The integration of artificial intelligence (AI) across Step One’s marketing processes has led to some strategic cost cutting.
Speaking with Ragtrader, founder and CEO Greg Taylor confirmed that his marketing team has been slimmed down to six people since signing on AI late last year.
He says the technology reduces the need for ideators and copywriters and allows graphic designers to reduce their workload.
“One person can just be far more efficient,” Taylor says. “You don't need a senior, a mid and a junior - you can just stick at sort of a mid-level, as an example.”
Taylor - who has been heavily involved with the marketing process since Step One’s initial launch - says while the company has reduced its marketing team, those remaining are being pushed into new ways of working.
This includes utilising generative AI programs such as ChatGPT and AI photo editing via Photoshop, which Taylor says has reduced the marketing output from days and hours to minutes and seconds.
In an example based on a recent campaign in collaboration with Surf Life Saving Australia (SLSA), he says his team produced ad copy and then put it into ChatGPT to generate up to 10 possible angles.
“So there'll be small iterations, and the difference these make could be 20-30% to the success of an ad,” Taylor says. “And that's just the ad copy.
“Traditionally, what you need is a copywriter who you'd get to write the different versions of copy, then we'd have to upload those all into Facebook. You have to wait a week, then you'd start to send out one or two. And then you'd add in a third, see which one's winning and vice versa.
“But we can put 10 in, and we can do this within literally a minute. Whereas that would take somebody an hour or two to think about it, create it, brainstorm the ideas… Here we can put those in and then within six hours, we can see which one's working, then we'll take that one, put it into AI, say this was the most effective, and it learns as you keep it in the same thread.”
This fast turnover of marketing content can also be used to make swift changes in imagery and advertising. In an example, Taylor says there may be four adverts, and his team can ask the Photoshop AI software to change the click buttons on each of them to various colours. Which it can do in seconds.
“Our ad server doesn't have to go back to the graphic designers, they'll download the image, put it in and say, Can you please change the ‘Shop Now’ button or the call-to-action button to four different colours? And then you'd get that back and see which ones of those work - because that can mean the difference of five or 10% more people clicking.”
It can also work for photographs, by removing unwanted things from real-life imagery.
“Obviously, we get efficiencies in the ad spend, but also efficiencies in the team,” Taylor explains. “So it means we can run a leaner team, and upskill people that potentially didn't have graphic design skills. So the ad servers are quite important because they can actually make the changes themselves now, without having any Photoshop experience.”
Taylor says some of its AI usage involves creating new advertising based on past successful campaigns, with its onboarded tools being able to reimagine an old ad in seconds.
“We spend about $30 million a year, and about 80% of that budget goes into digital,” Taylor says. “So for us to be able to make changes like this and optimise the ad itself, and do this in real-time - to do it in literally 10 minutes compared to two days means the efficiencies that we can get out of AI and digital marketing now are phenomenal.
“I certainly wouldn't be investing in ad agencies and things like that.”
Step One has prioritised a return to profitability over top-line growth this year, with aggressive cost reductions in areas such as inventory and advertising spend. In the 2022-2023 financial year, Step One spent $23.26 million in advertising and marketing. This was down from $32.08 million reported in 2022-2021.
The next area of AI integration is in Step One’s email marketing, which the company is preparing to target very specific cohorts within its customer base and automate the send-out process.
Through marketing software called Klaviyo - which utilises AI - the underwear etailer can add in key customer segments generated via AI-driven data analysis, as well as the aforementioned imagery and copy, and send it all off in small bursts to test performance.
For example, it can ensure that certain cohorts get sent advertising on desired products, all the way down to whether they prefer a certain colour of underwear or a desired amount.
“That would [usually] take two or three days to set up,” Taylor says. “Whereas now, it'll say, ‘Hey, are you okay for me to optimise this?’ It'll show you what it will do. You check it [to confirm], and then it just starts sending. But it is testing - so it won't send all the emails, it'll send 10% to see if it's working.
“And then if it's not reaching the delta - if we're not getting a +10% increase on the delta, then it will stop, it'll create more creative, and then run that test again."
However, marketing is just the beginning, with future plans to integrate AI technology into other areas of Step One, including stock forecasting and inventory management.
Taylor says the company is building out ways in which its operations team can “teach” AI software to forecast how many items of a particular product they would need, based on past sales data, and taking into account lead times.
“[For example], if we sell 100 pairs of black, large boxer briefs, how many do we need for three years? And here's the sales data for last year. Let's say it takes 180 days to get here, how many do I need to order each month? And [the software] starts to learn those things.”
In the end, Taylor does understand the fears of how AI could make certain roles redundant.
“If you're head of email and head of AB testing, and you click a button and turn AI on, and it does your job, and it does a little better, you're gonna be thinking, 'I don't really want to use that button too much'.
“Looking down the road, in five to ten years time, it's going to be interesting, where you look at what it's done in around the last 12 months. I've given you some examples now - and that's just the tip of the iceberg in how we use it.”