It exists for a reason. It occurs frequently, hence cliché. As a result of its frequency, we learn to preempt it. We account for it in the set of possible outcomes. In such a set, clichés would be weighted more heavily as the chance of its occurrence is above average.
Over time, these clichés are absorbed into generalisations. As offensive as you might find such crude categorisations, it is useful. It offers unique information about the likely characteristics of a particular subset of samples.
It’s how we hedge our bets to accommodate likely outcomes.
The challenge is that clichés are superficial. It segments based on fairly arbitrary metrics that are visible, but not necessarily useful. Let’s say we want to advertise beard oil, it would make sense to target an audience of men past school-going age. Sure, some woman might use the product, as would some men of school-going age, but they would certainly be the minority.
The cliché is useful and efficient.
To launch a campaign suggesting that intelligence is liked to race would be inaccurate and non-useful. Here, generalisations would not serve us. Our sample sizes, bias and preconceived ideas, language, tests, etc trip us up.
A far more individualised approach is necessary.
This is the biggest challenge we face in workplace psychology and testing. What is general informs generally, what is specific is useful.
Don’t confuse the two.