The reason is accountability
A machine can produce a number, but it cannot be responsible for it. Trial results inform treatment decisions and regulatory judgements, so a named person has to stand behind them. Human review is not a temporary limitation waiting to be engineered away; it is a requirement of the claim being made.
Verifying is faster than producing
The saving survives because checking a machine's output is quicker than generating it by hand. A reviewer who can see exactly which rules were applied and which source records were used can confirm the result in a fraction of the time it would take to reproduce it. This is why traceability matters more than raw speed: an output nobody can check is not a saving, it is a risk.
Where the saving stops
If the automated output is wrong in a way the reviewer cannot detect, the time saved is illusory and the error is worse than a slow one. The value of automation in this phase depends on the reviewer being able to see what was done, not just the result.