A: In study after study, consumers tell us that the single most important factor influencing their purchases is informative, accurate product descriptions, which makes sense because people don’t buy what they don’t understand. And this is true both online and offline. A recent Google study found that 87% of consumers do online research before ever visiting a store.
A: Because of how much time and resource writing product content requires, companies are desperately searching for new solutions to accelerate this process.
In my previous experience as an applied machine learning researcher at Amazon, I worked on deep learning techniques for NLP and saw how quickly the field was advancing. On the other hand, current platforms that tried generating product descriptions either require a lot of pre-written templates beforehand – resulting in unoriginal content, or can be difficult to set up without tedious amounts of input and training, but we knew we could do better — this is the gap we're solving for.
Hypotenuse can learn from your existing data, with human supervision, and that data doesn't have to be structured or processed in any particular way. You don't have to invest in the natural language generation technology that exists today, which requires a huge initial and ongoing investment. Hypotenuse trains on your existing data in a matter of days, not months, and consistently learns from your new data.
Another gap is that current solutions are either generic, or try to force-fit words into sentences that sound unnatural. This hurts both your brand voice and SEO—which is what sets you apart from your competition. That's where Hypotenuse comes in. We offer something that is tailored to your voice and content needs, all unique and human-readable.
A: I knew AI was going to be the future so I’ve been hunting for an AI partner for quite some time. I got really excited when I read about the power of Natural Language Processing (NLP) and GPT-3. But I was quickly deflated when I tried other writing tools built with GPT-3. I couldn’t use them as an e-commerce content solution for Talkoot’s customers because they couldn’t be trained to channel a specific brand voice.
Any market-leading brand knows that its voice is one of its greatest brand assets. For consumers, a generic, commoditized product description signals they’re buying a generic, commoditized product. We work with many of the world’s most loved and respected brands. For us, it didn’t make sense to offer them an AI solution that made their products sound like commodities.
With Hypotenuse, we have the ability to create AI writers that can channel a specific brand voice. In fact, we can spin up as many AI writers as a brand needs. If you sell to multiple audiences, like runners, basketball players and hikers, you might want an AI writer that speaks authentically to each. Likewise, if you have several very unique brand franchises like Disney does with Frozen, Marvel, Star Wars, etc., you likely need a writer that can speak authentically about each of those.
We’ve also built in an automated feedback loop so the more you use the AI, the better it gets at channeling your brand language.
On top of all that, the Hypotenuse team is incredibly smart, kind, humble and fun to work with.
A: Our models can be trained on a per-brand basis and is a process that gets down to the nitty-gritty, such as the use of a passive oran active voice, adding in specific SEO keywords or pivoting away from a list of words to avoid.
As the brand uses us, our AI continues to learn and adapt based on the data it collects, whether by positive reinforcement or to correct certain mistakes. This gives brands the ability to customize content that is specific to their voice, ensuring quality written content.
A: When I was the global writing director at Adidas, I realized a brand is a complex system of nested stories—from a company’s reason for being, to its product categories, to its collections and collaborations, to the individual products and technologies. And all those stories need to stay consistent and cohesive, one building upon another.
Even for brands that understand their story and are producing amazing products, it’s hard to keep that story consistent and powerful all the way down to the e-commerce product page. There are too many people, too many deliverables and it’s all moving too quickly. Even when everyone is doing great work, how you talk about your products can still come across as schizophrenic and disjointed to the end consumer.
That’s why I founded Talkoot. Talkoot is that one place—from product inception to product launch—where everyone working on your product copy can log in, see what’s been written before, including all the feedback, input and images, and add their own chapter to the evolving story of that product.
A: Paraphrasing the immortal words of Simon Sinek, people don’t buy ‘what’ you do, they buy ‘why’ you do it. Product descriptions describe what you’re selling. Product stories let people know why it matters. Who created it? What inspired them? Who was it built for? What problem does it solve? How is it built? Ninety-nine percent of the value consumers see in your product comes through in the backstory.
In a study by market research firm Origin, shoppers were 5% more likely to purchase from a product page with the winemaker’s story in addition to a standard description. They were also willing to pay 6% more for that bottle of wine.
Origin saw a similar trend within other product categories. For instance, shoppers were willing to pay 11% more for a painting when the artist’s story was included on the PDP and 5% more for a hotel room when the product page featured stories about real guests.
Joshua: We have a training phase where we gather all the information we need about the brand — style guidelines, brand books and everything their marketing team uses to get their collateral and content consistent, as well as copy that they have written before.
We get our models to learn their content needs and writing tone. After working together with many brands, we’ve managed to tie down this process so that the whole phase takes just 3-4 weeks. From there, brands can already start having copy generated for their products — but that’s not where it stops. We provide ongoing support and improvements, and as brands use the tool, the model continues to evolve and get more tailored to them, further enhancing the copy that is generated for them.
Brian: As Joshua said, it takes 3-4 weeks to create an AI writer that understands your brand language. But it won’t take 3-4 weeks of work on the client’s end. Mostly the work happens on our end mapping your data flows, gathering input and training the model. Brands provide feedback on the product copy the AI generates. It will also take 1-2 training sessions to get your team trained on Hypotenuse’s integration on Talkoot.
Interested to learn more? Sign up for a demo with Hypotenuse AI right now to elevate your e-commerce presence.
You can follow Talkoot on LinkedIn.