AI Ecommerce: Experts Share What Makes a Winning Strategy

The human mind's capabilities are nothing short of extraordinary — but even traditional intelligence has its limits. Similarly, artificial intelligence (AI), perhaps against popular belief, is not without limitations either.

So what happens when you combine the traditional and artificial, particularly in the case of building a winning ecommerce strategy?

The results can be killer — from boosting creativity and conversions to finding ludicrously efficient ways of complying with regulations and solving data problems.

Rob Gonzalez, CMO and co-founder of Salsify, and Steve Engelbrecht, CEO and founder of Sitation, joined Peter Crosby, vice president of corporate communications at Salsify, for a conversation centered on AI ecommerce — and all its power and limitations.

For AI Ecommerce To Work, 'We Need To Be Here'

As our experts and others (per Harvard Business Review) will attest, AI will not replace humans — nothing can be run on full automation.

"We’re talking about letting this technology support a business process where obviously we need to be here," Engelbrecht says.

In the case of AI writing assistants, for example, you can produce content at an extremely accelerated pace — but not without a need for significant edits and humanization.

"Humans are in the loop," Engelbrecht says. "We’re using the [AI] technology to support a workflow where we’re still making decisions, but it’s helping us to push faster, and do more, and leverage that data to create opportunities for ourselves and a competitive advantage."

But, if you can get firm with what rules your AI needs to follow or what it needs to look out for in even daunting amounts of data, it can prove invaluable.

"Imagine if you had an infinity of pretty capable interns working for you," Gonzalez says. "That gives you a lot of potential to scale things that you’re doing, especially things that can be well-defined."

For brands, there are only so many hours in the day to focus their best efforts. And, usually, those best efforts are applied to their highest-performing channels. But you might be losing a significant amount of steam — and potential sales — by not leveraging AI to fill in the gaps.

Employing these "pretty good interns" can lead to high- (or at least higher) performing product detail pages (PDPs) across the board.

"Instead of the 'Cs' that you’re getting on smaller retailers you could bring everything up to a 'B+' or an 'A-'," Gonzalez says. "And that would be a huge win."

AI as a Conversion Booster and 'Thought Partner' for Retail Media and More

AI doesn’t have all the answers — it’s a common misconception (more from Engelbrecht on that below). Rather, tapping into AI to leverage the data, insights, and talents brands already have shown it’s not only a valuable "thought partner" but a conversion booster as well.

Video Source: Salsify

Ecommerce boils down to two things: Get traffic, and convert the traffic. According to Gonzalez, the "explosion" of retail media has made these two facets a lot more related, especially for selling on retailers like Amazon, Walmart, Target, and Kroger.

"If you look at your top product detail pages through those top channels where you’re spending retail media, you can use AI as a co-pilot, as a thought partner, where you sit there and you say, 'Hey, here’s a product title, can you give me 15 other product titles that might convert higher?'" Gonzalez says.

This same approach can be applied to generating and optimizing product descriptions, images, videos, and even content featuring models wearing apparel and other products.

"None of the 15 it gives you is going to be a slam dunk necessarily, but they might get your brain turning, and they give you something to A/B test, and they give you something to experiment with," Gonzalez says.

How To Apply AI Ecommerce To Your Strategy

There are many ways for brands and retailers to use artificial intelligence in ecommerce, but what are the best ways to apply it to your organization’s productivity and operations to support a thriving AI ecommerce business?

AI Fixes Data Problems, Not Hard Problems

A perfect task for AI to do isn’t solving an enormous problem, according to Engelbrecht.

Instead, if there’s a well-defined process that you can scale up — even having the AI complete it a thousand times an hour — you can process a tremendous amount of data, and that’s where the value lies.

"Have it do the things that you know how to do that are data problems," Engelbrecht says. "That’s where there’s business value … where you can oversee it, and you can tell it, 'Yes, this is correct,' [or] 'No, this is wrong,' and bring that feedback back into the process so you can make it smarter and better and repeat it and fix your data."

AI Can Help You Get More Personal

Outside of the content generation and optimization applications of AI, companies are already seeing success with the machine-learning model for personalization.

Gonzalez references the Netflix model of optimization for personalization, which has been around for a while. More recently, companies have been leveraging data personalization to determine the best samples to provide their customers with.

AI can also further catalyze personalization, or, rather, differentiation, in how brands show up (and, with requirements, have to show up) on different retailers.

"You can have attitude, you can have style, you can have data differentiation," Gonzalez says. "The plane of competition is just going to move."

How AI Helps With Regulatory Compliance

When it comes to keeping up with ever-changing compliance and regulatory requirements — per retailers, the Food and Drug Administration (FDA), and everywhere in between — there’s no such thing as being too vigilant.

Engelbrecht shares a current application of AI for Sitation’s customer, Blue Mercury, a cosmetics retailer, whose suppliers need to tread carefully when it comes to making any healthcare or medical claims.

"There’s a fine line between saying, 'This product reduces wrinkles,' versus, 'This product, when used correctly, can reduce the appearance of wrinkles,'" Engelbrecht says. "One is a medical claim, which could cause issues with the FDA, and one is a marketing statement."

Though it’s a very difficult programming problem, according to Engelbrecht, it’s also a language problem in which the solution is very achievable with AI language models.

"It’s an absolutely perfect example of what we can program in with this sort of technology and build a rule to analyze the data at scale, build it right into a process, and say, 'Oh, sorry, supplier, you can’t do that,'" Engelbrecht says. "'But, here’s a recommendation of how we can reword this to make it more compliant with the brand voice, the rules, and ensure everybody in legal is happy.'"

How AI Can Help Cut Costs Today

Thanks to the wildly disruptive technology that is AI, on top of inflation and recession-related pressures, it’s been "a weird year for everybody," according to Engelbrecht. Reducing costs is arguably at the top of every organization’s priority list.

"Everyone’s very focused on cost and trying to do more with less and is kind of reluctant to invest in new systems," Engelbrecht says.

But, interestingly, according to Engelbrecht, this isn’t stopping companies from applying AI technology specifically to reduce costs. However, before application, brands are still asking how to manage both the expectations of their executive teams as well as the risks of this technology.

Engelbrecht stresses that companies say they want to derive the benefit of this exciting new technology but also don't want to publish things that are incorrect.

The solution lies in designing workflows that allow for human checks that data is correct and complete. And, in many cases, AI that’s checking AI.

"It’s not perfect, but it does work," Engelbrecht says. "And you can make money with it today as long as you’re smart about it and you’re building a workflow that allows you to look at it and be confident in the output."

The Risks of AI To Consider (And Not Worry About)

Feelings of uncertainty, hesitancy, or even avoidance of AI aren’t unreasonable. But, as our experts share, this technology can be used securely, and companies at the helm of it take concerns seriously.

Some of the main risks of AI, however, lie in the gravity of some of the assumptions made regarding its capabilities.

According to Engelbrecht, a lot of people jump to two conclusions about AI and related technology:

• One: "I’m going to tell it to do everything;" and
• Two: "I can automatically assume that it’s right."

People jump to these conclusions, according to Engelbrecht, simply because AI can speak in full sentences or spit out full paragraphs of output.

"There’s this baked-in assumption that it knows everything, somehow … or it’s automatically right," Engelbrecht says. "And we need to stop thinking that, it’s not automatically right."

AI and Security

Being secure when experimenting with AI tools is possible, according to Gonzalez, and the biggest brands behind these tools have it in their best interest to keep it honest.

Two ways to up your security are opting out of applications saving your data for training purposes — an easy toggle in ChatGPT’s settings, for example — and not sharing proprietary information.

"There’s a real fear there that the data you do put in there is going to be just absorbed into the borg and reused," Gonzalez says. "But you can also very easily protect against that."

AI and Copyright

The other fear that people have, Gonzalez says, is related to copyright — but that’s more or less covered by fair use in the U.S.

Video Source: Salsify

AI 'Can Always Be Checked'

Brands might also be concerned that AI can produce "crappy" content that doesn’t match their image and messaging and just put it everywhere online — but that’s not the case, according to Gonzalez.

"You need to have a human in the loop," Gonzalez says. "As long as you have somebody checking on it, it’s OK."

As Engelbrecht points out, it’s not any different than working with copywriters — or interns, either, according to Gonzalez — you have to have a process for checking content anyway.

Using AI to Work With What You Have (And Have Fun)

According to Engelbrecht, generative AI doesn’t necessarily mean creative AI — "generate" doesn’t necessarily mean you have to create something new.

"A lot of this can be applying this technology to changing something, analyzing something, fixing something, taking the data that you have, and using this technology to make it better," Engelbrecht says.

This reframe is a game-changer for established brands with tons of data to work with as well as up-and-coming brands.

It’s also important for brands to "step away from the keyboard" and consider all of the different touch points and applications that AI can be used for like search, SEO, taxonomy, analytics, channel syndication, and personalization.

And, when learning a new technology, according to Gonzalez, the best thing you can do is play with it and have fun to help you build a sort of "instinctual understanding."

"Play is how humans learn things," Gonzalez says. "The more you can build that into this, I think the more fun you’ll have and the more you’ll learn."

Video Source: Salsify

Some of the best tools for this play and experimentation, according to Gonzalez and Engelbrecht, are ChatGPT and Midjourney. (The latter can help any visual you imagine come to life, cigar-smoking cyborg parrots and all.) 

AI Ecommerce: The 'Capable Interns' Your Strategy Needs

Running a well-oiled machine now requires both the natural instincts of man and, well, machine.

Leveraging the traditional intelligence of your teams with artificial intelligence (i.e., interns) can get you where you want to be faster, do more, and create an ecommerce strategy that simply wins.