What problem is AI or ML solving for you or your customer

I wanted to write today about AI and ML, and take a day or two off from my recurring posts on lessons learned from many years of leading corporate innovation.  It may seem strange that I'm also writing about AI and ML, or digital transformation generally, but increasingly it's clear that these two management concepts - innovation and digital transformation - are linked and will influence each other over the course of the next few years.  They share common challenges in that they both have big cultural impact, but differ in that AI, ML and other digital transformations seem more "real" and demonstrable, while innovation is still viewed as more problematic and risky.

One commonality they both share is the "shiny object" problem - that is, it's cool to be "doing" innovation or digital transformation, but to what end or what purpose?  Both digital transformation and innovation have this common element - that the mere activity seems to be validation enough, and that every team and every executive should be doing something with innovation and digital transformation, regardless of how well defined the activity is, or how clear or certain the outcomes.

One question I wish more people would stop and ask themselves about innovation, and digital transformation, is:  what opportunity or problem are we solving that is 1) important to us 2) important to customers 3) drives new value or radically reduces costs or increases efficiencies 4) has the support of management if we get it right.  And yes, that is a compound and multi-part question, but still one that everyone doing innovation and digital transformation projects should be able to answer rather succinctly.

Drives new value or radically reduces costs

You'll notice that in the multipart question there is a multi-part answer:  drives new value or radically reduces costs or increases efficiency.  I put that statement there because of the flying car phenomenon:  everyone over the age of 40 has been promised a flying car or jet backpack in their lifetime, and yet it never appears.  Yet the advance of technology has been astounding.  It's just that many new technologies are first applied to existing problems - making cars more nimble or more safe or more fuel efficient, rather than making them fly, which is a newer and riskier application.

AI and ML, and much of the digital transformation that will be accomplished as it is first adopted will have the same tenor - it will be applied to existing processes to accelerate them, reduce variations and remove humans from the process, except to manage exceptions.  Only then, once these technologies are proven, will they be applied to create radical new capabilities or insights.

So the question becomes:  what can AI or ML do right now better than existing capabilities or processes?  This is why you'll see a lot of RPA - robotic process automation - improving existing processes using robots or ML applications.  Your big goal if you are trying to get an AI or ML project off the ground is to determine what key challenges your organization has that can be improved through the use of AI or ML, and how to scope and manage the expectations.

Problems and Challenges

Your customer - internal or external - has problems and unfilled needs that can and should be easily defined and prioritized.  Once you've done that you can then determine which opportunities are best suited for AI and ML applications.  Then you'll need to understand the cost of implementing a digital transformation solution (often not that expensive since there are many open source applications) and also determine the amount of process definition, learning and data that are available to get the digital programs to work at least as efficiently as the people and processes in place.  Here's the rub - how the processes are defined now may not be optimal for AI or ML, and may need to be reconfigured, which can have knock on effects to the processes upstream and downstream from the activity you are focused on.  Plus, having enough good, clean, validated data to train the AI or ML can also be problematic.

But these implementation questions are somewhat secondary to a more important question - have you defined an important problem or need that an internal or external customer wants to have solved and is willing to pay for when you start implementing AI or ML?  This is often the same question that is asked about half way through an innovation project - "what are we really trying to solve, and for whom" - that should have been the basis for the project, rather than a discussion question half way through.