Skip to content Skip to footer

Below is an excerpt of the article “When good AI development project fails” written by me, which first appeared on IDG TECH (talk), an exclusive online community brought to you by IDG (publisher of CIO, CSO, Computerworld, InfoWorld, Network World, and other technology sites).

While AI adoption is on the rise, according to 2021 Gartner reports, 85 percent of AI and machine learning projects fail. 

Why are they failing? Often it’s because of a lack of understanding of what AI can and cannot do, leading to unrealistic expectations. Most importantly, AI will never work completely autonomously without human intervention. AI can augment human insights and judgment, but humans will always be in the loop. Companies that fail to recognize this may already be set up for failure. Consider the other actions that can jeopardize successful AI outcomes.

When Good AI Development Projects Fail

Not understanding the problem you’re trying to solve 

Many companies see AI as the next shiny object that needs to be grasped, but the approach should be to start with the business problem and then determine if AI may be able to solve it. 

Keeping humans out of the equation

Not only do companies often lead with technology instead of a problem, but they also lead with technology instead of humans.  Any type of digital transformation must first understand the problem, and how it impacts humans.  

Having insufficient data

The key to successful AI is having the right data to train algorithms. Unlike other software projects, AI is only as effective as the data upon which it is trained, so sufficient data needs to be available or attained – 10,000 data points at least. 

Ignoring proper change management

AI is a major digital transformation that impacts an entire organization and should be treated as such. It requires a change in processes and people. Before letting an AI system loose in a company, it’s important to set expectations on what it can and cannot do. 

Expecting immediate results 

AI is an iterative process that in many ways has no beginning and end.  While an AI sprint can be conducted in 2-3 weeks, the real impact of AI can take months to be realized. And, as a data-driven solution, it can take ongoing training in order to reach and maintain a high level of accuracy.  

While there are many reasons why AI initiatives can be perceived as failures, those companies that lead with a human-centric focus almost always win. Understanding the real problem impacting humans and setting out to see how AI can help address it are the first and most critical steps in the AI journey – all the rest is just technology.

Sign up for our monthly newsletter:

Get the best blog stories in your inbox!