A trial runs its stages one after another, not side by side, so the time saved at each stage is subtracted from the same overall clock. That makes the total saving roughly the sum of the individual savings rather than an average of them. If design work is shortened, recruitment is shortened, retention is improved, monitoring decisions arrive sooner, and analysis and reporting are compressed, the trial finishes earlier by the combined amount.
The size of each saving is not equal, and it tracks two things: how long the stage normally lasts, and how much of that stage is repetitive checking work. Recruitment is the longest single stage, so even a moderate percentage improvement there produces a large absolute saving. Analysis and reporting is shorter but is almost entirely checking and restating, so it responds strongly to automation. Design sits in the middle. The clinical conduct of the trial — the period in which participants actually receive the treatment and are followed — barely moves at all, because it is bounded by biology and by the need to observe outcomes that take time to occur.
This is why a realistic picture is a shortened timeline with one stubborn block in the middle, not a uniformly compressed one. The savings cluster at the front and the back, and the middle stays roughly where it was.