O’Reilly, a source for insight-driven learning on technology and business, today announced the results of its 2019 Artificial Intelligence (AI) survey, “AI Adoption in the Enterprise” The report explores how enterprise organizations are planning and prioritizing AI implementations and how adoption patterns may change over the course of the year.

The findings suggest that there continues to be strong interest in democratizing AI in production regarding accountability, compliance and ethics. Despite this, a majority of businesses are not equipped with the data or AI/machine learning expertise they need for successful implementations. In fact, the need for AI talent has become even more critical, with 23% of respondents indicating lack of skilled people as a factor that slows adoption, compared to 20% of respondents last year who indicated a lack of skilled people as a bottleneck.

Other notable findings include:

  • Eighty-one percent of respondents work for organizations that already use AI.
  • More than 60% work for organizations planning to spend at least 5% of their IT budget over the next 12 months on AI.
  • One-fifth (19%) work for organizations planning to spend a significant portion—at least 20% —of their IT budget on AI.
  • The level of spending depends on the maturity of an organization. Those with a mature practice plan to spend on AI at a much higher rate than less-mature companies.
  • “Lack of data” and “lack of skilled people” remain key factors that slow down AI, along with “company culture” (23%) and “difficulties identifying use cases” (17%).
  • More than half of all respondents signaled that their organizations were in need of machine learning experts and data scientists.
  • Half of all respondents belonged to organizations that used AI for research and development (R&D) projects, while one-third used it for customer service or IT.
  • More than half (53%) of all respondents already using deep learning use it for computer vision applications, but many more use it for “enterprise data,” including unstructured data (86%) and text (69%).
  • In O’Reilly’s 2018 survey focused on deep learning, the top three tools used were TensorFlow (61%), Keras (25%) and PyTorch (20%). This year, usage rates for Keras (34%) and PyTorch (2%) grew.

“AI maturity and usage in the enterprise has grown exponentially over the past year and there are no signs of that slowing down,” said Ben Lorica, O’Reilly chief data scientist and AI Conference chair. “Mature organizations with plans to spend on AI tools that hire talent to identify use cases that fit those AI solutions will succeed, but for those less-mature companies with a lack of investment in AI, we expect the gap between leaders and laggards will only widen.”