The analysis phase is a sequence of steps, and they are not equally suited to automation. Steps that follow fixed rules and produce checkable output — cleaning and validating the dataset, generating the standard tables and figures defined in the analysis plan, and running consistency checks between outputs — can be automated reliably, because the correct result is defined in advance and a machine can be tested against it. Steps that require judgement — deciding what an unexpected result means, choosing how to handle a deviation from the plan, and writing the interpretation — stay with people. The practical picture is a chain in which the routine middle is machine-run and the judgement at either end is human. The time saving comes from the routine middle, which is also where most of the repetitive hours sit.
Learn MotionHow AI Makes Clinical Trials Faster: A High-Level Overview
Analyzing Results and Reporting Sooner
Which Steps a Machine Can Take Over
2 / 4
Look at the chain from left to right. The first steps — cleaning the dataset, checking it against the rules, producing the standard tables and figures — are highlighted, because those are the ones a machine can run. The reason is not that they are easy; some are quite fiddly. It is that the correct output is defined in advance, so a machine can be tested against it. The steps at either end are not highlighted. Deciding what an unexpected result means, or how to handle a deviation from the plan, has no fixed correct answer, so it stays with a person. Notice where the bulk of the repetitive hours sit: in the highlighted middle. That is why automating it moves the finish line.
0:00 / 0:00