AI at Work: Measure the Task, Not the Hype

Before asking whether AI can transform your business, ask which task it can improve—and how you will know.

The research

Anthropic’s January 15, 2026 Economic Index report introduced five dimensions for studying Claude use: task complexity, skill level, purpose, autonomy, and success. It analysed a November 2025 sample of Claude conversations and first-party API activity.

This is evidence about one provider’s observed usage. It is not a survey of all workers, and it does not establish what a particular Philippine business will gain from adopting AI. Its measures also rely on model-based assessments, which should be interpreted with care.

Sir Next’s perspective

For a small business, the useful starting point is a clearly defined activity. “Improve marketing” is too broad. “Prepare a first draft of next week’s product captions using our approved product facts” gives you something concrete to assess.

My suggested approach is to compare the full workflow. Count the minutes spent preparing instructions, reviewing the result, correcting mistakes, and approving the final version. A quick first draft has limited value when the team spends longer fixing it.

A practical pilot

Choose one repeatable task and record how it is done today. Define what an acceptable result looks like before testing AI. Use a small set of examples, then compare completion time and quality with the usual process.

Keep the person responsible for approval clearly identified. At the end of the pilot, ask: did we save effort, maintain quality, and make the process easier to repeat? If the answer is unclear, refine the task before expanding it.

These are practical suggestions drawn from the research, not measured outcomes for Filipino MSMEs. The decision to continue should depend on your own results.

Source

Anthropic Economic Index: Economic primitives · January 15, 2026. Research explainer, not a breaking-news announcement.