96% of organisations experience issues when pursuing ML and AI projects

A recent survey conducted by Dimensional Research on behalf of Alegion found that the majority of organisations pursuing ML and AI projects have run into problems

A recent survey conducted by Dimensional Research on behalf of Alegion found that the majority of organisations pursuing ML and AI projects have run into problems

According to an IDC report, 75% of enterprises will use artificial intelligence (AI) by 2021. Nevertheless, a recent survey of 277 data scientists and other AI professionals found that 96% of organisations experience issues during ML and AI projects.

The reality of ML and AI projects

Sponsored by Alegion and conducted by Dimensional Research, the survey contained responses from large companies across almost 20 industries. Overall, the research aimed to determine the current use and development of AI and ML projects.

According to the report, AI and ML projects are no longer novel ventures in the enterprise. On the contrary, nearly half of companies surveyed reported that they had undertaken four or more active projects.

Nevertheless, 7 out of 10 respondents stated that they had only started their project during the last two years. Just half of those projects were in production, while half of those surveyed said that their AI/ML teams possessed 10 or fewer members.

More enterprises are experiencing issues

Unfortunately, 8 out of 10 companies stated that training AI/ML algorithms is more challenging than they had initially expected. In fact, almost as many respondents reported problems with projects stalling.

The vast majority of companies experienced training-related problems with data quality, labelling required to train the AI, and building model confidence. 7 out of 10 companies thus stated that they utilise external services for their AI and ML projects.

Due to the expense and rarity of AI/ML talent, the report suggests that enterprises should consider using external solution providers. As a result, organisations could benefit from better quality data labelling and model scoring.

According to the report, teams that outsource data labelling get into production faster. Out of the 71% of companies outsourcing for AI and ML projects, the 3 of the top 5 services were related to training data.

Interested in AI and ML? Check out our latest AI podcast, in which Managing Director at Brainpool AI Kasia Borowska discusses the ethical challenges involved in AI projects