The AI revolution that vendors have been touting may not be as transformative as it seems. Recent studies indicate that while companies are highly interested in adopting generative AI, turning that interest into a reality is proving to be challenging. The main obstacles companies face are estimating and demonstrating value, as well as a lack of talent. According to a survey by Gartner, these issues are the top barriers to implementing AI solutions.
Another study by LucidWorks found that only 25% of respondents reported successfully implementing generative AI projects. Senior partner Aamer Baig from McKinsey and Company shared similar findings at the MIT Sloan CIO Symposium, revealing that only 10% of companies are implementing generative AI projects at scale, with only 15% seeing any positive impact on earnings. This suggests that the current hype around AI may be ahead of the reality experienced by most companies.
The complexity of implementation is a major factor slowing down companies. Baig emphasizes that even a simple project requires multiple technology elements, including the right LLM, proper data and security controls, and new capabilities like prompt engineering and IP controls. Ancient technology stacks can also hinder progress. Baig explains that too many technology platforms can be an obstacle to achieving generative AI at scale.
To successfully execute AI projects, organizations must address their readiness deficit, which includes data challenges. According to the Gartner survey, 39% of respondents expressed concerns about a lack of data as a top barrier to successful AI implementation. Baig advises companies to focus on a limited set of data that can be applied to multiple use cases, prioritizing high-priority business challenges with business value.
Data governance is another crucial aspect to consider. Akira Bell, CIO at Mathematica, highlights the need for caution when using data for generative AI. Companies must respect where the data comes from and whether they have permission to use it. Bell emphasizes the importance of being a trusted data steward and maintaining strong governance practices.
CIOs are naturally cautious, similar to when the cloud was emerging. While they recognize the potential of generative AI, they must also ensure governance, security, and measure ROI. Sharon Mandell, CIO at Juniper, states that measuring return on generative AI investment is challenging, and pilots are being run to determine if there is a true productivity increase to justify the costs.
Baig suggests a centralized approach to AI across the company and avoiding too many independent skunkworks initiatives. Having the proper company structure and top management visibility is crucial for success. It’s important for teams to avoid trying to do too much and focus on reusing what works to deliver impact and keep businesses happy.
Although there are obstacles to overcome, companies should not be paralyzed by challenges related to governance, security, and technology. It’s essential to start with something that works and shows value and then build upon that. Companies should also remember that they are not alone in their struggles with AI implementation.