By John Stroud and Jen Schellinck
Creating an AI Strategy can be intimidating. The field seems so new and technical, and it is topic on which so few people have experience, that it is hard to know where to begin.
The standard industry benchmarks can be downright scary. Some 70% of AI projects fail and some 95% of projects experience delay.
The good news is that you can learn from the experience of those who went before you. With a little bit of help, any leader can handle it, even if they lack the technical skills.
Even more importantly, the payoff to getting it right is enormous. AI will change every industry. Even the most traditional of industries, like agriculture, leverages satellite images AI to increase crop yields foot-by-foot basis. If AI can transform farming, it can transform almost anything.
The secret is to realize that you don’t have a technical problem to solve. You have a people problem to solve.
The best way to get people aligned on doing what needs to be done is by asking them the right questions.
If your team comes to an agreement on these answers, the technical matters won’t exactly just fall into place, but they will be so much simpler. You can hire technical people to solve these technical problems (just the way you hire specialized staff to solve all your other technical problems).
Here is a cheat sheet of questions to ask.
- Start with Your Corporate Strategy – Is it Still Relevant?
Start with your strategy. Does it still make sense? In particular, does it make sense in a world where (because of AI) we can have “smart” products and services?
Don’t skim over these questions with a “yeah, yeah.” Be mindful that AI is an exponential technology, whose capability has been doubling about every 3-6 months since 2011. This means that in the last decade the effective computational power of AI (taking the power of the computers + the improvement in the algorithms they use) is approximately 300,000x of what it was ten years ago. That is a lot to take in on its own. Now think about this: in 6 months the capability will be 600,000x what it was in 2001. And in a year it will be 1,200,000x of what it was in 2001. With all that extra capability, are you sure that your strategy still makes sense?
Implicit in this steps is a warning: please don’t start by picking the first or easiest use case comes to mind. It might work, but chances are that you’ll miss out on the biggest opportunities. You don’t want to try to get to the moon by building a ladder. Adding a few more feet of ladder every day might feel like progress, but you will never make it.
- What is a Problem Worth Solving? Picking it is Half the Battle
Having decided where you want to go, the next question is how will you get there? Now is when you start thinking about use cases.
Ask yourself, “What question do we want answered? What is a problem worth solving?”. This may be surprising, but getting your team aligned on this is half the battle.
Here are some tricks to finding this kind of question:
- Start with the customer in mind.
- What decision would the customer ultimately care about? What is most valuable to the customer?
- What should the end experience be?
- With all that in mind, what micro decisions are embedded in in your products, services or back office/corporate shared services. These micro decisions can be automated, which is where AI comes in
- Is the Problem in a Core Competency or a Cost Centre?
People often assume that if they want AI they need to build it themselves. That’s not the case. There are incredible number of AI-as-a-service offerings that organizations can use to get started. Picking the right one can save you an incredible amount of time, effort and money.
When does it make sense to develop a bespoke solution, and when does it make sense to use AI-as-a-service. We suggest this framework.
|Core Competency||Cost Centre|
If the problem is in your cost centres, AI can be a way to reduce costs while maintaining or improving operations. Our advice is to take that win, and move on to solving bigger problems.
But if the problem is in a core competency, the more likely it is that you will want a bespoke solution. Why? This is where your competitive advantage comes from.
- Build a Data Pipeline
With the right questions in mind, now you turn to getting the data. Think of an AI like an electric lamp. A lamp needs electricity to work. Well, the data is the electricity that powers the lamp.
The good news is that companies often have more data (potential sources of electricity) than they realize. It can be in the form of text, audio or video files.
The bad news is that it is probably not organized the way the AI needs it to be.
You need to build what’s called a data pipeline. (We know that we are mixing electrical and plumbing metaphors; our only defense is that that is a standard terminology. Maybe it is because data scientists were not also English majors).
Brace yourself: Getting properly organized can be a huge undertaking. In terms of the total number of hours on a project, perhaps as many as 80% of them will be spent here
Thankfully, when teams are pointed in the right direction, with the right tools, and with support from leadership, these are very solvable problems
- Think Like a Startup
Throughout this process, think like a start-up. Be ambitious in your vision (to transform the organization), but start with problems that you can solve. Solve them, and then iterate.
Every time you solve a problem, you will get better at the process. Your capability and confidence to take on more problems will only grow.
- Due Diligenc
To make sure the project is a success you can’t cut corners. AI has its own unique challenges around finding the right talent, change management, protecting privacy and acting ethically.
We’re parking those issues for now because there is only so much a person can digest in a single blog.
Bring your team together to start the conversation. They’ll enjoy it (who wouldn’t want to transform the organization). As the leader, you are now guide them even without knowing the answers in advance. Remember, the best leaders run their businesses on questions, not answers.
If you’d like us to help facilitate that conversation, don’t hesitate to reach out.