Let us do some basic statistics here in the hyped world of AI. Let us for a moment assume that the 95% failure rate of AI projects claimed in this MIT study is correct. That means you will need to launch ~32 AI projects if you want an 80% chance to get at least one success. If you want to be almost absolutely sure to secure a successful launch (95% certain) you will need to run ~60 projects.
For a 4 in 5 chances of getting it right, you have to get 32 business cases evaluated, meet with the AI governance and CoE board 32 times, convince people to join your project (because this one is surely the winner) 32 times. That is a waste of resources, it is fatiguing the organization, and it makes you look rather incompetent.
Let us assume for a moment that you cannot change the 95% failure rate, the game now is not about getting success, it is about methodical failure. It is about getting to learnings faster than your CEO can say co-pilot, and about breaking the governance sandbox over and over again to make all of those painful learnings that will make the next project a bit more likely to succeed.
What organizations have to realize is that AI is moving fast, few if any across your organization really knows the full potential of AI, your processes and governance is not up to the task of handling this new technology, and your are learning your way forward. So learn! And learn fast!
This is where you can take a few pages from running a startup. It is not about getting it right, it is about lowering your uncertainty that you are heading in the right direction. It is about asking the most critical questions for any hypothesis and testing your most risky assumptions first. Do people even want this? Is this the right problem to address? ROI is such a lagging indicator, what are immediate KPIs that can guide us?
As a technical co-founder I have time and time again told my commercial co-founder the following: No we are not there yet, but I can now tell you 20 different ways we won’t get there. And those 20 ways I failed, might mean I only have 12 more failures to learn from before I get a success – at least with 80% certainty. That means my only job is to fail forward, fail cheap, and fail fast. This is the mindset you need when you approach the unknown – and AI is for most of us still an unknown.
So drop the powerpoints, drop the grand business plans, and get out of your office to get messy. Dare ask the questions, challenge your own assumptions, talk to those that knows more than you do. Instead of talking and calculating and playing innovation theater – go do. Doing is the first step of learning, and failure is an inevitability. Embrace it, adopt that growing mindset and develop some grit. Just like any sport – everyone starts a noob but with enough repetitions you can get really far.