Only one in three AI projects reported to succeed

Register now

Only one in three AI projects are currently succeeding, according to a new study from market researcher IDG.

Data incompatibilities and organizational frictions are at the root of the problem, say the 200 U.S. and European IT executives surveyed. They indicate that 96 percent of businesses with 1,000 or more employees face data-related issues such as silos and inconsistent data sets. At 80 percent of the businesses surveyed, tensions between data scientists and data engineers are undermining the level of collaboration needed for a successful AI deployment.

As a result of these issues, the study found that even when artificial intelligence projects are successful, they are taking more than six months to go from concept to production.

Other survey results reflect the sundry other challenges surrounding corporate AI initiatives:

  • 98 percent of those surveyed are having difficulty aggregating and preparing large data sets;
  • 96 percent of respondents experience problems with data exploration and iterative model training;
  • 90 percent cite obstacles to moving AI models into production challenge,
  • And 87 percent say their organizations invest in an average of seven different machine learning tools, adding to the complexity of their AI initiatives.

Better integration
So, what would help these IT executives overcome their AI obstacles? Ninety percent of them say that the key is to better integrate their data science and data engineering efforts.

“Data is the fuel that powers AI, [and] large amounts of reliable data that data scientists can iterate on is the key to AI success,” notes Bharath Gowda, vice president of product marketing at Databricks, the corporate sponsor of the IDG study and an analytics platform provider. “But organizational silos between data science and engineering cripples the iterative model development process.”

Making matters worse, Gowda adds, the divide between data and AI technologies increases complexity throughout the AI lifecycle. A unified analytics platform addresses this dilemma, he says, by integrating big data with AI and fostering better collaboration between data science and engineering.

When asked, nearly 80 percent of the survey respondents agreed that an integrated platform would help support their AI initiatives.

To view the complete IDG study go to:

For reprint and licensing requests for this article, click here.