One of the big problems is companies with "too many" customers not having enough humans to deal with them interactively. It's probably possible, but it always s

One of the big problems is companies with "too many" customers not having enough humans to deal with them interactively. It's probably possible, but it always seems like a side-job or afterthought, often shipped to low-wage countries, and hidden behind a wall of 'support documents' that hardly cover the many cases you might need to get in touch with a vendor. Google being Google (and this probably goes for many tech leaders) probably thought that using an algorithm and some documentation would be enough, even if it means about 1% of customers slips through the cracks and loses all of their Google-stuff. The problem is that in the real world this generates enough backlash that people don't want or trust your services anymore. While a cloud and even a more classical MSP is supposed to take what they are good at and make it available at a reasonable price (so you can focus on what you are good at instead of everything you might ever need to know to function) at this point you need a second service to backup your first service just to be sure you don't lose everything over some stupid automation you cannot influence. It's like an older HN post (week or 2 ago) about someone being transferred between systems during a company merger and the system deciding that he is fired. No human could influence it, everything was revoked and removed automatically and security just acted on what the computer was saying, and escorted him out. They had to wait until the automated process was done before he could re-enroll in the system and continue to work. That, and GC (and to an extent others like PayPal, Amazon, Apple, MS) are examples of trying to put nuances in human interactions into algorithms and failing, have bad real-world effects.