As the chief research officer at IDC, I lead a global team of analysts who develop research and provide advice to help our clients navigate the technology landscape.
Back in 2023, at the CIO 100 awards ceremony, we were about nine months into exploring generative artificial intelligence (genAI). We were full of ideas and possibilities. Fast forward to 2024, and our data shows that organizations have conducted an average of 37 proofs of concept, but only about five have moved into production. It’s been a year of intense experimentation.
Now, the big question is: What will it take to move from experimentation to adoption? The key areas we see are having an enterprise AI strategy, a unified governance model and managing the technology costs associated with genAI to present a compelling business case to the executive team.
Our research indicates a scramble to identify and experiment with use cases in most business functions within an enterprise. The challenge is that each function within an organization might identify five or six use cases. When you look across the entire organization, the organization may be running dozens of disconnected use cases. To determine which ones will impact the business, organizations must develop an enterprise use case roadmap that prioritizes the highest impact use cases AI use cases. We encourage leaders to look for “AI super use cases” — those that will deliver the most significant business outcomes for the investment. Consider which ones will make your organization more resilient, will support overall organizational health and key goals such as innovation, adaptability or sustainability.
This approach requires a partnership between business and IT. Our data shows that nearly 40% of organizations don’t have close collaboration between these two areas, which makes it harder to move use cases into production. It’s time to get organized, partner with the business and create that enterprise use case roadmap.
Build or buy?
Another area where enterprises have gained clarity is whether to build, compose or buy their own large language model (LLM). The path of least resistance is to purchase genAI capabilities through existing applications. However, not all use cases are being addressed in commercially available apps. More will be in the future, but for now, most organizations need to do some composing. This involves grounding a commercially available or open-source LLM with your own data. Our research shows that very few organizations are building their own LLMs. A year ago, many thought they had to, but now they recognize there are other options.
Another realization enterprises had is just how important data is to AI initiatives, especially those composing their AI services. Organizations are finding they have outdated data or incomplete data sets. Companies tend to invest heavily in the data plane — where data is stored, organized and managed. Now, they need to invest in data engineering to prepare data for grounding and fine-tuning their AI models.
Predictions around future growth and concerns for AI
There is a lot at stake for organizations. AI will reshape enterprises and industries. Within the enterprise, AI will act as an assistant, advisor, agent or all three, changing business processes, applications and daily work tasks. Industries will innovate, engage customers and deliver value in fundamentally new ways. But without a vision and enterprise AI strategy, backed with a use case roadmap and strong business cases, this cannot be recognized. We expect some organizations will make the AI pivot in 2025 out of the experimentation phase. In doing so, they will begin recognizing the exponential benefits of their collective AI use cases starting in 2027. For those organizations that do not pivot in 2025, their experimentation phase will slip into 2026 as they fall behind their competitors. The difference between these two paths will be significant, impacting productivity gains, speed of innovation, customer relationships and financials. It’s crucial to keep moving forward on this journey.
The good news for CIOs is that you have an opportunity to take a leadership role with AI, especially as organizations mitigate risk by keeping AI model development centralized. Don’t forget to consider the support employees will need to adopt AI and develop a change management plan to bring everyone along.
Meredith Whalen is IDC’s Chief Research Officer and a member of the senior management team. She is responsible for IDC’s Research Global Product Line of subscription products for tech buyers, suppliers and investors, and the analysts responsible for delivering them. Meredith sets IDC’s annual thought leadership theme and research agenda for IDC’s global team of 1,300+ analysts.
Read More from This Article: IDC chief research officer: GenAI, from experimentation to adoption
Source: News