Despite all the work I'd put into this system to make it simple, to make it about a hard personal choice that players needed to make with every action, and to make the mechanics about the character, in the end the person requesting the game kept referencing D&D. I'm not sure if that's because we haven't played many other fantasy games (come to think of it, I can't think of any others I've played with him, and my own experience is limited to some very short sessions of Warhammer, despite owning several others), but that left me with one major gap in the current system: spellcasting.
One thing that we agreed on, though, was that neither of us liked the Swiss army knife way magic worked in D&D. There was a spell for everything, and with enough levels or scrolls, you could have a massive battery of utility spells to render you immune to every hardship you could imagine. Add to that the massive killhammer high level magic became, and fighter characters become good for show only. Neither of us wanted that.
So I set about creating a system of magic that offered some decent choices without completely invalidating every other character option in the game. There are plenty of ways to limit magic in an RPG. There's drain, limited casting capacity (also known as Vancian casting), components, even insanity risks. In the end, I went with making magic slow. Yes, you can throw a fireball into the midst of an enemy horde and roast all of them, but you won't be doing it every exchange. It takes time to build up that much power.
Instead of writing a separate tome chocked full of spells (I'm too lazy for that kind of work), I broke a spell into several categories. This includes aspects such as range, area of effect and duration. Each of these has a few ratings, and the more you want to do, the higher the cost.
What cost? Effort cost. To cast a spell, you decide what you want it to do, assign it a rating in each of the categories, and then tally up the cost of all the aspects. That gives you a number, and that's the effort you need to generate before you can cast the spell. Now, that's just to gather the energy to cast it. Effort spent gathering the power doesn't actually go toward casting the spell.
Once you meet a spell's effort cost, you're free to cast it just like you would perform any other action. You generate an effort pool and look to beat the challenge rating of the task. If you succeed, your spell is a success. If not, you still cast the spell, but it doesn't have the desired effect. An attack spell misses. A telekinetic effect lacks the strength to perform the action. Etc.
However, win or lose, there's no further side effects. No drain, no loss of energy. If you want to cast more magic, you can do so, but you begin by gathering up the power again. On average, if a mage wants to throw an effect that's on par with a warrior's sword swing, he can do so every exchange. If he wants to roast an entire battalion, he can do that too, but he'll be gathering energy for a good number of exchanges. He can do other things during this time if he likes, but that requires splitting his pool, and thus gathering energy at an even slower rate.
And that's my attempt at introducing magic, keeping it powerful, keeping it fun, and keeping it in line with everything else. It looks good on paper, but only testing will tell how it actually works.