Like I’ve said many times before automation is a wonderful thing when properly applied and tested. But there are many examples of automation not only failing to do what it was expected to do but also actively wastes huge amounts of time and money by their very existence.

There is a plethora of cases illustrating this phenomenon but one that really stands out is the Michigan Data Automated System which was known by the acronym “MiDAS”. This is from the website entitled “When automation goes horribly wrong” and the book Humans vs. Computers by Gojko Adzic and very much a case of “computer overreach”.

The MiDAS system was supposed to review unemployment claims for payment and determine which ones were fraudulent and which weren’t. That wasn’t a problem in and of itself. What was a problem is that the computer was allowed to unilaterally bring legal collection actions against those it thought were “scamming” the system without first being reviewed by a human.

Even that wouldn’t have been a huge problem had it not been for the fact that this computer system was WRONG 92% OF THE TIME! The author notes that automating a process only makes it faster, not better. He also says that automating a bad process makes it spiral out of control without any oversight. I couldn’t agree more.

I’ve seen this general scenario repeated over and over in industry with the problems well known and documented by those who had to work “with the system”. Or more correctly “try and make the system work”. Soon after the equipment went “on line” the only unknown was how long management would take to “pull the plug” on the whole operation.

For the “bottom line” perspective The MiDAS system cost $40 million dollars in 2013, the year it was “brought online” It left in its wake a trail of damage and destruction that could easily top $400 million or ten times what it cost. And the original computer system had to be discarded as well!

The resulting fallout from this flawed system is still ongoing so this story isn’t over but so far but it was reported that this “MiDAS touch” had overwhelmed the courts with a backlog of 30,000 cases waiting to be heard by a judge. One federal class action lawsuit had been settled but another was still working its way through the courts and a third lawsuit was filed against Fast Enterprises which is the company which sold the software to the state.

This is yet another case where the proper testing procedures would have revealed these deficiencies without exposing the user to great liability. What’s wrong with testing this software on a small group of, maybe, several hundred claimants? The same problems would quickly become obvious but the issues dealing with this would be minimal.

There are no shortage of “automation going wrong” horror stories out there. We use seat belts when we drive our cars and helmets when we ride our bikes to prevent injury but getting automation “right” is even simpler. Only due diligence is required for success which is defined as proper operation with minimal problems.