One thing that has become apparent to me over the years is that most of these exploration and visualization tools will be pretty ineffective unless the data is modeled correctly. Some of the tools mentioned here will actually do a lot of the modeling before the exploration and visualization kick in, but the tool is probably only doing the best it can but will most likely not understand a lot of the structure and business nuance that have been accumulated throughout its existence.
Moreover, if you plan to adopt and build upon one of these tools that infers and generates the models as well as provide the explore and visualization functionality, you might be painting yourself into a corner and forcing all current and future workloads to use this layer. Otherwise you'll be having to reinterpret and reimplement your models all over the place; one off SQL scripts/reports, web analytics, dashboards/visualizations/reports on other analytics tools. Then you'll also end up having to scale this tool up in both compute and storage to handle the load that grows over time. This can end up being quite costly in time, money and responsibility.
While these tools will offer a lot of value providing visibility and insight into your data, it'll probably be worth circling back and seeing if the data can and should be modeled correctly (semantic layer) before hitching your wagon to your first choice.
Once your data is all modeled, it might be worth re-evaluating all the tools that you started with and see how they manage now that your house is a little more in order.
Remember, your modeling doesn't have to be done by the same tool that does your exploration and visualization.
This is a great article related to these ideas: https://benn.substack.com/p/is-bi-dead