Why AI is failing in the enterprise and two ways to correct it

Enterprises across the globe are struggling to prove positive ROI from their AI efforts. Studies over the past 12-18 months have shown only 5-25% positive ROI from AI projects. While not every project is failing, most are. So what is the magic formula to ensure that AI efforts move from the negative to the positive?

First, it is important to understand that we are still largely in the first horizon of AI opportunities. There are two other horizons coming. More on that in an upcoming post. For now, let’s focus on how to get positive ROI from the first wave opportunities.

After analyzing many enterprises and their AI efforts, I’ve discovered there are two pathways that turn most AI efforts into positive ROI outcomes.

AI is not just another technology

Before getting into the two pathways, it is important to understand that AI is not just another technology. In the past, technology has largely been ‘dropped’ into enterprises will little training or support. Most new technology causes a behavior of learning to adapt to using the new technology. Interestingly, much of this comes from its inherent ability to be understood. Think about the google.com search engine. google.com is an incredibly powerful too. What do the training instructions look like for the search engine? There aren’t any. Why? It’s so incredibly easy to use.

This was common with many technologies over the past decade or so. If there was training required, it was only because the usage patterns were new and change was relatively easy. Hit the green button in application A. Use the blue button in application B.

AI is different.

AI fundamentally changes how we think about and perform work.

For decades, changing user behaviors has been one of the biggest hurdles for technology adoption.

AI requires a different approach, and way, to engage users. That is why many enterprises are struggling with AI adoption today. Most enterprises are simply dropping the technology into existing processes and expecting widespread adoption and results.

When analyzing this further, there are fundamentally two ways that get over the hurdle and lead to success with AI efforts.

Pathway 1: Slipstream AI into existing processes

The first way is the easiest. Infuse AI in a way that users don’t even know they’re using it. The more you require users to change their behaviors on how they work, the more problematic adoption becomes. This is one of the core reasons solutions like Microsoft’s CoPilot have struggled out of the gate.

The more you can avoid changing a user’s behavior, work stream and processes, the less they’re involved. In this pathway, AI’s power sits largely behind the scenes. Essentially, users are leveraging AI without actually understanding they are using AI.

Pathway 2: Training

If you can’t embed AI behind the scenes, then training needs to be involved. Many user work streams have developed over years or decades. Changing those work streams does not happen overnight or magically with a technology solution. This is one of the fundamental challenges enterprises have faced by ‘dropping’ AI in front of users without training nor understanding of how it will impact the user’s work stream.

AI is a powerful tool. Users, however, are unsure of how exactly it can help their work streams. Work is needed to understand these processes, where AI can help and how best to leverage it. AI is not a solution that should be applied evenly and expected to produce consistent output.

Once the work stream(s) are understood and where AI can help, then the piloting and training starts. Users need help in understanding what changes are needed to their work stream and what the benefits are. You can’t just say ‘use it and it will improve processes’. The problem is that when this approach is taken, users are left asking how to use it and unsure how it will improve their processes. Hence why training is so critical to success.

In Summary

Enterprises are struggling to achieve positive outcomes with AI. There are two pathways to success that are surfacing. 

Enterprises that deploy AI either behind the scenes without impacting user-impacted processes or provide training to users find that their adoption rates and success with outcomes are markedly better than the alternative.

As we move to the second and third horizons of AI innovation, new pathways will develop.