Polite AI customer service feels as rare as a warm greeting in a bustling market. I experienced that contrast first‑hand in a French market and then with Microsoft’s AI support. If we want AI to be truly helpful, it must learn empathy, not just scripted apologies.
I’m in France on holiday, wandering a local market and conversing in terrible French with stallholders. Their politeness and the warmth of our interaction stood in stark contrast to my first conversation with an AI “person” while sorting out my Microsoft Authenticator app last week. It’s all very well being polite in a small French village, but is it possible when a business is dealing with millions of people?
In the market, the stallholder and I used all our senses to understand each other beyond the words we spoke. Gestures, posture, expressions and the surrounding atmosphere were unconsciously interpreted. I could tell the stallholder was busy; she could tell I was a tourist from my hesitant French. We adjusted our behaviour and the interaction felt natural and pleasant.
My call to Microsoft began with the usual automated telephone “triage”. I was cut off several times, growing frustrated and confused about where I was in the system. Eventually I was connected to an AI “person”, whom I shall call Mr AI, with no introduction or explanation of what to expect.
Can an AI understand when I’m frustrated?
Yes, by analysing tone and language, AI can detect stress and adjust its responses. A simple introduction would have been both polite and helpful.
Perhaps it was a beta version, because the interaction was dreadful. I was cut off repeatedly, faced long delays, and had to answer the same questions several times – accompanied by many apologies. Eventually Mr AI informed me a case had been opened, then the call abruptly ended without explanation, leaving me frustrated and uncertain. Only later, after checking my email, did I discover a human would contact me. I had assumed the supposedly intelligent Mr AI would actually resolve my problem.
It is easy to program apologies into a machine, but personally I find machine apologies irritating and hollow. A machine cannot feel empathy; it cannot truly be sorry. Only people can.
How can AI be polite without sounding robotic?
Politeness works when AI introduces itself, explains next steps, and uses context‑aware language. Then again, I also find myself annoyed by “Sorry, not in service” notices on buses, as though the vehicle were a relative of Thomas the Tank Engine.
So, a note to all service designers valiantly trying to make sense of this remarkable new technology: don’t focus solely on functional usability. Consider the emotional side of interactions too. Simply sprinkling “sorry” and “please” into automated dialogue does not create empathy; from an unfeeling machine, they are empty expressions.
Surely, intelligent systems can move towards something resembling an understanding of the human condition. By gathering and analysing contextual data, they could begin to recognise what emotions a user may be experiencing and respond appropriately. If apps such as Merlin Bird ID can recognise thousands of intricate birdsongs, then machines can be trained to detect stress, confusion or frustration in a person’s voice and language. The topic being discussed provides further clues. Someone seeking medical advice requires a different tone and level of care from someone buying a season ticket. The difficulty of a conversation – and the answers being given – could indicate rising frustration and trigger a more supportive path of interaction.
Conversely, we humans are also losing some of our ability to be polite because of machines. Technology often creates barriers to normal human behaviour. How often have you seen someone on a video call casually scrolling through their phone – something they would never dare do in a face‑to‑face meeting? Or witnessed the rudeness people display from the safety of their cars, blaring horns and gesticulating angrily from behind a windscreen – behaviour rarely seen while walking down a crowded street? Even email has made discourtesy easier. Replies are simple to send, yet many messages go unanswered. Even a brief “no” feels far better than silence.
Perhaps technology could also help remind us to be more considerate, offering prompts or tools that encourage more thoughtful interactions.
Technology allows us to connect with more people than ever before, yet it can also leave us feeling more disconnected. We do not all live in a pleasant French village where everyone knows one another and has time to chat, but we all still want to feel valued and respected. Technology has dramatically expanded our interactions with the world, but in doing so it has also made many of us feel like just another cog in the machine. We need to design systems that anticipate emotions and respond more thoughtfully to the people using them.
