That's a huge question. I think that if you haven't read much about David Allen's Getting Things Done (GTD) task management approach, it would be helpful for you to do so.
I have OmniFocus set up with 10 top-level folders representing different areas of my personal and professional responsibilities. Within each, I put projects (anything that takes more than one step to complete), and most have their own list of single actions.
For instance, I'm a history professor, and I have a folder called "Book and manuscript reviews." Each time I review a book or a manuscript, I create a project for it. In the project, I put the first couple of steps that I need to do, often beginning with a "waiting for" item until the publisher sends me the manuscript or the review editor sends me the book. I assign a "waiting" context to those items, so that when I go over my context lists, I can quickly see everything I'm waiting for, and when in planning mode, I know what the status is at a glance.
Once the book arrives, I tick off the waiting for item. Usually I've entered the next task--read the book--and I'll add a time estimate after glancing over the book to see how long and how difficult it will be. Sometimes, though, I won't have entered a next action.
This is where the review feature is great. I use the Inspector to set a review interval for each project (usually every week, but sometimes more often and sometimes less often). Every day, I open the Review perspective and look at every project that's set for review today (and any previous reviews I might have missed--my system isn't perfect). When I review an existing project, I ask myself a series of questions:
--Have I completed the project? If so, I mark it complete.
--Have I defined a next action? If not, I define one.
--Have I been making sufficient progress on this project? If not, I might shorten the review interval. A review is a form of reminding yourself that you've made a commitment to do something.
--Is this still a project I should be working on now? I.e., have my other commitments changed so that this project is less important? Depending on the answers, I might put the project on hold and set a review date in the future when I think I might want to make it active again. If the project is unlikely to ever get my attention, I might drop it.
--Is this project becoming urgent? If so, I'll flag it.
--Is this project's due date (if it has one) still realistic? If not, I'll add an action to renegotiate the due date with whoever needs the results. I don't like doing this, but it's better to do so than to simply miss a deadline.
The great thing about the review process is that it allows you to do minimal planning while still ensuring that you get stuff done. You don't have to plan out a whole project at once (which is a great invitation to procrastinate); instead, you just have to ask yourself, "what is the next concrete step (action) I can take toward finishing this project?" Once you've defined that, you can rest assured that you'll have something in your context lists that will move the project forward. If you review the project frequently enough, you'll be able to add additional steps to keep the project underway without overplanning.
Of course I'll often enter multiple steps, especially for a complicated project like teaching a semester-long course, with action groups to separate different kinds of tasks and to sequence them properly.
When it comes to doing things, my approach varies. Because I took my vacation late this summer, just before the semester started, I'm working a lot from my Due Items perspective, to make sure I don't miss any important deadlines. Otherwise, though, I tend to do two things:
1) In Context mode, select (command-click) the contexts in which I find myself. I'll then look at flagged items in those contexts, do them if I have time and energy, and then move on to non-flagged items.
2) In Planning mode, use the Focus command on a folder, then scan the projects to see if there's something that needs attention in a different context. I'll then switch to Context mode, while keeping the focus, so I can select the contexts I'm in and see only actions that are connected to that area of responsibility. That's how I keep certain administrative jobs from taking up all the time that I really ought to be devoting to working on articles and books--or at least, that's how in theory I do it!
At the beginning and end of the day, I try to quickly scan the Due and Flagged perspectives as a way of ensuring I keep on top of things that are urgent, important, or both. I also borrowed from Curt Clifton the idea of a Tickler perspective that is set to show actions that begin today and tomorrow, so I know what I should start to consider--and, if necessary, push the start date forward when circumstances change.
I hope those examples give some idea of the real-life possibilities of OmniFocus. I've been using it for nearly two and a half years and it's become indispensible; I had used Life Balance before but OF is just so much better for my needs.