The Experiment

I performed 4 real work tasks manually first, then repeated each task with AI assistance (Claude by Anthropic). I recorded time taken for each and rated the quality of the output on a scale of 1–10 using a consistent rubric. The goal was to move beyond statistics and test AI's impact on real tasks directly.

Summarising a 500-word legal paragraph
Writing a short news article (150 words)
Analysing a small dataset to identify trends
Brainstorming solutions to a business problem

Results

Task	Manual Time	AI Time	Quality (Manual)	Quality (AI)	Notes
Legal summary	12 min	45 sec	7/10	8/10	AI was faster; missed one nuance I caught
News article	18 min	1 min	7/10	7/10	Similar quality; both needed editing
Data analysis	25 min	2 min	6/10	9/10	AI found patterns I missed entirely
Brainstorming	15 min	1 min	8/10	7/10	Human ideas more original and context-specific

        AI reduced task time by 92–94% across all tasks. Quality gains were strongest in analytical work; human judgment remained superior in creative and contextual work.
      

What I Learned

Speed is not the whole story — a 45-second AI summary is only useful if it's accurate enough to rely on.
AI outperforms humans on tasks with clear rules and lots of data (analysis, summarisation of factual text).
Human judgment still wins on tasks requiring originality, context-sensitivity, or understanding of unspoken stakes.
The most effective approach was using AI to generate a fast first draft, then applying my own critical review — this is augmentation in practice, not automation.