You'll run into a problem: evocation is already a very powerful ability.  Letting it do 'other stuff' will quickly move it out of the top tier of abilities and into the best of all abilities category.

Sample things you generally shouldn't let evocation do without any extra refresh spent:
Duplicate other powers (examples: flight, supernatural sense, etc)
Do something that a skill roll would normally be needed for (examples: picking locks | note that there is already the rules explicit veil ability, which is problematic enough)

If one of the players wants their character to do something frequently, they should pay refresh to put it on their character sheet.  If they just want to do it rarely, charge them fate points for 'temporary access to powers'

Wizards do not seem all that better to me.

Here's a sample combat wizard
Submerged (10 refresh, 35 skill points)

5: Conviction, Lore
4: Discipline,
3: Endurance,
2: ,
1: , , , ,

-1 the sight
-3 thaumaturgy (+1 crafting strength)
-3 evocation (spirit, , ) (+1 spirit power)
-1 refinement (+2 Spirit Control)
-1 refinement (2 item slots)

6 item slots:
enchantment focus (+2 crafting power, 2 focus slots)
offensive focus (+2 offensive spirit control, 2 focus slots)
defensive item (power 8, 3 uses, 2 enchantment slots)
defensive item (power 8, 3 uses, 2 enchantment slots)

So, this attacks with control 8, power 6 evocations, and defends with 2 power 8 layered defensive items.

Now, imagine that a ghoul attacks this character from ambush.

The ghoul tags an aspect, rolls +2 on the fudge dice, for a total of 8 to attack, and has a base damage of 4.

The wizard activates his first defensive item, for a power 8 defense.  The ghouls attack matches it, so the attack gets through, and the block provided by the defensive item it shattered (important if there are more attackers).  The wizard is looking at 4 damage, so he activates his second defensive item to provide 4 armor, reducing the damage he takes to 0.

Then, the next round happens.  The ghoul goes first, and the wizard probably has to use a defensive item to bring up another block, but it's highly unlikely that the ghoul will roll a +4 on the dice and break through.

Then the wizard lets fly with an evocation.  He has a rote that's a power 8 attack, so he does that an pays 3 stress.  Assuming everyone rolls a net of 0 on the dice: the ghoul defends at 5, so it's looking at an 11 stress hit, so it would have to take 7 points of consequences to stay in the fight.  But random unnamed ghouls aren't likely to do that, so the fight is over.

The wizard is down 3 item charges and 3 mental stress.  If there had been two ghouls, the wizard would probably be down 6 charges ad have his 3rd and 4th mental stress boxes marked off by the time the fight is over.  Or if the wizard had room to move, he has a chance of ending the fight in first non-ambush round with a supplemental move and a zone wide attack.

Thing I can think of:

A stunt or minor power related to resisting mind control and mental attacks.
Most sorts of mind control are for human targets, and your mind now has awesome improved bits stuck on where all those human weaknesses used to be.

A purely mental beast change.
Always wanted to do better in negotiation, but having trouble with your hunger for blood getting out of hand? Now you can swap in some new mental software and only want to kill your foes and drink their blood after the talky bits are done!

Or just some stunts or powers related to you being a really good method actor.  Or, at least, you've eaten the brains of some pretty good method actors, and you've been itching to try out your new acting skills.

A stunt or minor power to improve the sight/make resisting it easier.
"I used to have trouble looking at things like that, but now that I see the same sort of thing in the mirror, I find it's not so much of a problem anymore."

Right but the concession is a group decision so it stands to reason that whatever makes sense in the current scene is how the latter part will play out. Even if the guy kills himself he could have a pack of matches from the hotel he is staying at, or an address scrawled on a piece of paper (both these things being declarations in most cases) just because you failed to have him give you information directly does not mean you have failed to find anything out

No, I'm saying that one way of doing 'success' or 'failure' is to make the fiction layer say whatever you want 'after' the dice determine that result.  And as long as you get what you were after, 'success' can be narrated in whatever way you find interesting.

So, as long as the dice give me victory, the fiction layer can give me a victory in whatever narrative that the players find acceptable, and if the dice give me failure, then my failure also happens in a whatever narrative way people find interesting.

Example:
GM, this fight will end up with you captured by the villain.  The stakes are how much you find out about the villains plan.

So:
1: the characters aren't risking death, even if the in fiction layer of the game says that the villain wants to kill them
2: even a victory by the players here will result in their capture.
3. what the PCs get for winning is finding out about the villains plans, but how their winning actually happens in the fiction layers can be described in whatever way people find entertaining.

This allows things like the fictional layer describing narrow margins of escapes and skin of the teeth victories without needing the dice system to promote the 'you die now!" mathematics that that tends to require.

-------------------

Example, in the game, any contest you get into with a black court vampire is going to be a big deal.  They can deal large amounts of damage with a lucky hit, so you may need to have a lot of fate points in reserve to deal with them.  So, if we'd expect the characters to want to avoid fighting them, wherever possible, and prefer to do things like burn down the building they are in and stake anything that manages to crawl into the sunlight.

Bur we, the players, might prefer our characters to get into fights with black court vampires, to show how awesome we are.  But that's a pretty (eventually) lethal thing to do if we the fiction layer drives the game.

------------------

So, how can all this be applied within the scope of a DFRPG game?

You can set the stakes of the contest to be something other than the stakes that appear in the fiction layer.

For example
1 the stakes in the fiction layer are 'the city blows up!'
2 none of the players in interested in 'the city blows up!' outcome

So, if we let the fiction layer drive things, we need to arrange things so that the dice don't allow the city to blow up, because no one wants that.  This can lead to the players being able to crush the end villain like a bug, because eve a 10% chance outcome of 'the city blows up!' in that fight is unacceptable to the players.  They want to keep playing in the city, after all.

-----------

But if we set the 'actual' stakes as something other than 'the city blows up!' then we can have a lot more flexibility in how much chance of 'failure' the dice system allows.

For example, at the end of Fool Moon, the fictional layer says the stakes are "Harry dies horribly" if things go wrong.

But instead, if the set the stakes as something like "How much Marcone respects Harry when this is over" then the players can allow for 'failure' because it doesn't end the game.  After all, there's plenty of interesting stories to tell when Marcone doesn't respect Harry.

[edit]
Some edits for spelling, and stuff.

And to make people 'able' to play in these sorts of games, you really need to do one of two things:

--------------------

Be able to explicitly communicate to the players that the loss in this encounter isn't really a loss, which very few games actually do. 

Were we to play a game 'naively' we'd have two things to guide us.  What the rules say, and what the game fiction says.  That is, if we get into a fight with people who want to kill us, we'd normally expect the consequences of losing that fight to be death.  And we'd expect the rules to model the winning or losing of that fight to agree with the fictional presentation of the fight. 

So, for example, if the dice say we won the fight, we'd also expect the fiction to say we won the fight.  Though it's possible to have, say, the fiction say that we lost of fight, were captured, and learned more about the villain's plans by listening to his henchmen talk before we made our daring escape.

I don't know any games that have meta rules that explicitly say things like: the stakes of this fight are a minor advance or setback.  So you probably don't want to spend too many of your metagame tokens on the outcome of the fight.  You want to save your metagame tokens to influence a more important conflict later.

[edit] whoops, Dogs does do this.

----------------

Or two, separate the game fiction layer from the dice results.

---------------

There are a few games that try this, I can think of Dogs in the Vineyard as an example.
Briefly: you roll the dice and describe what you do, but the success or failure of the dice roll doesn't need effect your description.  So, if you want, your character can actually always succeed in the vast majority of stuff that happens in the fictional layer.

Example: your character is in a struggle with one of the villains henchmen.  You want to  find out more about the villains plans. 
Sample story outcomes of a failure in the dice layer:
1. you manage to beat the henchmen down, but he has a poison capsule and you learn nothing.
2. you manage to beat the henchmen down, but "he'll never talk!" and your character won't use torture.
3. you manage to convince the henchmen of the moral superiority of your cause and he's about to tell you  everything, when another of the villains henchmen shoots him.

Sample wins in the dice layer:
1. you are defeated by the henchman and taken prisoner.  While you are tied up, you hear the henchmen talking about the villains plans.  Then you make a daring escape.
2. The henchmen gets away, but you use your investigative skills to deduce what the villains plan is.
3. You defeat the henchmen, and the threat of getting his remaining teeth knocked out is enough to get him to tell you everything he knows.
4. You convince the henchmen to be a proper and moral nihilist, like yourself.  In a fit of feeling for his fellow man, the former henchmen tells you everything.

---------------------

So we could have a loss in the fictional layer just be the 'interesting' way of narrating our victory in the dice layer.

But that takes some careful rules design and writing.  Notice how often the 'concessions are metagame constructs, not something that the in game players have to agree to' keep coming up.  The fact that you are in a fight with people that want to kill you, but a concession where you don't die is a perfectly valid result to end that fight seems to be something totally unexpected to a lot of people because the vast majority of game are driven by what happens in the fiction layer.

If the characters never face the chance of losing, they do not face adversity and challenges. If they never face a major setback and basically win at everything they do not face adversity and challenges. What they're facing is boring. What they're facing is softball validation.

Adversity and challenges mean the characters have to work to beat them. Otherwise it's, to put it as bluntly as possible, one big circle jerk. Fanfiction is filled with this sort of "We're all awesome and nobody can beat us ever" stories, and except for the ones that are parody, they generally suck.
It's not about your character succeeding. It's about your character always succeeding without there being any chance of failure.

Part of the problem here is that the vast majority of game designers and GMs do not understand compound probability.

An example:
What base chance of success does the character need to have on any give roll to succeed at 8 rolls in a row with a 75% total probability?

More in depth coverage of the example: the adventure has 8 'critical points' the the PCs have to succeed at.  Everyone wants to PCs to succeed at adventures at least 75% of the time, or even higher, because the stakes of the adventure are often something like "and the city blows up if you fail".

So, what percent chance of success at an individual test would the characters need?
If the PCs succeed on an individual test 90% of the time, they'll succeed at the adventure only around 43% of the time.   To succeed at the adventures 75% of the time, the PCs would need a success rate on an individual test of around 96.5%.  But books and cinema tend to support the characters only succeeding by the narrowest margins.  And that's the feeling that game designers often try to emulate, which means that they set the expected PC success rate to maybe 67% and the whole adventure goes off the rails the first time the dice hit the table.

That's why most stealth systems in games are completely useless.  They ask you to succeed several times in a row to get anything done with stealth, and unless you've really cheesed out your stealth skill, you'll fail one of those rolls sooner rather than later.  Which means planning for succeeding in sneaking isn't something you do.  Sneaking is just something you plan to do to let you get as close as possible to the enemy before you switch to the combat engine.

My advice is to be generous with FATE points.  The system tends to be considerably less interesting without a ready supply of them, because people are holding on o their last point or two to stave off disaster, rather than using them to do anything interesting.

The balance point to keep in mind is that having a high starting refresh should still be advantageous.  So you should be thinking about giving out somewhere between refresh and half refresh in fate points out between when you call for a fate point pool refresh.

So, for example, if you are playing 10 refresh characters, and refresh FATE points at the start of every 4 hour game session, then you should be thinking about arranging everyone to get around 5-7 FATE points per session.  To meet this ideal, you have to give out a FATE point every 10 or 15 minutes.

But how do I give out that many FATE points, you ask?  Develop the art of creating the minor cost for an action.