Vendors want us to consider that we’re within the midst of an AI revolution that’s fundamentally changing the best way we work. But the reality is, in line with several recent studies, that the matter is rather more nuanced.
Companies are extremely keen on generative AI as vendors point to potential advantages. But translating that desire from a proof of concept right into a working product is much tougher: They run into the technical complexity of implementation, whether as a result of technical debt from an older technology stack or just an absence of staff with the appropriate skills.
In fact, a recent Gartner study found that the largest obstacles to implementing AI solutions are an absence of ability to properly assess and reveal value (49 percent) and an absence of talent (42 percent). These two elements could prove to be key obstacles for corporations.
Consider this a study by Lucidworksan enterprise search technology company, found that just one in 4 respondents said they’d successfully implemented a generative AI project.
Aamer Baig, senior partner at McKinsey & Company, said on the MIT Sloan CIO Symposium in May that his company had current survey that only 10% of corporations are implementing generative AI projects at scale. He also reported that only 15% saw positive impacts on revenue, suggesting that the hype could also be way ahead of the fact most corporations are experiencing.
Why the delay?
Baig sees complexity because the foremost factor slowing corporations down, as even a straightforward project requires 20 to 30 technology elements and the appropriate LLM is just the start line. You also need things like proper data and security controls, and employees may have to learn recent skills like prompt engineering and implementing IP controls, to call just a few.
Outdated technology stacks may hold corporations back, he says. “In our survey, one among the largest barriers to implementing generative AI at scale was actually the sheer variety of technology platforms,” Baig says. “It wasn't the use case, it wasn't the information availability, it wasn't the trail to value; it was actually the technology platforms.”
Mike Mason, chief AI officer at consulting firm Thoughtworks, says his company spends lots of time preparing corporations for AI — and their current technology setup is an enormous a part of that. “So the query is, what’s your technical debt, what’s your deficit? And the reply is at all times going to be: it will depend on the organization, but I believe organizations are increasingly feeling this,” Mason told TechCrunch.
It starts with good data
A big a part of this lack of readiness is as a result of data. 39% of respondents to the Gartner survey expressed concern that an absence of knowledge is the largest obstacle to successful AI implementation. “Data is a large and daunting challenge for a lot of, many organizations,” said Baig. He recommends specializing in a limited dataset and keeping reuse in mind.
“A straightforward lesson we've learned is to really give attention to data that helps you with multiple use cases. In most corporations, that's normally three or 4 areas which you can actually start with and apply to your necessary business challenges with business value and deliver something that may actually get into production and scale,” he said.
Mason says an enormous a part of successfully implementing AI is expounded to data readiness, but that's only a part of it. “Companies quickly realize that usually they should put some work into AI readiness, some platform constructing, data cleansing, all of those things,” he said. “But you don't must take an all-or-nothing approach, you don't must wait two years before you possibly can see value.”
When it involves data, corporations also have to respect where the information comes from – and whether or not they have permission to make use of it. Akira Bell, CIO at Mathematica, a consulting firm that works with corporations and governments to gather and analyze data related to varied research initiatives, says her company must tread fastidiously in the case of using that data in generative AI.
“When we have a look at generative AI, there are actually opportunities for us, and once we have a look at the ecosystem of knowledge that we're using, we’ve got to watch out about that though,” Bell told TechCrunch. That's partly because they’ve lots of private data with strict data use agreements, and partly because they're sometimes coping with vulnerable populations and wish to concentrate on that.
“I got here to an organization that takes the role of trusted data steward really seriously, and in my role as CIO I actually have to be very rooted in that, each from a cybersecurity perspective and by way of how we treat our customers and their data, so I understand how necessary governance is,” she said.
She says that without delay, it's hard to not be enthusiastic about the probabilities that generative AI brings; the technology could offer her company and its clients significantly higher ways of understanding the information they collect. But it's also her job to proceed fastidiously without getting in the best way of real progress – a difficult balancing act.
Finding the worth
Much like a decade and a half ago when the cloud emerged, CIOs are naturally cautious. They see the potential that generative AI brings, but in addition they have to worry about staple items like governance and security. They also have to see an actual ROI, which is usually difficult to measure with this technology.
In a January TechCrunch article on AI pricing models, Juniper CIO Sharon Mandell explained that measuring the return on investment (ROI) of generative AI investments is difficult.
“In 2024, we'll be testing the GenAI hype because if these tools deliver the promised advantages, the ROI is high and may also help us eliminate other things,” she said. So she and other CIOs are running pilot projects, proceeding cautiously and trying to search out ways to measure whether there may be actually a productivity increase that justifies the additional cost.
Baig says it's necessary to take a centralized approach to AI across the corporate and avoid “too many skunkworks initiatives” where small groups work independently on various projects.
“You need the framework of the corporate to be sure that the product and platform teams are organized, focused and dealing quickly. And in fact you wish the visibility of senior management,” he said.
None of this can be a guarantee that an AI initiative might be successful or that corporations will find all of the answers immediately. Both Mason and Baig said it's necessary that teams don't try too hard, and each emphasize reusing what works. “Reuse leads on to faster execution, which makes your corporations pleased and creates impact,” Baig said.
However corporations implement generative AI projects, they mustn’t be paralyzed by governance, security and technology challenges. But they also needs to not be blinded by the hype: There might be loads of obstacles for virtually every organization.
The best approach could be to create something that works and provides value and construct on it. And keep in mind that despite the hype, many other corporations are struggling too.
