Most AI projects are abandoned - 5 ways to ensure your data efforts succeed

Wong Yu Liang/Getty

Almost two-thirds (63%) of organizations are unsure they have the right data management practices for AI, according to research firm Gartner. This lack of readiness has an impact: the analyst predicts 60% of organizations will abandon AI projects through 2026.

1. Expose people to new tools

AI is nothing new to the airline industry. Masters' organization has used AI and machine learning for many years, such as to predict load factors on a plane, including the likely number of passengers for a flight.

He said airlines have used emerging technology to analyze other operational considerations, such as monthly revenues, competitor strengths and weaknesses, and airplane reliability.

"The evolution of statistical analysis through AI has been gradual," he said, before suggesting an upward trend that will be common to other business leaders: "Now we notice that evolution keeps getting quicker and quicker."

Also: What are AI agents? How to access a team of personalized assistants

Masters said the key turning point was the release of ChatGPT in November 2022, which marked the start of easy access to chatbots and opened up the potential benefits of AI.

"More people across the organization, who weren't necessarily in the analytical teams, were realizing what they might be able to do with this technology," he said. "They started thinking about what AI could give them access to, how it could reinterpret and explain data in different ways, providing analytical concepts to parts of the business that might not normally have access to the same tooling before."

Masters and his team have taken responsibility for considering how professionals can use these tools to boost their productivity and the operational effectiveness of the business.

"We can push that approach to different people and expose insight through different tooling, such as bots that we've got on our data platform."

2. Answer key business questions

Encouraging people to make the most of AI is just one part of the challenge. The bigger issue is ensuring people use the right tools effectively.

Masters said the return on investment in some areas is clearer than others: "When it comes to dynamic pricing tools, such as our partnership with Fetcherr, or things like predictive maintenance or fuel prediction, you can quite easily do the maths for making or saving money. Those targets were easier to go after."

In other areas, such as improvements to decision-making processes, where the ROI is less clear, his team has focused on its Databricks data platform and considered how to react quickly to new technologies and models that emerge.

"We've been focusing on getting ready for the questions to come in, so that we can then provide the right algorithm or piece of data," he said.

Also: 4 ways to turn AI into your business advantage

Those areas include analyzing net promoter scores, digging into all the different survey results the airline receives, and combining that information with other data, such as customer opinions on in-flight experiences.

"We can start to combine that information now because we've got a platform in place, versus scrabbling for the data and putting it together," he said. "The platform is letting us answer lots of small questions without having to have this huge prioritization exercise."

3. Establish a unified approach

Pressure to dabble in AI comes from all directions, including tech vendors who push their latest AI-enabled systems and services.

"That creates a lot of noise," said Masters, who added that his organization has established a process to assess new suggestions for AI tools. "Given our ability to talk across the silos, we can say, 'You don't need this AI module. We've already got this tech coming to the platform.'"

Also: You've heard about AI killing jobs, but here are 15 news ones AI could create

Masters explained how the process works. Every week, a group of specialists, including architects, business analysts, and the central product management office, explores technology proposals.

"That's the front door for technology," he said. "Then we've got me and our VPs for technology, digital, and data who assess tools at the higher level."

These discussions about AI-enabled services can also include other senior executives if further clarification or prioritization is required.

"Our VP of customer journeys and engineering, for example, might look at how this technology helps them achieve part of our wider business strategy," he said. "That joined-up approach helps us to allocate the right resources to the project."

4. Exploit your platform

Airlines traditionally hold customer and operational data in separate systems. This disparate approach to data collection and storage makes it tough to run cross-business initiatives.

Virgin was eager to create a consolidated approach to enterprise information and has brought its information together via Databricks' Unity Catalog. Masters said this approach makes it much easier to generate insight.

"We can run various simulations with the data and run different predictive models on the platform that helps us be nimble and agile," he said. "If you get a disruption, you can start to ask questions, such as, 'How many connecting passengers do we have at this airport? What can we do for these connecting passengers? Are they going to be tight on time for that connecting flight, or are they going to be fine?'"

Also: AI won't take your job, but this definitely will

That insight helps the airline operationally as the organization can move dynamically, changing teams on the ground to assist with passengers' requirements.

"Our teams can go down to the lowest level of granularity about what's happening every day on every flight," Masters said. "You can roll that insight up in whatever dimension you need, which is pretty revolutionary for us."

5. Keep fostering curiosity

The key lesson Masters has learned from these explorations into data and AI is that digital leaders should spend less time tinkering with technology and more time understanding what the business does.

"Data teams can get a little bit stuck in the world of the platform and tooling," he said. "However, it's becoming more the case that the specifics of the tooling, or building a better database here or there, isn't so important anymore. Data automatically appears on the platforms today, and they're becoming easier to connect."

Also: 5 ways you can plug the widening AI skills gap at your business

The result of this higher level of integration is that technology is less of a day-to-day priority for data leaders.

"The focus, instead, becomes being curious about the organization you're in, and thinking about what to do," he said. "You can start to remove a lot of the noise and get to what's important. You can spend a good chunk of your time articulating those priorities back through your teams to make them curious."

Get the morning's top stories in your inbox each day with our Tech Today newsletter.