Someone had a problem with a traffic light in their town. Instead of firing off an angry email to city hall, they asked Claude to translate their everyday frustration into the kind of formal, technical language that municipal traffic engineers actually respond to.
It worked. The town reprogrammed the light.
This is a small story, but it is a useful example of something that gets lost in the noise around AI coding assistants and enterprise deployments: language models are genuinely good at register-shifting. Taking an informal complaint like "this light takes forever and nobody is even coming the other way" and turning it into a structured request with proper terminology - signal timing, traffic volume observations, cycle length - is exactly the kind of task where Claude performs well. The user knows the problem. The AI knows how to phrase it so the right people take it seriously.
Local government offices deal with vague complaints constantly and ignore most of them. A clearly written request that uses the correct technical terms and presents specific observations tends to land differently. You do not need to be a traffic engineer to write like one if you have a model that can handle the translation.
The practical takeaway here is narrow but real: if you need to communicate with a bureaucracy, a building inspector, an insurance company, or any institution that responds better to formal language than casual language, asking an AI to rewrite your message in the appropriate register is one of the highest-value, lowest-effort uses of these tools. No prompt engineering required. Just explain your problem in plain English and ask for a formal version.