Some bots do not have significant data in their repositories to resolve complex queries from customers across a wide range of scenarios. We do not trust our AI devices. Sometimes the customer wants an opinion from another human, and a bot that goes round the same loop will not improve customer satisfaction at all.
Chatbots are superb for simple sales queries, ordering products from Amazon or other online stores, and are great for customer service. But large organisations often are not ready for AI. They have complex issues to resolve and sales will only close due to relationship building, nurturing, and negotiation.
Although nine out of 10 of us are ready to take orders from robots at work, taking instructions from chatbots is still in its infancy. Although chatbots can provide background checks on employees, we want to give instructions to our voice assistants -- getting our fridge restocked and controlling our lights -- not to have a virtual manager advising us about our career.
Although AI bots can be primed with the entire set of statutes and historical cases, they would be unable to tell when someone is being economical with the truth.
They would be unable to discern nuances in speech and arguments for an against the case. They could not cross examine, nor argue. But chatbots with humanity could analyse emotion often used at hearings to come to a rapid decision.