No question that Amazon is a smart operator. But it's also capable of some astounding blunders, particularly when automation and big data take a hand. That seems to be the case now as a lot of Amazon customers have found their accounts locked without much in the way of explanation or efforts to rectify the situation.

For at least a week, many Amazon customers have complained on social media that their Prime accounts -- the ones that people pay for -- were either locked or closed with little to no explanation. Some are affiliates, meaning they help drive some business to the company.

In some cases, people have received some reply from Amazon. In others, nothing.

One user, who claimed to be a prolific reviewer on Amazon, tweeted that the company did eventually unlock her account, but all her reviews were gone. Another said that Amazon sent a link as help that won't work because the account would first need to be unlocked.

Amazon told a portion of the affected customers that they had broken the company's terms of service. Except, apparently Amazon didn't tell them how or what to do to avoid future trouble.

It isn't clear exactly how many people are affected. Some reports suggested hundreds while others said thousands. Whatever the number, the results have been a lot of bad word of mouth on social media along with a hashtag of #amazonclosed and terrible press. It also appears to have happened before, although without as much publicity and whether at the same scale, no way to know.

In a way, this sounds similar to the case where some high-profile websites lost access to Google AdSense when a combination of AI and claimed human review decided the sites had shown inappropriate content. Except, Google ultimately reversed the decisions, suggesting that even with human oversight (often gained through gig economy assignments), decisions can be a toss up.

Dependence on automated systems becomes a greater problem the more companies rely on them. Learning systems are only as good as the information they receive as the models of action. Software systems can be buggy. Developers might not understand the particular business function they are to recreate. In fact, it can become a demonstration of the Dunning-Kruger effect, where highly trained developers assume they know more than they do. Expertise in one area can have a crippling effect if not balanced by requisite humility.

Automation of business processes can be fine if someone keeps close watch over the result. Companies get carried away and assume that systems put into place once will continue to work correctly. Machine learning systems have to keep learning as conditions change.

It all means reworking the corporate mindset about automation. Yes, reduction of workforce will be one of the intents. To pretend otherwise is absurd. One of the standard excuses that companies give when implementing automation is that they will redirect employees to activities that offer more value. Only once have I seen someone in management admit a real aim was to reduce headcount.

Some might argue that should be the intent. Companies frequently have many needs that need tending. To pretend that they can do without help may be profitable in the short term, terrible in the long run. That isn't to say companies are never overstaffed. At times they are. Business conditions may change and require a response.

And yet, so often executives and investors are short-sighted. They want quick solutions, easy money, and no criticism. As economic and social forces struggle with changes that technology evokes, corporations cannot plead indifference. Not should they be short-sighted toward the needs of the company. Perhaps it is time for many companies to think hard about their strategies and steps they assume are wise but that may not be.