There are no 100% isolated systems - even the universe itself may not be truly isolated (depending on how you define universe).
However, in practice, many systems are "isolated enough" that we can treat them as conserving mass and energy, especially over short time frames or small distances.
For example, we often treat a well-insulated and closed reaction vessel as isolated when doing thermodynamic calculations, and this approximation works very well for fast reactions.
Another example where this works is an adiabatic process. Many processes can be successfully modeled as adiabatic even though they happen in poorly-insulated containers - for example, the rapid expansion or compression of a gas.
Another example is the use of small sub-domains of a non-isolated system, which over short time scales behave as isolated systems themselves. For example, in finite element methods (FEM), a system is broken up into very small sub-domains. Each one acts as a small system in which mass and energy can transfer across the boundaries. Locally, an element is an open system, and over long time scales the entire system is open, but on short time scales, the sets of domains that share boundaries are treated as isolated. If energy leaves one domain, it enters the neighboring domain - it doesn't completely leave the system. In effect, you can model an open system as a large collection of isolated systems, with the outermost having open boundaries.
As you can see, even though the assumption of an isolated system is not really true, you can find many, many places where it is "true enough" that we can use it practically. In fact, it would be very hard to do anything in thermodynamics without it!