Every spring, we send families a climate survey. We draft the questions, translate them, send three reminders, and call a one-in-five response rate a good year.
Meanwhile, families fill out a different survey all year long, and nobody has to remind them. It's the questions they ask your website.
Take a single week of them. Thirty families ask when winter break starts. Nineteen ask how to report an absence. Twelve ask, in Spanish, how to register a kindergartner. Six ask about the new bell schedule, and nobody can answer them, because the memo went home in backpacks and never made it online.
None of those families meant to give you feedback. Still, each question is a small confession about your website: this was hard to find, this was never posted, this was written for someone else.
A survey tells you what families think when you ask them; their questions tell you what they needed when you weren't in the room. Which would you rather plan from?
I'd argue the questions nobody could answer are the most valuable of all. They're a to-do list your communications office didn't have to build: post the bell schedule, move the calendar link up, translate the registration page. Fix those, and next week's questions change.
This is what gets lost when we think of an AI assistant as an answer machine. Answering is the easy half. The more useful half is listening—seeing, in your families' own words and in their own languages, exactly where the website falls short.
Your families have been filling out this survey all year. Somebody ought to read the results.