Artificial Intelligence (AI) cannot be trusted to handle human tasks on its own because it is not actually intelligent. What AI is great for right now is things computers do best, like processing data or handling programming languages or formulas in Excel, but its usefulness as a writing tool for communicators is still limited. In fact, if you try to use AI for complex, high quality work, you can waste a lot of time trying to tweak your prompts to get it to do what you want when it is not yet actually capable of that level of complexity at this point. In my experience with AI, it hallucinates a tremendous amount, so, as SyncPR addresses in their article “AI in PR and marketing: Use, benefits and limitations,”  if you don’t research and get backing for all of the information it has put into your communication, it can make you look foolish or even publish something dangerous or something that puts you at legal risk. There have been some famous cases where AI made up authors who didn’t exist, fabricated court cases, or gave the wrong information, causing embarrassment or financial losses for the organization who used the erroneous AI. Sometimes, the process of checking the accuracy of everything it has stated can take longer than writing the communication without using AI. Communicators will need to be trained on how to quickly vet the statements that AI makes for accuracy and to determine whether they are real. Training on how AI tends to fail can help communicators identify ways to eliminate its problems from a final composition. Common AI errors include fabricating/hallucinating sources, quotes, and information. One way to identify this in quotes AI pulls is to check the source it states to see if that quote actually exists. If it doesn’t, you are dealing with a hallucination. Another pitfall of AI use is that it uses language from its sources without attribution. Training on how to use a plagiarism checker will help protect communicators from accidental plagiarism due to AI. Training should also include examples of situations in which AI errors have resulted in negative consequences so communicators understand that the threat is real.

The reality of AI now is that its power and programming is not sufficient for it to perform complex human tasks like skilled communications at this time. The following video is an insightful look at what AI is good at and what its current limitations are, detailing why it falls short on complex communication tasks.

The other issue regarding the use of AI is that people who use it frequently get a somewhat intuitive feel for how AI sounds when it communicates. So, it stands to reason that there are going to be people receiving your AI-written communications who will notice that you used AI to write or respond to a post and feel offended that you did not take the time to write the communication yourself. So while AI is not currently best for complex writing tasks, it can be useful for simple ones. The simpler the communication, the less likely it is to give itself away, so restricting its use to short-form communications like Twitter posts or brief, repetitive internal communications is best at this time. You can also use it just to brainstorm communications you write yourself or find new ideas when you get stuck. I do feel that these problems will disappear as the quality of AI improves, but at this time, it is only useful for simple communications, and using it for more complex compositions places you at risk of AI plagiarism or of publishing false or misleading information.