How to use frustration to pick AI use cases

I’ve spoken with many leaders of maritime companies that are aware of their urgent need to move forward with AI. A recurring challenge at the early stages is how to find good use cases that:

1. Create short term ROI
2. Build internal AI capability & maturity
3. Don’t involve too much risk

These are all meaningful goals and while it can be tricky to satisfy them all perfectly, here is one practical approach that you may find helpful. 

Together with your COO and the managers of each key activity in your business – e.g. customer support, manufacturing, marketing – follow this procedure:

1. Map out the key business processes making up the activity
2. Give each process a score from 1 (perfect) to 5 (awful) for each of these aspects:
   1. The time it takes to complete one iteration of the process
   2. The cost of each iteration
   3. Typical accuracy 
   4. How personalized the outcome is to the consumer of the process
   5. Consistency
3. Sum up the scores of all the aspects to give each process a frustration score

The top scoring processes are strong candidates for improvement through AI. 

This exercise is about turning the biggest frustrations that hamper your operations and into drivers for generating AI success stories. 

The scoring is relative, of course, and should be ‘calibrated’ to what makes sense in your particular business. On cost, for example, you would rate a process as ‘awful’ if the ability to bring the cost close to zero would make a serious difference, and as ‘wonderful’ if the cost is already so low that further reductions would mean very little. 

An example of poor consistency, which is a problem in manual processes, could be that the exact same production item is rated as ‘fine’ by one human rater, and ‘poor’ by another. Potentially, AI could completely eliminate such inconsistency – inconsistency that can lead to major internal and customer frustrations.

Once you have these scores, you have a good foundation for moving forward. The next steps include thinking through how AI could be implemented for each process, for which see my writing on the AI Value Catalyst.

If you have any questions or comments about the above, I’m here to help.