The example he used was the sort of problem where there's a obvious, but wrong, answer to a problem (e.g. "If a bat and ball cost £1.10, and the bat is £1 more than the ball, how much does the ball cost", to which a lot of people instinctively answer 10p - using what he refers to as "System 1 thinking"). These sorts of question were presented to two groups of people, one with the questions clearly printed and one with the questions badly printed in very light colours, requiring the readers to concentrate to be able to even read the questions.
The surprising result was that people in the latter group did significantly better on average than those in the former, suggesting that because they are forced to start using the more conscious, and more conscientious, "System 2 thinking" to be able to read the questions, they are primed to use that sort of thinking on actually answering the question.
This isn't always true though - if you're already concentrating heavily on something, then introducing another problem that requires a lot of concentration, you will stop concentrating on the first issue.
So, introducing something that forces your brain into a low-level System 2 mindset (such as slightly wonky offices) might be a good way of reducing the number of "obvious but wrong" decisions being made, but wouldn't be a good way of improving problem solving for tasks where people are already concentrating on the problem.
For those unaware, I recommend reading up on "inattentional blindness". Attention is a very valuable resource.
Basically, when I'm at risk to lose focus from task at hand, this steals it, and it's rather boring compared to what would most likely be highly abstract contemplations, totally new ideas and so on, so your focus returns to original task since immediate alternative provided is dull affair.
It somehow "eats up" part of my nature that nudges me to lose focus. It's maybe analogous to how rain noise gobbles up various distracting sounds.