AUSTIN, Texas – Nearly eight out of 10 enterprise organizations currently engaged in artificial intelligence or machine learning projects have reported stalled progress, according to a survey released today by Alegion and Dimensional Research.

Furthermore, 96% of the companies survey said they have run into problems with data quality, data labeling required to train AI, and building model conference. The new report, “Artificial Intelligence and Machine Learning Obstructed by Data Issues”, indicated that data issues are causing enterprises to quickly burn through AI project budgets and creating project hurdles.

“The single largest obstacle to implementing machine learning models into production is the volume and quality of the training data,” said Nathaniel Gates, CEO and co-founder of Alegion, which develops a training data platform for AI and ML. “This research reinforces our own experience, that data science teams new to building ROI-driven systems try to tackle training data preparation in house, and get overwhelmed.”

Gates added that many companies encounter challenges early in the process, especially around accurately and efficiently labeling and annotating enough data to train their algorithms. “We believe that as enterprise data science teams gain more experience with machine learning projects, they’ll be more likely to offload activities such as data labeling, model validation and scoring, and in doing so speed their path to model deployment.”

New concept, harder than thought

While large businesses with more than 100,000 employees are the most likely to have an AI strategy, only 50% of them currently have one, according to MIT Sloan Management Review. Alegion said in its survey, AI is still a new concept:

• 70% said their first AI/ML investment was within the last 24 months.
• More than half of enterprises said they have undertaken fewer than four AI and ML projects.
• Only half of the enterprises have released their projects into production.

Additional responses indicated even more challenges within their projects:

• 78% of AI/ML projects stall at some stage before deployment, with one-third stalling at the proof of concept stage.
• 81% said the process of training AI with data was more difficult than they expected.
• 76% attempt to label and annotate training data on their own.
• 63% try to build their own labeling and annotation automation technology.
• 71% said they ultimately outsourced their draining data and other ML project activities.

When asked the types of problems that they’ve experienced with AI training data:

• 66% said they had bias or errors in the data
• 51% said they didn’t have enough data
• 50% said data was not in a usable form
• 28% said they didn’t have the people needed to label the data

The electronic survey was conducted by Dimensional Research of 227 participants, representing five continents and 20 industries. Participants represented enterprise data scientists, other AI technologists and business stakeholders involved in active AI and ML projects.