In recent years, enterprise artificial intelligence has grown significantly. The research firm IDC estimates annual business investment in AI came in at $341.8 billion in 2021 and expects that figure to hit $500 billion by 2024. Comparatively, according to the Enterprise AI Trends to Watch in 2021 study from market research firm CB Insights, AI companies raised a record $33B in equity funding in 2020.
But are enterprises getting what they are paying for? A survey released today suggests maybe not, as most enterprise respondents cited increased innovation as their primary goal with AI, yet that’s not the most likely outcome.
To get a sense of where enterprises stand regarding their AI deployments, AI provider SambaNova surveyed 600 directors (or higher level) within enterprise AI, data, research, customer experience, and cloud infrastructure about their AI experience. The survey focuses on the financial services, healthcare, life sciences, manufacturing and auto, retail and e-commerce, public sector and oil and gas industries. The survey found while progress in AI is being made, steep challenges remain.
Investments in AI on the Rise
The survey found that most respondent organizations (70%) plan to invest more than $100 million of their technology budget toward technology goals they consider strategic, such as AI. While the report didn’t provide the overall revenue size of respondents or the size of their technology budget, 32% said that more than 20% of their IT budget is allocated to AI. Additionally, two-thirds claimed that their organization plans to increase their AI investments throughout the next five years significantly.
The top three reasons respondents are investing in AI are powering innovation (40%), improving operational efficiency (25%), and keeping up with the competition (22%).
Interestingly, how enterprises measure AI success isn’t aligned tightly with those objectives. Consider that 72% of respondents cited cost savings as their top key performance indicator, followed by revenue growth (67%), time savings — often a proxy KPI for cost — at 60%, new product development at 56%, and time to insight at 52%.
While enterprises want to rely on AI to help them cut costs and drive innovation, they have several complex challenges getting there.
AI Challenges Remain High
At 50%, the biggest challenge enterprises cited in this survey is the difficulty respondents said they have in customizing their AI models. Current AI technologies require constant training, tending and tweaking to become useful and accurate, and then to stay that way over time. This challenge is also likely related to the AI skills shortage. Finding the AI talent necessary to build and maintain these models was cited as a challenge by 28% of respondents.
Other top challenges cited by survey respondents include the complexity of working within the constraints of current computing architectures (35%), insufficient compute capacity for existing data (28%) and lack of buy-in and trust from enterprise leadership.
These findings track with a survey of 5,154 respondents conducted by O’Reilly Media and published in April 2021 that found a lack of skilled people and difficulty in hiring to be the biggest problem for 19%, closely followed by data quality at 18%. According to O’Reilly, the most prominent skill shortages exist for machine learning modelers and data scientists, 52%, as well as understanding business use cases at 49% and data engineering at 42%.
Additional Executive Leadership Education Required
One of the essential actions enterprises can undertake for future success, according to the SambaNova report authors, is educating business leaders on the nature and potential of AI. “Successful AI/ML initiatives are driven by the desire to achieve defined business outcomes, which requires a clear understanding of specific use cases,” the report authors said.
“Relatively few business leaders understand what deep learning is or grasp its transformative potential for business, and this lack of awareness can prevent deep learning initiatives from receiving the support they need. With improved education on the business side, organizations can unleash the potential of AI/ML to unlock innovation, achieve scale and drive business outcomes across the enterprise,” the report’s authors concluded.
Of course, fixing executive leadership understandings of AI won’t solve everything. The very real compute capacity limitations pose a challenge and will likely increasingly do so if not solved. About half of respondents, 53%, strongly agree that they’ll run out of computing power in the next decade without new architectures. That prediction may be an understatement considering that 65% of respondents said they already struggled with limiting server space.