For numbers, 5 and 51 undo each other: 5⋅51=1, and multiplying by 1 changes nothing. For matrices the role of 1 is played by the identity I=[1001], since IA=AI=A for every 2×2 matrix A. The inverse A−1 is the matrix with AA−1=A−1A=I.
Take A=[3423]. Its determinant is 3⋅3−2⋅4=1. Swap the diagonal entries and negate the other two to get [3−4−23], then divide by the determinant, which here is 1. Check by multiplying: row 1 of AA−1 is (9−8, −6+6)=(1,0) and row 2 is (12−12, −8+9)=(0,1). The product is I.
The division by ad−bc is why the determinant decides everything. If detA=0, the formula would divide by zero. Geometrically, A has flattened the plane onto a line: different points have landed on the same spot, so no matrix can send each one back. Such a matrix is called singular.
Two look-alikes are not inverses. The transpose flips rows and columns, turning [3423] into [3243], and their product is [13181825], not I. Taking the reciprocal of each entry fails too. Only the product test AA−1=I decides.
Think of it like this. A matrix and its inverse are like a scrambling code and its unscrambling code: encode a message with A, decode it with A−1, and you are back where you began. A matrix with determinant 0 is a shredder rather than a code: it sends different messages to the same output, so no unscrambling code can exist.
Core idea. A−1 is the unique matrix with AA−1=I, and it exists only when detA=0.