Vendors would like you to believe that we are in the midst of an AI revolution, changing the very nature of how we work. But some recent studies suggest the truth is much more nuanced than that.
Enterprises are excited about generative AI as vendors tout the potential benefits, but turning that desire into a proof-of-concept and working product is proving to be much more difficult. Whether it's due to technical debt from an old technology stack or simply a lack of talent with the right skills, companies are faced with the technical complexities of implementation.
In fact, a recent Gartner survey found that the biggest barriers to implementing AI solutions were finding a way to estimate and demonstrate value (49%) and a lack of talent (42%). These two factors can be major roadblocks for businesses.
A study by enterprise search technology company LucidWorks found that only one in four people surveyed reported successfully implementing a generative AI project.
Speaking at the MIT Sloan CIO Symposium in May, Ammar Baig, a senior partner at McKinsey & Company, said his firm's recent research found that only 10% of companies have implemented generative AI projects at scale, and only 15% have reported seeing a positive impact on their bottom line, suggesting that the hype may be far outweighing the reality most companies are experiencing.
What's the delay?
Baig believes complexity is the main factor slowing down companies' growth. Even a simple project requires 20-30 technical components, and a good law degree is just the start. You also need proper data and security controls, and employees may need to learn new competencies, like how to implement rapid engineering and IP management.
And outdated tech stacks can hold companies back, Baig says. “Our research pointed out that one of the biggest obstacles to achieving generative AI at scale is actually having too many technology platforms,” Baig says. “The problem wasn't the use case, the availability of data, or the path to value, it was actually the technology platform.”
Mike Mason, chief AI officer at consulting firm Thoughtworks, says his firm spends a lot of time preparing companies for AI, and the current tech posture is a big part of that. “So the question is: how much technical debt do you have? How much of a deficit do you have? And the answer is always: It depends on the organization, but I think organizations are starting to feel this pain more and more,” Mason told TechCrunch.
It starts with good data
A big factor in this lack of preparation is data, with 39% of Gartner survey respondents saying they fear a lack of data as the biggest obstacle to successfully implementing AI. “Data is a huge challenge for so many organizations,” says Baig, who recommends focusing on limited data sets with reuse in mind.
“The simple lesson that we've learned is to really focus on data that serves multiple use cases, and that's typically going to be three or four domains for most enterprises, and really be able to hit the ground running and apply it to high-priority business challenges that have business value and deliver something that can actually get to production and scale,” he said.
Mason says that successful execution of AI has a lot to do with data preparation, but it's only part of the story. “Organizations often quickly realize that they have to do all that AI prep work, platform building, data cleansing, all that work,” he says. “But you don't have to take an all-or-nothing approach and it doesn't have to take two years to get value.”
When it comes to data, companies also need to respect where it comes from and whether they have permission to use it. They need to tread carefully when it comes to leveraging that data for generative AI, said Akira Bell, CIO of Mathematica, a consulting firm that works with companies and governments to collect and analyze data related to various research activities.
“When you look at generative AI, it's certainly a possibility for us, and when you look at the whole ecosystem of data that we use, we have to do it carefully,” Bell told TechCrunch, partly because we have a lot of personal data with strict data usage agreements, and partly because we're dealing with sometimes vulnerable populations and need to be aware of that.
“I came to a company that is serious about being a trusted steward of data, and my role as CIO requires me to really understand that, not just from a cybersecurity standpoint, but from how we handle our customers and their data, so I know how important governance is,” she said.
Right now, she says, it's hard not to get excited about the possibilities that generative AI offers — the technology could offer a much better way to understand the data her organization and its clients are collecting — but it's also her job to tread carefully so as not to impede real progress, a tricky balancing act.
Finding value
Just as when the cloud emerged 15 years ago, CIOs are understandably cautious: They understand the potential that generative AI brings, but they also need to consider basics like governance and security, and they need to understand the real ROI that this technology can offer, which can be difficult to measure.
In a January TechCrunch article about pricing models for AI, Juniper CIO Sharon Mandell said measuring the return on investment in generative AI is proving difficult.
“In 2024, we'll test the genAI hype because if these tools can produce the kinds of benefits they say they can, their ROI will be high and they could help eliminate others,” she said. So she and other CIOs are running pilots, treading carefully, and trying to figure out how to measure whether the productivity gains are really big enough to justify the increased costs.
Baig says it's important to adopt a focused approach to AI across the company and avoid what's known as “too many skunk works efforts” — small groups working independently on multiple projects.
“You need footholds from the company to really make sure that product teams and platform teams are organized, focused and working quickly, and then of course you need executive visibility,” he said.
None of this is a guarantee that AI initiatives will be successful or that companies will have all the answers right away. Both Mason and Baig stress that it's important for teams to avoid trying to do too much, and to reuse what works. “Reuse directly translates to speed of delivery, keeping companies happy and effective,” Baig says.
No matter how your company approaches a generative AI project, don't be overwhelmed by the challenges around governance, security, and technology. But don't be fooled by the hype: There are plenty of obstacles for nearly every organization.
The best approach may be to start with something that works and demonstrates value, and build from there — and remember that despite the hype, many other companies are also struggling.