Before your school buys another AI tool, start with the job it needs to do
A useful school software conversation starts with an ordinary task: who does it, where time goes, and what would make the work easier.
By Jeff Carstens · AI and education · Published 2026-10-23
Imagine a school looking at an AI tool for family communication. The demonstration turns a few notes into a polished message in seconds. It looks useful. Then someone asks who checks the translated version, how staff correct a wrong detail, and whether the message reaches the system families already use. That’s where the evaluation starts to get interesting.
Describe the Tuesday you want to improve
‘We need an AI strategy’ is a difficult place to begin a software purchase. ‘Our office team rewrites the same event information for three different channels every week’ gives you something you can examine.
Sit with the person doing the work. What starts the task? Where do they get the information? What do they copy, check, or chase? What happens when a detail changes after the message goes out? A tool that helps with one sentence may leave the rest of that process untouched.
For this imagined communication project, the goal might be to prepare a consistent draft from approved event details, ready for a member of staff to review. That’s specific enough to try. It also gives the reviewer a defined job rather than a vague instruction to ‘check the AI.’
Give the demonstration a few awkward examples
Ask to see what happens when the input is incomplete. Try a changed date, two events with similar names, or a request the tool shouldn’t answer. Use invented information for the evaluation. Watch how much work it takes a staff member to find and fix an error.
A reasonable trial would compare the current process with the proposed one from beginning to end. Include setup, review, corrections, and moving the result into the place it needs to go. Decide beforehand what improvement would make the purchase worthwhile, then record what the trial shows.
The result may be that the AI draft is helpful. It may be that a better form, a shared template, or a clearer process solves most of the problem. Finding that out early is a useful outcome.
Ask who owns the extra work
New software arrives with jobs attached. Someone manages access. Someone learns the settings, answers colleagues’ questions, and deals with the day the integration stops working. A short demonstration can hide that ongoing effort.
Ask the vendor to walk through the details that affect your staff: how accounts are removed, how records are exported, how a mistaken action is corrected, and how support works during the hours your school needs it. Ask what the price includes and what would increase it. Get important answers in writing.
For any proposed use of student or family information, involve the people responsible for your school’s data decisions before entering real records. Establish what information would be sent, why it is needed, who can access it, and what the vendor does with it. A practice trial with invented records lets you examine the workflow while those questions are being answered.
Keep a decision small enough to explain
You should be able to explain the proposed purchase in a few sentences: this team has this recurring task; we tried this approach; here’s what improved; here’s the remaining work; and this person will own it. If the explanation depends entirely on the phrase ‘AI-powered,’ the evaluation probably needs another pass.
At SHL, the software directory is a starting point for discovery. Listings summarize vendor information and link to sources. They aren’t a substitute for trying a product against your school’s workflow, and we haven’t published hands-on editorial scores.
The next useful step could be quite small. Pick one recurring task, describe how it works today, and ask the people doing it what they would change. You’ll have a better brief for a vendor conversation—and a better way to tell whether the demonstration answers it.
Jeff runs SchoolHouse Logistics and writes about building with AI and choosing useful tools for schools.