💡 Busy Executive Summary: Choose business problems that are data-rich, operationally relevant, and recurring. But weigh the value to the business and risk when starting with AI.
Words no one wants to hear. When starting with value-driven AI, choosing the right initiative - where your company can quickly prove value without expending a ton of resources - is vital.
I will focus on starting AI with your company's proprietary data vs. Generative AI (i.e. ChatGPT). Why? Because a company’s data is one of its most important assets and competitive advantages (and people!), putting that data to work with AI is critical for future success.
Also, the internet is clogged with articles on using Chat GPT (guilty!), and businesses need practical AI solutions today.
"Companies that layer machine learning and generative AI over their own proprietary data sets stand to gain actionable insights that will help them jump ahead of competitors in the next decade." (Source: Crunchbase News)
AI and Machine Learning (ML) won’t solve every business problem, but it can unlock tremendous value when applied correctly.
Typical problems that AI and ML can help a business with can be categorized as
If you find a specific business problem or workflow, use case where AI might be of value. There is a quick test to apply to see if this is the type of problem where AI/ML can be a useful tool.
Try this simple test to see if your business problem is a good candidate for AI/ ML:
➡️ Data Rich - Do you have a problem or workflow that does (or could) capture data?
ML needs data to find patterns and predict outcomes. Fun fact: You may be surprised at how little data you need to get started.
➡️ Operationally Relevant - Does the business care if you solve this problem? It doesn’t mean starting with the most critical workflow in your organization, but make sure that your starting point rises to the “is the juice worth the squeeze” test.
➡️ Recurring - AI automation is most valuable if a problem occurs regularly.
If you have a problem that rarely occurs or is a one-off, this is a better use of “human analysis” than ML automation.
If your workflow or problem qualifies from the above, then there is one more sanity check you would want to apply before proposing it as a starter project for your organization. Weighing the risks of the project with the potential benefits, aka. The Goldilocks test.
Although your organization might be tempted to swing for the fences at its first time at bat with AI, this would be a mistake. Your starter AI test case will quickly become an 18-24 month slog with too much risk and too little chance of success. The best starter use cases are those that are "mid-size" and, at best, "mid-risk."
“Companies think they need a “Big Win” to get value out of AI. These projects require big budgets, long runways… and often fail” (Source: IIOT).
Conversely, there is such thing as a problem that is too small. If you choose a business problem that is essentially irrelevant, it usually means there is little ROI in solving it with AI… or any other tool. Your efforts will be met with the collective shrug of the organization saying, "Okay, so that's all AI can do?"
No matter what you choose, ensure that you can attach some very measurable and meaningful metrics to your AI initiative. This was a great point made recently by Frank Casale in a panel to the effect of "Make sure the organization even agrees to what that metric is… 'productivity', that is a meaningless term because no one can agree what that means." Without measurable metrics, it will be nearly impossible to prove ROI to the overall business.
In 15 minutes see how SAVVI Al can benefit your business.