The majority of AI projects fail, right?

By Mark Pfaendler, La Mobilère

The transition towards data-driven business models has become highly relevant to most companies and industries. However, according to a Gartner publication back in 2018, managers find that 85% of AI use cases have not lived up to expectations or worse are deemed unsuccessful. Given this number being rather high, we were eager to validate this finding within our Expert Group “Data Driven Business Models” – held during the last meetup of this year – the Expert Day 2022. To bring our experts together, we organized a well visited workshop with 25 experts & practitioners from both industry & academia.

Our main undertaking was to first discuss the high failure rate reported by Gartner, followed by drivers behind successful AI projects. For the latter, we developed an AI project assessment framework covering 5 dimensions (Strategy, Talent & Leadership, Ways of Working, Data & Governance, Technology & Tooling). Each dimension comprises a set of best practices serving as prerequisites for running AI projects successfully, which D ONE has collected over the years from working with numerous clients from different industries. In this blog post, we would like to summarize the most important talking points & key takeaways.

First things first, our experts were skeptical about the AI project failure rate of 85%. In their opinion, a failure very much depends on how it is defined & the perspective the responsible team together with stakeholders take on failure in general. This is something interesting to better understand for which we will collect responses from a wider audience within our network.

The main goal of the workshop however was to discuss the AI project assessment framework. As an exercise, we asked the participants to apply the framework to their project experience & prioritize the set of best practices per dimension. The following table summarizes the results collected.

To wrap up, our 3 key takeaways from this workshop are:

1. An AI project failure rate of 85% has been heavily challenged – this needs further investigation.
2. According to our respondents, managing stakeholders & their expectations seems to be the strongest driver of AI project success.
3. The project assessment framework was well received which allows us to go more in detail next time & start conducting a survey to collect solid results worth sharing with the public.

The workshop was the perfect opportunity to sound & refine the presented AI project assessment framework with industry experts who deal with success stories & challenges day by day. With this, we would like to thank all participants for the active discussion & contribution to make this workshop a success. We will keep you posted on the next steps taken!