It's a common complaint in industry: if you're good at building things, the odds are you'll get promoted to team lead, at which point you'll be too busy herding cats to build things any longer. Similarly, professors don't actually get to spend a lot of time thinking deep thoughts; most of the hours are sucked up by administrative duties (the CS department is a 1000-strong division of a 50,000-strong firm, and stuff on that scale takes a fair bit of administering), writing grant applications, teaching, herding TAs, and so on.
But every once in a while, it's worth it. I spent an hour yesterday morning talking about possible thesis topics with four graduate students; it was the second time we'd done something like this, and it was a lot of fun. Today, two of my undergraduate students gave their client a quick demo of the system they've been building, then spent an hour or more talking about what the interface should look like, and why. The signal-to-noise may not be all that I'd hoped for, but there definitely is some signal.
(Written partly in response to Michelle's post earlier today.)