Adventure design is a tricky thing. Certain things can make it easier, most notably the buy-in from your players. Few things forgive a bad adventure design like players willing to go "Ok, the adventure is clearly over there. C'mon guys, let's head on over." And they'll do it, because your players are great and you're lucky to have them, but that might make you feel a little guilty about giving such great players such a half-assed setup.
So, sooner or later, you want to give them more head. Not that you were railroading them before, but rather you were giving them pretty linear adventures where their primary role was to react to bad things and eventually go in the dungeon. Clearly, they want something with a few more moving parts, and just making dungeons more complicated doesn't seem to be doing it, so what do you do?
This is the point where you might be looking to turn a corner from encounter-driven adventures to dynamic adventures, but don't turn it just yet. Let's make sure we choose the right tool for the job.
Encounter adventures are familiar to most of us, as they represent the bulk of dungeon games (however the dungeon is dressed up). The GM has prepared a number of scenes which will be encountered then resolved through play. There are numerous ways these encounters might be sequenced from a strict dungeon map to a wild and loose random table to a sophisticated set of if-then statements, but the formula is pretty much the same.
While this approach is most familiar to us in dungeons, you can find it in every game system out there, with shadowruns and vampire parties constructed in this same fashion, as a sequence of encounters. Part of this is familiarity, but part of it that this model really _works_. Scenes allow for focused events that are (hopefully) cool and specific. Their real strength (and weakness) is that they draw bright lines around the event and allow it to effectively be resolved in a vacuum. You're fight in a dungeon will usually have little bearing on your fight in the next room of a dungeon, and your trading of barbs with the duke will have little bearing on your tea with the chancellor.
While this means that you can really drill down and focus on the scene in question, it also can really end up suspending disbelief if things get too clearly delineated. This is only so much of a problem with dungeons - we're trained to overlook how horribly fake those feel - but it's part of the double edged sword of trying to apply the dungeon model to things that don't have hard walls, like social situations.
One of the classic problems with trying to run a city campaign is the question of _how_ to do it without just treating is as a crowded space filled with small aboveground dungeons. The historical instinct has been to gravitate towards turning manors and sewers into interesting dungeons and treating the rest of the city like an oversized version of the town outside the dungeon.
For a GM trying to let go of the dungeon and loosen up play, there are a few tricks to easing the transition, but the most important one is to loosen your grip on the encounter model. That is not to say you need to stop using encounters entirely, but rather that you should focus on the encounters that will really rock, and put less work into building "tunnels" between them. Allowing players to meander a bit makes their arrival that much more satisfying.
It will also, somewhat paradoxically, make everything feel a lot more realistic. If you can find a module that uses someone's manor house as a dungeon, this will illustrate it very well, but if not, just consider it. People do not live in houses the way we imagine monsters living in dungeons. We do not sit in rooms and wait for things to happen - we move around, do things, read books, eat meals and so on. If someone were to break into our home or office, the response would be much more dynamic than it is in a dungeon.
With that in mind, when we describe a home or office in terms of a dungeon, we get something stilted that feels entirely off. If, in contrast, you just have a particular scene in mind for the whole house (rather than on a room-by-room basis) then you can usually play that in a way that seems reasonable. A fight breaks out in the ballroom and things will happen as a result of it, but if you do it as one scene, there's no need to them move on and clean out the kitchen.
Now, here's the dirty trick - you can take this thinking back to the dungeon to bring it to life. No to say you need to do this with every dungeon and every encounter, but taking a part of the dungeon and thinking "Ok, here's how things work here, whose involved, and how things will play out" you can make a bigger, more freewheeling encounter which will probably be more interesting and more fun than usual. For a good example of what such an encounter might look like, listen to Mike Mearls talk about his lunchtime games sometime. As he describes them, they are basically giant, multi-element encounters, and there's a lot of juice in that model.
Now, all that said, you may eventually find yourself wanting to move onto another approach, possibly one where the PCs are more proactive than reactive. And when that happens, it's time for tomorrow's topic; KWORC.
1 - Though he may not think of them as scenes, that's what they really are.
2 - And is, even more than many indie games, a story technique. By breaking the adventure into discrete scenes, you make it flow better and make each scene better and more memorable. That is to say, if you've run a dungeon, and you don't details the hallways as much as the rooms, then you may be a story gamer. If the boss fight was in one of the last rooms, then you're definitely a story gamer.
3 - There are exceptions to this, but even they are fairly telling. Sometimes results from one encounter will have spillover results (like a guard triggering the alarm resulting in subsequent encounters not being surprised). It might even bring in monsters from the next encounter. But it is a rare adventure where the alarm seriously disrupts the structure of the dungeon.
4 - And intrigue, of course, which means "Going into dungeons for important people".
5 - Especially if you start wit a dungeon by someone like Owen KC Stephens, who does a great job of writing dungeons which are basically extensions of the villain.
* - This isn't a footnote on anything, but if you really want to learn how to run a good dungeon, read and play Dogs in the Vineyard. I know you may have heard all sorts of wild stuff about it, but it's worth getting to know if only to see how it structures adventures. Its towns are, for all intents and purposes, dungeons. They're constrained spaces with a problem to be solved. The big difference is that instead of being a dungeons of rooms, its a dungeon of people. That makes things both simpler and more complicated, and to a DM, it turns certain ideas on their ear while maintaining a very familiar framework. If you can absorb DitV, you will have a new and very powerful tool in your design toolbox