AI is a wonderful tool.

It has, for example, accelerated scientific research and rapidly developed the practice of medicine. In a recent edition of Strong Message Here (BBC Radio 4, 2 April 2026, available of BBC Sounds) an audience member shared that AI had picked up an abnormality on a scan that has been missed by 2 medical professionals. AI can help to save lives. 

There are also limitations to AI. Two examples were discussed in the programme and should set alarm bells ringing:

“A glass of wine overflowing”

Chat GPT was asked to produce an image of a “glass of wine overflowing”. AI couldn’t do it and persisted in producing images of half full glasses of wine. When asked again (No, I said a full glass of red wine), AI remained relentlessly (possibly irritatingly?) optimistic: Let’s get it perfect… Let me know if this finally hits the mark. Eventually, AI produced an image of a glass of wine that was vertically half full, then diagonally half full. The reason for its failure? There are very few images available of completely full glasses of wine. AI hadn’t been given the image and therefore couldn’t retrieve it. It couldn’t imagine what it hadn’t already seen. What is more, it could not comprehend that it had no ability to answer the question.

Palantir’s bomb dropping technology

The second example is more chilling. Some of you may have already heard of Palantir which develops data programmes for use by the US and UK military. The technology ultimately directs where bombs should be dropped. One of the presenters on Strong Message Here read out the instructions for use - they sounded rather like the instructions for a toaster. The technification of process and language, wholly data driven, potentially separates the user from the human consequences of deploying the technology. Moral judgment is absent. Palantir needs a human operator to engage brain in the wisdom of its use. Without that, Palantir is a killing technology that cannot weigh human cost.

So how does this relate to the writing and giving of speeches?

At present, AI often lacks imagination, empathy and judgment. AI is like a clever friend who remembers everything s/he has ever been told and can reproduce knowledge with superhuman speed. But AI has no emotional intelligence. It struggles to think through how your audience will react to your speech or argument. It struggles to develop arguments that will really engage the particular human beings who are going to listen to what you say. It lacks subtlety and crucially, the insight to know when a particular argument or proposition might offend, irritate or bore. AI is no diplomat. Human beings are more sophisticated than that. We live with others, develop empathy and learn (hopefully) from “putting our foot in it” and experiencing the discomfort of broken relationships.

When writing speeches, use AI if you must, for some initial ideas or to get you started, but use your own wisdom (as opposed to knowledge) to draft the script. Engage brain and be a human being, not a machine. It goes without saying that you will be more authentic in your delivery if you have crafted your own material. 

We need to continue to invest in AI but we also need to invest in EI (Emotional Intelligence).  Communicating (that includes listening!) is the best way to develop EI. AI may save lives but it is can also break relationships.

Wishing you an enjoyable Bank Holiday weekend. I hope (if it is your thing) that you get the glass of wine you ask for!