When you change thinking, your hope is that people, in time, will do things differently. If you'd rather not hope, but do, then you need to change behaviour.
When you change behaviour, you decide this is how things work now and you build that into the process. That takes commitment more than time.
In practice, this means you add accessibility to your definition of done. You're not allowed to ship anything without it being checked. That means making it part of code review, part of design critique and part of QA. And it needs to happen every time.
Here's a practical example.
Say you decide that from now on, every UI component needs to pass a basic accessibility check before it moves to done. The first few sprints are going to be painful and slow. All those checks you've never had to do before will slow you down. Everyone around you will grumble. There will be a general consensus that it was better before and subtle nudges to go back to the way it was.
You decide to stick with it nevertheless. And after a few sprints, developers adopt that as part of their day to day. They start building components thinking about accessibility from the start because they know it's going to be checked anyway.
The behaviour changed before the thinking did. You didn't need buy-in to get started. You just needed a process and the willingness to stick to it.
That's the upside of this approach. But that's also where it can fall apart.
If the new commitment isn't backed by someone with authority, people will find ways to work around it. If it's added as extra overhead without clear guidelines and why it matters, it breeds resentment. And if it's enforced inconsistently, e.g. some sprints it matters, some sprints it doesn't, people will treat it as optional.
Behaviour change works when the bar is clear and when the bar doesn't move all the time.
You can change behaviour or you can change thinking. Both approaches work. But not together. So which one should you pick? Because you do have to pick. Trying to do both at once is how accessibility initiatives go to shit.