7A – Testing the Hypothesis, Part 1

Problem: My job (pharmaceutical data entry for transcribing prescriptions) is too tedious for humans to do, and there's a lot of room for error, and it'd drive most people nuts to sit 8 hours a day, 5 days a week (or 10 x 4 in my case) for months on end. There is probably an easier way to do this to shorten prescription processing time, error possibilities, and tediousness. It's possible this can be improved by automation, but this presents issues with the employed and may not help much. There is still plenty of room for improvement especially if I think my own job is unnecessarily tedious and repetitive. Questions: 1) Who do you think has this problem, aside from the data entry technicians, pharmacists, and doctors? Who could benefit from improving prescription processing? 2) What do you think the issues are with this? Is this really an issue? Given that about 500 people would lose entry level jobs over it if the system is greatly improved, is a loss of 500 data entry technician jobs worth an increased speed in prescription processing and accuracy? 3) Why do you think the people in "who" have this problem, and how does it differ from the others with the problem?

Interview 1

1) Processing speed increase could also shorten time for patients to get the medicines they need as automatic systems could get the work done faster.
2) There is no real problem with automation. The loss of jobs would effect the local economies of the stores but overall the net gain the company would save from automation outweighs the cost of keeping the 500 workers.
3) Patients can get their prescriptions filled faster and more accurately, and the pharmacy can save money and time from the 500 employees.

Interview 2 [coworker]

1) Automation of prescription processing could probably help people in the AI industry by training computers to understand doctor's horrible handwriting and the variation between prescriptions. The detection technology could be deployed elsewhere too.
2) The data entry system never had a way for the technicians to circle stuff that related to the related entry field. This isn't necessary for the technicians to do for obvious reasons. But it would be necessary so that the computer knows what parts of the pre-loaded correct script are associated with what parts of the image.
3) AI people would be using the overall technology for handwriting detection and sig code translation for something totally different from pharmacy I guess.

Interview 3

1) Maybe patients could see a speed benefit but it doesn't seem valuable.
2) I don't think this is a problem because it isn't worth putting all that effort in to save money from just 500 people. The research costs to develop any technology may be too high to be worth just paying the 500 employees their wages. There's already AI teams working on the technology elsewhere. Patients might see a slight increase in speed on their end but it's just not worth it.
3) Maybe it can be seen as a problem for any employee who finds their job numbing. In that case they're just not working the right job.

Interview 4 [coworker 2]

1) I think the technology may be very useful to the employees themselves without having to terminate the position entirely. They could have more fields auto-populate for them. I find it especially hard to read some stuff for the handwritten prescriptions, and some of the calculations are hard because I don't always remember every drug that is an exception or stuff like that.
2) I think we'd lose some positions as the AHT [note: average handle time] would decrease dramatically as we would just error check instead of data enter except for a few missing fields maybe.
3) i mean, I work the position. Obviously I'm aware of this issue. I'm sure pharmacists have different uses of it.

Interview 5

1) With literally millions of prescriptions filled out and databased somewhere, and all it takes is a research team to re-process the prescriptions of interest (i.e. mostly handwritten not those digital ones) so that the related items are highlighted per field. This can be fed into a neural network of some sort to train the algorithm to detect and have it start to detect others on its own through supervised learning. Much of the technology already exists to image detect, but the hardest part is obtaining metadata from millions of images. But you guys already fill that stuff out and I'm sure it's all on record somewhere. That can be used to create some pretty powerful algorithms benefiting both machine learning researchers and Walgreens (or whatever other pharmacist that can use that centralization thing).
2) The hardest part would be figuring out which prescriptions are the funkiest, that need to be used for training. I think if they worked a bit harder they may be able to use an unsupervised learning algorithm that takes all the prescriptions of one type (like from one doctor, or for one medicine) and learns from that on its own. That technology also exists but it is far more difficult to train. But with such a large training set it may work.
3) I think the problem would be hoisted onto whomever Walgreens would hire to process the prescriptions. That's a very difficult job to do, creating a neural network from either unsupervised or supervised learning algorithms. But since they could benefit from a MASSIVE training set of correct data and associated images, both Walgreens and any machine learning team could rejoice over a huge data set.