what percentage of accuracy is satisfactory for a machine learning model.

whenever I create a model. I usually get around 70 to 80 percent accuracy.And if I try to increase the accuracy of the model to 90 percent using GridSearch, using ensemble algorithm, by proper scaling and features selection. I hardly get to that 90 percent.The highest I can get is 1 or 2 percent increase. Sometimes the accuracy remain constant without any increase.

My question is:
Does that mean the accuracy can not increase far beyond what I have(that is 90 percent). And so, any increments might result to overfitting the model?

Or

is there another methods of increasing model accuracy that I don’t know?

so, at what point can I say I have a good model