We understand the appeal of trying to shave cost and automate as much insight as you possibly can. We also understand that true data science talent is hard to find and expensive to keep. But most importantly, we suspect that your company never truly mastered the data quality part of the data management equation. Data management is not sexy… and not a task for a highly-paid analytical resource. But… if you don’t have the data management fundamentals in place to manage and govern your data, you’re simply creating a spaghetti monster, not AI.
Artificial “Intelligence” may not be so smart after all.
Let me give you an example of a simple automated process that was designed to save customers and employees time and energy… and shave some cost from the bottom line. It was most likely based on a narrow AI application known as a Pharmacy Management System and incorporates Natural Language Processing…and it’s failing.
I use a locally based chain pharmacy to fill my prescriptions. I happen to live with a 5-minute walk of my local pharmacy but have chosen to move my business to a competitor that is a 10-minute drive away. I also sold my stock in said company after repeated failures. Here’s why: I refill my prescription monthly via an automated system. I allow 1 day for the pharmacy to refill my prescription thinking that this should be plenty of time to take a bottle of the shelf, count out 30 tablets, print a label and put it in a bag. Each month, I get a series of automated messages reminding me that my prescription is ready for pick up. I faithfully show up, only to find that not only is my prescription not ready they don’t even have it in stock.
This is a repeated waste of my time and energy. I wait for 15 minutes in line, 5 minutes for the pharmacist to call in the script, 5 minutes discussing said missing prescription with the pharmacist. This is typically done during rush hours because the company, in an effort to shave costs has reduced their pharmacy hours to peak working hours (9-5) … and the line is always 10 people deep. The pharmacists explain to me that they don’t have it filled despite the repeated texts and calls from the automated prompter assuring me it is now time to pick up my prescription.
We have the same conversation each month. Oh, sorry we need to call your doctor for approval. No, you don’t - it says X refills before date X. Oh, well then maybe we need to wait on our shipment of meds. I called this in 2 days ago so I don’t think that’s the problem. Oh well then, maybe it’s on back order. The excuses and confusion get more elaborate with each passing each month.
Nobody could explain why I was getting calls and texts from the automated system telling me my prescription has been ready for 2 days.
This could be a result of any number of design or automation flaws… perhaps Pharmacy X didn’t master the concept of the feedback loop. They may not capture the failures at all. Perhaps they decided that the majority of the customers pick up prescriptions during the workday so they condensed their hours and inconvenienced their customers. Perhaps Company X uses tools to manage their prescription inventory and supply chain and it automates the delivery of meds based on refill dates that don’t correspond with the physicians’ refill instructions. Somehow, despite having all the relevant data, Pharmacy X is unable to create a positive customer experience. Insights coming from AI have not translated to operational efficiency. Whatever the case, they seem to have forgotten who their customer is (surprising at a time of unprecedented availability of data and tools) and certainly haven’t been meeting any level of service expectation.
Ultimately, automating a bad process is going to cost you customers, frustrate your employees and damage your bottom line. So, before you invest in AI, make sure that your data is accurate, your information flow matches your company’s capabilities, and your first focus is the customer experience. Otherwise, you will end up using AI to help you understand why you’re losing ground against the competition.
If you’re considering machine learning, NLP and AI, please take a strategic and measured approach to assessing whether it makes sense for your firm and answer the following questions:
Does our organizational structure reflect a strong emphasis on data quality? Whether decisions are based on traditional measures or gleaned from AI, garbage in still yields garbage out.