Seizing the generative AI moment
How four startups are assessing and harnessing the potential of this disruptive technology.
That data play set up Yanolja for this moment. Tapping generative AI for analytics, Yanolja is using that data to create better customer experiences by combining disparate services into a more cohesive, relevant landing page that provides tailored results for its customers.
“We made a decision to start the cloud-based solution business because it makes it possible to integrate all the data from the providers, and to realize new, innovative AI-based dynamic packaging from generative AI,” says Jong Yoon Kim, CEO of Yanolja and Yanolja Cloud.
With billions of data inputs from across its customer base—60,000 travel industry clients, from hotels and leisure facilities to airlines and travel agencies—Yanolja has built a data moat of travel information that allows it to outperform rivals. It’s the fuel that powers a generative AI engine that can offer a more unified experience for travelers and a profit-driver for providers, as it helps to reduce their dependence on the usual online channels, which typically take commissions of around 20%. The company’s goal now is to replace limited travel experience altogether with a dynamic, personalized experience built atop the travel providers’ data.
“The problem is that most of the travel platforms just provide one item, like a hotel booking,” says Kim. “Even when they provide a bundled package for travel, the data is not deeply integrated to revolutionize the overall travel experience.”
The plan isn’t to stop at transforming the customer experience at the time of booking. Kim also sees an opportunity to expand customer touchpoints as they travel. Yanolja has developed a generative AI-powered tool that keeps in touch with travelers throughout their journey, checking in with them and providing recommendations on activities they might enjoy. “It's ridiculous because it's a travel service, but nobody uses the travel service during the actual travel,” he says. “Now, generative AI can help.”
“AI has a place and it drives value in certain areas, but in certain areas, you don't need AI,” says Rishi Khosla, co-founder and CEO of OakNorth. Rules-based approaches, he adds, are sometimes sufficient.
But Khosla says that while generative AI is on his radar, the company has no immediate plans to integrate the technology into its core product. “Some technology is great from the get-go, some is exciting but wobbly,” he says. Because of problems with accuracy and reliability, he says today’s generative AI models are not ready for prime time in a credit intelligence business that faces compliance and regulatory requirements. That said, his engineering team keeps testing the waters through experiments, like a recent AI hackathon focused on how generative AI could be used to enhance the predictive industry forecasting model.
Outside of the core product, however, OakNorth is going all in on generative AI. The company is finding a wide range of generative AI use cases in marketing, sales, finance, and operations that can increase productivity and find efficiencies. In marketing, generative AI helps the company to create content such as press releases, corporate blog posts, and other company information. The technology’s application to content has turned out a surprising dividend. As inaccuracies come up in text generated by AI, it helps OakNorth identify misinformation about the company or its people on the internet. It can then track it down to make sure it’s updated.
“If you ask the generative AI to write a bio of an executive, you might get 90% or 95% of correct information,” says Khosla. “But maybe some of the information is incorrect and that's because somewhere, perhaps on page 27 of Google, there's something that's incorrect. And so it can actually be helpful in terms of discovering if there is some information out there that is not correct, and then you can go and find that and have it corrected.”
But one of the biggest impacts generative AI has had on OakNorth is through the changes it has sparked in the talent market. “Getting data analysts has always been hard,” Khosla says. “The interest in generative AI has made that harder.”
“We've been working in an industry that has traditionally been very manual, with humans and static documents, and we came in with a completely different approach, to be completely automated,” says Amper. Ironically, he adds, “what we're seeing right now is that there needs to be a human component in order for our models to learn.” With increased automation and more generative AI integrated into models, humans are being brought in to handle anomalies within edge cases, and as a final double-check, to continuously improve the models and help them learn. “Human input is key if you want to have high-quality data,” says Amper.
Generative AI is also playing a crucial role in training future models to accurately identify and verify individuals' information. Incode employees utilized a generative AI tool to generate simulated identities for training purposes, significantly reducing the time required compared to manual efforts. “Embracing these new technologies has already improved our capabilities, and they will continue to enhance both our work processes and product performance,” says Amper.
Its customer-facing offering, ParentingGPT, is a case in point. It’s a generative AI–powered chatbot that provides instant answers to parent queries regarding their children’s well-being. The answers draw on data from FirstCry Parenting, its community of experts and everyday people, and are annotated with references to professional content, pediatricians, and other experts in child development.
Behind the scenes, FirstCry is also putting generative AI to work on its core e-commerce platform in a pilot test to power product recommendations and improve semantic search. Maheshwari says that artificial intelligence has also been helpful in speeding up the steps involved in adding new products to the catalog, by copywriting and automating quality assurance functions. “It is helping us enhance some processes that are currently completely dependent on human skills,” he says.
Even as generative AI takes on a growing role across the business, Maheshwari is proceeding deliberately and with caution. “We’re starting with small AI projects to gain experience and test the waters,” he says. “All of our pilots in operations are around enhancing human output and productivity.”
How will he know when any of these tests are ready to come out of beta? Ultimately, the measure will be in whether customers are happy, he says. “We continuously evaluate the return on investment of our AI pilot projects. We measure their impact on customer satisfaction, operational efficiency, and revenue and bottom line impacts.” And when parents are happy, it’s not hard to tell.
