Vendors would have you believe that we are in the midst of an AI revolution that will change the nature of how we work. But several recent studies suggest that the truth is much more nuanced than that.
Enterprises are very interested in generative AI as vendors pursue its potential benefits, but translating that desire from proof of concept to a working product is much more difficult. Whether it's due to technical debt or not, you're facing technical complexity of implementation. Either they come from an outdated technology stack, or they simply lack people with the right skills.
In fact, a recent study by Gartner found that the two biggest barriers to implementing AI solutions were finding ways to estimate and demonstrate value, and (49%) lack of talent. These two factors can be major obstacles for businesses.
A study by enterprise search technology company LucidWorks found that only one in four of those surveyed reported successfully implementing a generative AI project.
Speaking at the MIT Sloan CIO Symposium in May, Aamer Baig, senior partner at McKinsey and Company, said his firm's recent survey found that only 10% of companies were implementing generative AI projects at scale. He also reported that only 15% had a positive impact on profits. This suggests that the hype may be far ahead of the reality that most companies are experiencing.
What is it?
Baig sees complexity as a major factor slowing down companies, even for simple projects that require 20 to 30 technical elements, and the right LLM is just a starting point. Additionally, things like proper data and security controls are needed, and employees may need to learn new capabilities, such as how to implement rapid engineering and IP controls, among other things.
He says outdated technology stacks can be holding companies back. “Our survey found that one of the biggest obstacles to achieving large-scale generative AI was actually too many technology platforms,” Baig said. “This was not a use case, it was not a data availability, it was not a path to value. It was actually a technology platform.”
Mike Mason, chief AI officer at consulting firm Thoughtworks, says his company is spending a lot of time getting companies AI-ready, and their current technology setup is a big part of that. “So the question is, how much technical debt is there, how much is the deficit? The answer is always this: It depends on the organization, but I think organizations are feeling the pain more and more,” Mason told TechCrunch.
It starts with good data
The biggest part of the lack of readiness is the data piece, with 39% of respondents to a Gartner survey expressing concerns about lack of data as the biggest obstacle to successful AI implementation. “Data is a huge and challenging challenge for many organizations,” Baig said. He recommends focusing on limited data sets with reuse in mind.
“The simple lesson we learned is to focus on data that actually serves a variety of use cases. And typically for most companies, that's three or four domains that they can really start with and apply to their high priority areas. It solves business challenges for business value and delivers something that actually reaches production and scale,” he said.
Mason says a big part of being able to successfully implement AI has to do with data readiness, but that's only part of it. “Organizations quickly realize that, in many cases, they need to do all sorts of things: AI prep work, building platforms, cleaning data, etc.,” he said. “But you don’t have to take an all-or-nothing approach. You don’t have to take two years to get any value.”
When it comes to data, companies must also respect where the data comes from and whether they have permission to use it. Akira Bell, CIO of Mathematica, a consulting firm that works with companies and governments to collect and analyze data relevant to a variety of research initiatives, says companies must tread carefully when applying that data to generative AI.
“If we look at generative AI, it's certainly a possibility, and we'll look at the data ecosystems we use, but we have to do it cautiously,” Bell told TechCrunch. This is partly because we hold a lot of personal data with strict data use agreements, and partly because we are dealing with sometimes vulnerable populations and we need to be aware of this.
“I come to a company that takes being a trusted data steward seriously. As a CIO, I need to be grounded in this, not only from a cybersecurity standpoint, but also from a customer perspective and how I deal with them. “We know how important governance is because we have the data,” she said.
She says it's hard not to be excited about the possibilities that generative AI brings today. This technology can provide organizations and their customers with a much better way to understand the data they collect. But it is also her job to move cautiously without impeding real progress – a challenging balancing act.
Find value
Just as they were when the cloud emerged 15 years ago, CIOs are understandably cautious. They see the potential that generative AI brings, but they also need to take care of fundamentals like governance and security. You also need to see actual ROI, which is difficult to measure with this technique.
In a January TechCrunch article on AI pricing models, Juniper CIO Sharon Mandell said measuring the return on investment in generative AI is difficult.
“In 2024, we will be testing the hype of genAI, because if these tools can produce the types of benefits they talk about, the ROI on them is high and could help weed out others,” he said. she said So she and other CIOs are moving cautiously as they run pilots, trying to find ways to measure whether there are real productivity gains that can justify increased costs.
Baig says it's important to have a centralized approach to AI across the company and avoid “too many skunkwork initiatives” where small groups work independently on multiple projects.
“You need a foothold in your company to really make sure your product and platform teams are organized, focused and working at pace. Of course, you need visibility from the top management,” he said.
Nothing guarantees that AI initiatives will be successful or that companies will find all the answers right away. Mason and Baig both said it's important for teams not to try to do too much, and they both emphasize reusing what works. “Reuse leads directly to delivery speed, keeping our business happy and providing leverage,” Baig said.
But even as companies launch generative AI projects, they should not be paralyzed by issues related to governance, security, and technology. But they shouldn't be blinded by the hype. Almost any organization will have many obstacles.
The best approach is to move forward with what works, shows value, and can build from there. And remember, despite all the hype, many other companies are struggling too.
