START WITH A DECISION
Write down the task, tool and version, date, cost and what a usable result must include. Compare your usual method with the assisted workflow, counting preparation, review and corrections.
TIME
Does it reduce total time to a usable result? Record both times and what each includes.
QUALITY
Is the work accurate, complete and usable? Keep the outputs and record errors.
RELIABILITY
Does the benefit hold across repeated trials? Record failures and conditions.
EFFORT
Is setup, prompting and supervision reasonable for how often you do this task?
VALUE
Does the practical benefit justify the actual cost and expected frequency of use?
SCORE ONLY WHAT YOU CAN SUPPORT
1: clearly worse or unacceptable. 2: limited benefit with major drawbacks. 3: acceptable with meaningful tradeoffs. 4: useful with minor limitations. 5: strong benefit under the tested conditions. Use “not assessed” where evidence is missing. Explain each rating; these are judgments, not universal measurements.
EXAMPLE: ROUTINE PROFESSIONAL EMAILS
Test #001 compared 10 manual times with 3 comparable AI-assisted times from one participant. Median time was 151.5 seconds manually and 16 seconds assisted. Its provisional score was Time 5, Quality 5, Reliability 4, Effort 5 and Value 5: 24/25 overall, provisional.
That score is an exploratory assessment, not an independently validated rating. Manual quality examples were written after timing; review was not blind. Reliability has only three timed assisted trials; effort was not separately quantified, and no paid-subscription cost-benefit calculation was performed. Do not use the Value rating as proof that a paid plan is worthwhile.
CHOOSE A VERDICT
USE IT · USE WITH LIMITS · SKIP IT · TOO EARLY TO TELL.
A serious factual, privacy or commitment error can outweigh a high total. The score never chooses the verdict for you.