AI….AI…everywhere but not a single practical solution to use
The tagline above, rhymes with “Water water everywhere but not a single drop to drink”….since it kind of captures the same situation. The “Water” metaphor was for a scenario where someone stranded at sea is extremely thirsty, sees water everywhere but can’t drink it. With all the AI hype we are experiencing these days, I feel we are in the same scenario. Every product has suddenly “transformed” into an “AI enabled” product but we still don’t see many examples of true AI capability- specifically in the area of Supply Chains. Products that can help us tackle conventional challenges that have plagued Supply Chains for decades.
So what is true AI ?
What will be considered a true Artificial Intelligence enabled Supply Chain solution ? And this question has become more important than ever post pandemic. As the importance of technology comes into limelight, Snake Oil AI vendors are cropping everywhere.
Let me quote an example- an article in WSJ mentioned that a Smart AI solution was able to detect the change in demand/supply as early in February. Was it possible ? Very much- by analyzing the supply data, you could see supply dwindling but you can do that in Excel, look at the numbers. Did any AI anywhere predict that hand sanitizers will become one of the most elusive items in American retail history ?
And that is why I insist, if you really want to harness AI in your Supply Chain, first, understand what true AI capabilities are. Second, understand what can be a realistic first step. So here is a high level definition that I use for a true AI solution:
- React AND
How to evaluate a solution ?
What can be a good example ? Say -A Smart algorithm tracking shipments. Let us evaluate a solution like that with the criteria defined above.
On the Sense aspect- it tracks the location and route status of, say a container, from IoT data
Intepret -It then uses the data senses to interpret where the container currently is, compares the “actual” with “predicted location” etc.
Analyze: It then analyzes if the container will reach the destination on time.
React: It either triggers a status report, a warning, or a rerouting suggestion- based on how you design the algorithm.
Learn: Over time, it learns that if a container is showing a specific pattern of being at a certain location at a certain time during the transit, its probability of getting delayed will be say 80%. Eventually, the solution, when trained over a very substantial data size, can predict delays very early in the process, allowing managers to plan accordingly.
Even though this is a simple example, I will still put an algorithm like this in my AI category bucket. Remember, AI label is used very vaguely so depending on who is selling what, anything can be labeled as AI. As a Supply Chain leader, YOU have to evaluate if what is being labeled as an AI functionality, is something that can or is already being done by some existing solution – or will it really allow you to develop a new capability, a true AI capability- that meets the criteria mentioned before.
Views strictly my own.