educate, or to “skill up” (as was so often used at this show), something about AI feels significantly different. Bigger questions are invariably at hand. What are we willing to replace? Are there things we are willing to fight to retain? And who among us really has any ability to fight for anything that fits the latter? On a more macro level: who will determine if AI technologies are truly open to the public, and if not – who determines what we can and can’t access? Can we trust our elected officials – the majority of whom seem less-than-cutting-edge, from a tech perspective – to regulate here?

Unfortunately, when it comes to AI, those decisions are not democratized. Very few people will be making those decisions for the rest of us. One of them happens to be Salesforce’s owner. Another would be Meta’s owner, who has just announced he’ll be making Llama, that company’s large language model, open to the general public – despite a chorus of voices suggesting this is premature. Either way, It certainly feels like the automated genie will never again see the inside of the bottle. 

And of course, because this is Salesforce, the conference still had some fluffy moments, for sure. As my partner Brian Kilcourse wrote after a not-too-distant Dreamforce event, much of the company’s industry-facing messaging can be boiled down to “Just trust us.” There was quite a bit of that still on hand at Connections in 2023. “Trust that we’re building the most equitable world” and “trust is the core of our corporation” – both direct quotes from the keynote. Each would go a long way further if not accompanied by “we’re net zero” and “we’ve achieved 100% renewable energy now” at an event we all just flew to (on airlines making the same, who-do-you-think-you’re-fooling claim).

But as it relates to AI, Salesforce is at least addressing the trust component in ways few technologists seem to be right now. They aren’t presenting hypotheticals, and they aren’t just ignoring the fact that we’re talking about people’s jobs and livelihoods here. Ms. Franklin began her keynote by stating that “The number one concern is can we trust AI?” Later in the day I watched her present on a panel with CMOs from First Horizon Bank and Tailored Brands where she got a lot of laughs by telling the audience that though it was once boring to make an online FAQ page, due to the shifting nature of online shoppers, it’s now one of the most important things a brand can do. And that AI can help make that content kitschy, catchy – completely aligned to whatever your brand’s particular vibe might be.

And while that might seem trivial at first, for me it was the perfect embodiment of what this show was all about: there’s this enormous potentially-terrifying-potentially-amazing technology that the entirety of the world can’t stop talking about: while they buzz, let’s find a really practical application for it. Right now. For something people would have completely overlooked a year ago.

I’m already looking forward to the next event.