Know before you buy: Designing a more transparent shopping experience
When prompted, Chat GPT advised us against drawing parallels between Chat GPT and Covid-19.
One is a global pandemic, it said; the other is a technology tool developed to facilitate conversation and provide information to users.
But the two “fundamentally different entities” have wrought a similar impact on our collective, human psyche. Each made headlines first as a whisper then as a tsunami; each holds massive implications for how we work and live; and each has taught us hard lessons in flexibility.
Everyone should be ready to pivot. This is particularly clear in respect to all things artificial intelligence. Even experts working with large language learning models have to stay on their toes.
Case in point — our client, Ethica AI, needed to pivot earlier than expected to adjust its business model when Open AI announced that plugin developers on the company’s waiting list could use its documentation to create a plugin for ChatGPT.
The initial idea behind Ethica AI was to create a platform that would help consumers quickly find products that met their unique needs and values. For example, someone shopping for sunscreen might make purchase decisions based on the absence of certain chemicals, whether the product is tested on animals, whether it’s reef safe, safe for sensitive skin, cost, and how quickly it would ship. The platform would serve up an item that met those qualifiers — and nothing else.
Now, that might sound like Amazon, but consider the last time you looked for something to buy from the e-commerce giant. Even with filters applied, hundreds of options appear, and many of those items are “promoted,” complete with potentially fraudulent reviews.
Ethica AI wanted something they could publish on a browser-based website that would let people interact with different products, understand details about them, ingest all of the metadata that’s scraped off the internet, and help consumers buy without hassle. This solution would use new AI technologies to generate custom product recommendations based around interactions with a large language model.
Our team spent four weeks doing user testing to inform the development of a product detail page that would map legacy systems to people’s pattern recognition, resulting in a familiar customer journey flow, as follows:
Someone wants to view certain products, they click into those products, get served info about those products, and buy those products.
When the Open AI plug-in announcement hit, Ethica AI realized the opportunity for them was not to build a standard ecommerce platform but rather to create a shopping plug-in for Chat GPT.
For lead designer Ted Schaller the immediate challenge was clear.
“We needed to explore how we could utilize their same model and technique for product scoring and translate that into something that would be an inline interaction in Chat GPT,” he says.
Our user testing also revealed the need to add an educational component to the final product. While many people will actively seek out sunscreen, for example, that’s reef-safe and free of parabens, others might not think to look for those factors in their SPF. By working with Ethica AI’s data scientists, we leaned more heavily into the large language model, scraping the internet for expert resources to generate product guides (think J.D. Power or Wirecutter lists — trusted sources of information that narrow down the best choices for your product decision).
The result is the following experience:
Individual searches for sunscreen. Ethica AI serves up sustainable, non-toxic options along with an article explaining why Ethica AI thinks those are good products for that individual. If that person shares that they’re allergic to a particular ingredient, the next time they come back to look for face wash, Ethica AI takes that into consideration to deliver increasingly relevant recommendations and expert insights.
Ultimately, our final deliverable was better suited for a market that looked very different from the one in which we started our engagement with Ethica AI.
Our ability to shift gears on demand is thanks, in part, to our approach to resourcing, which is rather uncommon in the agency world.
“Sometimes agencies will have multiple designers on a project or a designer on multiple projects. We are usually assigned to projects on a 1:1 ratio, which meant my entire time and focus was dedicated to this client,” explains Ted.
This is helpful not only for startups navigating fast-changing emerging technologies like AI, but also for enterprises that face equally challenging external and internal dynamics.
“All it takes is one stakeholder to shift direction and your entire project gets flipped over,” says Ted. “We’re well suited to help clients pursue alternative paths.”
As a studio, we’re also staying ahead of the curve by leaning into systems design, because that’s what we’re designing around — how people communicate with AI in conversational or chat-like mode. For Ethica AI, we delivered sample building blocks to inform their model and to help the model understand the structure it should follow.
We’re optimistic about AI. We are actively incorporating it in our design process and can help clients apply it to their businesses.
“We should all be playing with these tools, interacting with them, figuring out how they work, and bringing that kind of thinking into their design process,” says Ted. “Because just as much as we know how the internet works or how CSS works, and how to use it in our standard design processes, we need to have an understanding of how AI technology works.”
Plus, there are many applications for Chat GPT that many businesses might not consider. We encourage our clients to think about how that technology could apply to the problem spaces we’re working with.
Ultimately, we’re here to be partners in design, making it so clients can focus solely on the technology or products that they’re building.
We live in an era defined by the unexpected. Before 2020, most Americans didn’t think about social distancing or mask-wearing as a public safety precaution. Cities and states could generally prepare for extreme weather events knowing that wildfires or hurricanes tended to impact certain regions more than others.
And with the right understanding of artificial intelligence, a startup could develop an AI-based tool without the underlying technology quickly shifting and upending their business model.
But few things are so certain anymore.