I recently had to write a simple web app to search through a database, but full-text searching wasn't quite cutting it. The underlying data was too inconsistent and the kind of things people would ask for would mean searching across five or six columns.

Just the job for an AI agent!

So what I did is this - I wrote the app in Django, because it's what I'm familiar with.

Then in the view for the search page, I picked apart the search terms. If they start with "01" it's an old phone number so look in that column, if they start with "03" it's a new phone number so look in that column, if they start with "07" it's a mobile, if it's a letter followed by two digits it's a site code, if it's numeric but doesn't have a 0 at the start it's an internal number, and if it doesn't match anything then see if it exists as a substring in the description column.

There we go. Very fast and natural searching that Does What You Mean (mostly).

No Artificial Intelligence.

All done with Organic Home-grown Brute Force and Ignorance.

Because that's sometimes just what you need.