4 ways to turn generative AI experiments into real business value

Two years after OpenAI unveiled ChatGPT, generative AI (gen AI) has reached a point where industry experts openly question whether the emerging technology has been overhyped.

Gartner recently predicted at least 30% of gen AI projects will be abandoned after the proof-of-concept stage by the end of 2025. Such failings mean the analyst says the technology has reached the “trough of disillusionment” in its latest Hype Cycle for Emerging Technologies.

This slip into the innovation abyss is a big fall for grace for gen AI. Last year, chief executives of tech companies started their keynote conference speeches talking breathlessly about a wave of AI-enabled products.

Also: Gen AI as a software quality tool? Skepticism is fading as more organizations implement it

Today, it’s common to hear the same CEOs apologize for even mentioning gen AI. These tech bosses recognize that professionals are becoming disillusioned by gen AI. A technology with so much promise for boosting productivity is often stuck at the starting gate.

However, some executives are going beyond the exploration stage and helping their organizations embrace gen AI. Take Rahul Todkar, head of data and AI at travel specialist Tripadvisor, who told ZDNET the technology is a game-changer.

“I’m a believer in generative AI technologies,” he says. “There’s a lot of talk about, ‘Is this real or hype?’ I’m in the camp of, ‘The impact and the benefits of this technology are real.’ And we already see it in our company.”

Also: Technologist Bruce Schneier on security, society and why we need ‘public AI’ models

Todkar explained how Tripadvisor uses gen AI in production services to benefit its customers and the firm’s bottom line. 

1. Find a suitable challenge

Todkar said Tripadvisor recognized that planning a successful trip is a common pain point for travelers. “We’ve had a lot of different products and solutions in this space,” he said.

“But with gen AI, what we have been thinking is, ‘How can we reimagine that process from a lens of making it super-easy, delightful, and highly personalized for our travelers?'”

Also: Perplexity AI’s new tool makes researching the stock market ‘delightful’. Here’s how

Todkar said his team at Tripadvisor has been on a data-enabled mission to transform the customer experience.

They’ve taken the company’s huge stores of customer data and used gen AI, large language models, and internally developed recommendation engines to create personalized content for travelers.

As a result of this privacy-compliant development work is Tripadvisor’s Trip Planning Solution.”We’ve been able to launch, iterate, and perfect this tool over time in nine months,” he says.

“The adoption of this tool has been fantastic. We recently crossed one million-plus trips created by our gen AI solution, which is, in my mind, a massive milestone.”

Todkar said these adoption numbers are just one measurement of the tool’s success. The technology is also helping to boost customer experience and engagement — and the techniques that have been exploited help to point the way to further success.

“It’s been fascinating,” he said. “I think it’s a win-win for the business and our customers. It’s a great innovation, where we’ve married the latest technology with user experience.”

2. Use the right technology

A successful gen AI solution like Tripadvisor’s Trip Planning solution is just the visible tip of a much larger iceberg — and Todkar said his company’s underlying technology effort has focused on two areas: foundations and models.

“It all starts with having the data in the right place,” he said, referring to the company’s deployment of the Snowflake AI Data Cloud for Travel and Hospitality.”Many times, it’s easy to not pay enough attention to the foundational data required to make gen AI solutions successful,” said Todkar.

Also: The secret to successful digital initiatives is pretty simple, according to Gartner

“So, moving from experimentation to production starts with data — having all your data in one place, which is where we use Snowflake. That’s the underlying technology.”

Todkar said his data team uses those cloud-based foundations to build a “thoughtful approach” to gen AI that combines the best of big-name LLMs with internal development work.

Also: CIOs must also serve as chief AI officers, according to Salesforce survey

“We use of-the-shelf commercial models — which is OpenAI, we’ve used a lot of the GPT series of models — and then combine that technology with our internal recommendation system models, which we’ve patented and built over time,” he said.  

“We can create the best experiences for travelers by taking different models and packaging them in a way that makes contextual recommendations.”

3. Turn the use cases into revenue

The right use case for gen AI produces other benefits, too. Todkar said happy customers boost Tripadvsor’s revenue.

“We’ve seen that when travelers find the right recommendations — which are personalized to them and contextual to what they’re trying to do — the engagement with those specific recommendations is significantly different,” he said.”The technology provides up to three times more engagement — and that engagement can lead to increased stay on our site and more revenue.”

Also: Gartner’s 2025 tech trends show how your business needs to adapt – and fast

He gives the example of a family of four vegetarians, including two children, using the Trip Planning tool to receive personalized holiday recommendations.”If I’m traveling with a family of young children to Napa, and I’m asking for recommendations, wineries will be off-limits to me,” he said.

“So, help me recommend a place I can stay in Napa. And when I’m looking for my restaurant dining options, if I’m vegetarian, help me pick a place that works.” ‘Todkar said AI-enabled recommendations tailored to travelers’ needs lead to a big increase in engagement.’

Also: Generative AI doesn’t have to be a power hog after all

“What we’ve seen is our users love that level of personalization. That process is all backed by the massive amount of data we have,” he said.”We have tons of data in our forums where users chat, interact, and discuss travel with each other. All that insight is incorporated into our personalized recommendations.”

4. Make long-term innovation plans

Tripadvisor is exploring other use cases for generative AI. One of these cases, review summarization, has been in production for about six months. Todkar said the application illustrates how the company plans to do more with emerging technology.

“Let’s say there’s a fantastic pizza restaurant in Chicago with hundreds of reviews. Nobody has the capacity and time to read through all those reviews,” he said.”So, how do you summarize that with key, meaningful indicators, such as ‘Here’s the best thing about the restaurant, and here are the things to watch out for?'”

Also: 3 ways to build strong data foundations for AI implementation, according to business leaders

“We’ve done all that review summarization using generative AI and large language models.”

In the longer term, Todkar thinks generative AI can be used to create interactive and multimodal experiences for users. He said Tripadvisor has about half a billion images on its site, including user-submitted photos and pictures from property owners. These images can be used to create tailored recommendations and new levels of engagement and experience.

Also: 5 ways AI is changing the future of air travel

“We have a massive collection of photos on our site that we can use for models,” he said.

“If you’re looking for a restaurant, and you’re vegetarian, you don’t want to look at images of steak. So, the images we present can be personalized — and that’s the next frontier we’re tackling. It’s on our roadmap to embrace these types of opportunities.”