Distributed authority
Algorithmic management does not always replace a manager with an identifiable machine. It often distributes authority across a model, an interface, configured targets, historical data and the people expected to apply or challenge a recommendation.
This can speed up coordination. It can also make it difficult to identify who actually sets the norm, explains a decision or owns an error.
Preserve the capacity to discuss
An assisted decision must remain understandable to those affected. They need to know which data mattered, who can correct the system and how to raise a situation the model does not represent.
The prevention question is not whether the algorithm should be called a manager, but whether the organisation retains an accessible, competent and accountable human authority.