Despite the fact that it is not so popular in the workplace and it is still more a buzzword, there are companies implementing AI (Artificial Intelligence).
However, often the output of standalone AI tools is not directly actionable, because it is disconnected from other business process systems and may be taken out of context as a result.
How to make sure that AI is embedded in the core business services and the true potential is unlocked, rather than using technologies in 'patches'?
My experience of AI is limited to applications related to operational decision support.
I do think that at this stage step changes will only be achieved by big-budget R&D (such as IBM Watson).
I think that at a small company or departmental level the old rules still apply. If you want to use any form of automation to improve a process, first make sure that you fully understand the process and have designed a manual process that reliably produces the desired outcome. Then look for a means of automation.
In relation to decision support, the biggest limiting factor in applying AI is lack of consistent and reliable data. I would advocate looking at all data that is, or could be, captured and ensure that this is being done in a systematic and cost effective way. Even if you cannot make practical use of some data today, it may prove invaluable in the future.