Like all elements of startup planning, the only thing you know about a budget or forecast when you finish it is that it's wrong. You just don't know how wrong it is or in what way it's wrong.
This is a difficult reality but it isn't a problem. Startups aren't about reading the crystal ball; they're about your ability to react. That doesn't mean you shouldn't make predictions, you just need to be willing to recalibrate them constantly based on new data. There are two things that you can predict: revenue will take longer realize than you expect and expenses will be higher than expected. Budget accordingly.
Every CEO—startup or otherwise—should be able to roughly model their business and understand its main financial drivers. These might change over time. To be more precise, these will change over time, probably quite drastically. That isn't an excuse to avoid making predictions about your cash flow and your expenses. It's the reason why you have to continually adjust those predictions based on new realities.
Financial modeling is not only a task to be strictly delegated to your CFO. That person may own your financial model but you need to be extremely comfortable working your way around it, changing variables, understanding how little changes you make in those variables ripple through from concept to cash.
“That new client is going to be huge” isn't a satisfactory prediction of cash flow and “We're going to ...