I see two natural extremes worth exploring. The first extreme's first thought would be something along the lines of.
Oh, no. Not again.
But we'll get back to that one in a moment, because the second extreme is oh-so-interesting. It would be reasonable for the building to develop as an infant does. This would imply that "what would be the first feelings and thoughts of a living building" is the same question as "what were my first thoughts when I was an infant." But what were your first thoughts? Can they be written down? It's not currently clear what distinguishes our first thoughts from the chaotic soup of neuronal activity from which they come forth. It would be hard to show clearly the first thoughts of a building for the same reason.
Of course, raising a building from an infant state takes a lot of work. If you've ever raised a newborn you'll understand when I say the ahem plumbing doesn't always work for the first few months. It turns out it's not easy to manage a gastrointestinal tract, and it certainly can't be easy to manage indoor plumbing. To say one would have to endure a few cold showers would certainly be an understatement as this infant building learns to control its functions.
Which leads us to our first line of reasoning. Those making these living buildings clearly know they need to sell their product, and would not want to sell infant buildings. They need to sell adult buildings, ideally past the rebellious age (it turns out grounding a building for bad behavior is more tricky than you'd think!). They certainly couldn't afford to raise them for years before selling them to a contractor for installation. They would need a fast process to make quick money. The solution would be to "seed" these buildings with training regiment that quickly grow them into the building we want.
There would certainly be times during this training where there's things a building would have to learn that only another building could teach them. We would have noticed this, and fed them what training we could. Not knowing what else to do, we'd probably let each generation of building develop its own content to train the next building with. Those buildings which did not collapse or go insane would be permitted to pass this information on. This would not be a kind cuddly process. This is business. The lessons each building would have for the next generation would clearly show the brutality of growing up in a building mill. So, we would expect the first self-aware words out of our building's mouth to reflect this:
Oh, no. Not again.
Of course, once the building is installed, as you guessed, they would engage in self modification. Which self modification they choose to start with is really quite open ended and depends on the environment the building is placed in. When deriving fiction, I always start from where reality has gone first. If you look at the world of self modification for humans, you can see everything from wearing make up to tattoos to plastic surgery. A common trait of these is that all of them are designed to present an image to others, defining our identity. Buildings don't have too much trouble with this in general -- they are long lasting products, each one unique. However, you may decide that AI's could be swapped in and out of a building. This would make the AIs more transient, and you would expect to see similar body modifications to what humans do.
Likewise, for the buildings that are secure in their identity, their modifications would be very dependent on their interactions with people. One building may modify themselves to be more amenable to humans. Another may grow fearful and replace locks on the doors to protect its core circuitry. Still another may dream of one day being free of this concrete and steel shell and to fly among the birds. Make no mistake, the responses our living buildings would have would be varied indeed.