Talking to the Machine
I once wrote that AI could be either a great equaliser or a great divider. A year on, that framing feels incomplete. It is not whether AI helps or harms. It is who can make it help them and that turns out to depend on the oldest advantage there is: being able to say what you mean.
AI is becoming the way we reach everything: jobs, services, information. But what you get out of it depends on how you ask. Getting a good answer is a kind of writing: knowing what you want, saying it precisely, noticing when the reply has missed. Some people have always been better at that than others. The person who can frame a clear request gets a useful answer. The person who cannot gets a vague one and often cannot tell the difference. That is a new literacy, and like every literacy before it, it is not evenly distributed. Those never taught how to make themselves understood are the ones it serves worst.
It shows up in small, invisible ways. Job applications now pass through AI screeners before any human sees them. By 2025, 48% of hiring managers were using AI to filter applicants. Someone who knows this writes their resume to be read by a machine and gets through. Someone with the same experience who does not know this is filtered out before the first conversation. The same goes for the chatbots now standing between people and their bank, their clinic, their government. The person who can phrase the question gets helped. The person who cannot gets stuck in a loop or gives up. Neither person was less capable. One of them just knew how the system listens.
It could go the other way. The same tools can translate, simplify and draft for people who struggle to do those things alone, which is the most hopeful version of this technology. UNESCO frames AI literacy as a new digital divide and has called for global action to close it. Whether it equalises or divides depends less on the machine than on who we bother to teach. Access to the tool is the easy part. Teaching people to use it well is the part we usually skip.
Whether AI is good or bad for fairness is not the question I find most useful. It is who gets to be understood. We have built a technology that listens best to the people who were already the most fluent. Unless we are careful, it will not close the communication gap. It will just automate it.
