I've often found that by far the best way of teaching him to do this kind of thing relies on finding something he wants his computer to do for him.

This could be just about anything - if he likes sports, it could be a sports results and stats database, if he likes RC modeling it could be an interactive application for setups for his radiocontrolled cars. Your only real role here is to ensure he chooses something feasible in a reasonable timeframe (don't suggest writing Quake5 :) )

The thing I like about this approach, is that it will teach him far more than just "how to do it" - you can start it with a discussion about how he wants to go about it, to start with which language (pros and cons, quick GUI development vs. old school stuff - basically just see what ticks his boxes) and it'll then take you through the basics of data models, and the fact it'll be useful will keep him motivated. Help him break the task up into little bits, and use the first few to teach him the ropes, and then let him try some on his own.

Make it clear it's his project, that you're there to help whenever and wherever you can - but don't judge. If he wants to start with an Access DB - by all means point out the pros and cons, but it's his toy - let him do it, he should be old enough now to see for himself whether it's "right" or "wrong".

I can't figure that out either.