“I wish more brands would take more small swings to see what kinds of new experiences, products, interactions they can create [with generative AI],” said Jenny Nicholson, founder at creative AI agency Queen of Swords. The agency worked with wedding planning site Zola to develop its Split the Decisions chatbot last year.
Zola didn’t set out to create an AI chatbot. It aimed to solve a problem in wedding planning, and generative AI happened to be a useful tool for that solution.
“Everything that we do comes from trying to address a consumer pain point,” said Allison Cullman, vice president of brand marketing and strategy at Zola. “We identified the pain point first and then tried to think about creative solutions that would allow us to address this in a way that was accessible.”
For Zola, accessibility meant offering consumers an interface they could understand and a product anyone could use.
“One of the biggest problems with AI and large language models and chatbots in particular, is they expect a lot of the user,” said Nicholson.
Only 1 in 3 US adults will use generative AI this year, per our forecast, and many of those who do use generative AI will not be experts.
Rather than opening on a blank box, Zola’s chatbot leads with specific questions and points the user toward a specific path. The product is still customized, but because it is tailored for a specific use case, it is easy for novice AI users to interact with.
Use of a creative generative AI tool should lead the user somewhere like a purchase on a retailer’s website or further engagement with the brand. Zola’s chatbot offers users a downloadable CSV with a breakdown of wedding planning tasks, assigned owners for each task, and links to more resources on Zola’s website.
“It was a new way to get some of these tools and these articles and these resources in front of people who maybe wouldn't know about them otherwise,” said Nicholson. As a result of that actionable framework, Zola saw an increase in traffic on its site, per Cullman.
AI can produce biased results, which poses a risk to brands implementing generative AI, the way it has for Meta and OpenAI. Rather than shying from the tech, Zola set out to train this bias out of its chatbot. “We wanted it to be really inclusive. We wanted it to make sure that it wasn't just sort of giving the guy the ‘guy stuff’ and giving the lady the ‘lady stuff,’” for heterosexual couples, Nicholson said.
For Zola, the solution to this problem was testing and retraining. The team tested varying combinations of genders and names until it was satisfied with the results.
This was originally featured in the EMARKETER Daily newsletter. For more marketing insights, statistics, and trends, subscribe here.