Not sure I get the context or how slack operates -- but wouldn't most of the big companies have functional tests already in place to catch erratic ML models in production. If they are deploying ML model for the first time for some new features, wouldn't they require to test the functionality and SLAs..