08-12-2026, 08:16 PM
Every week there is a new AI product promising to replace a workflow. The problem is not a lack of tools; it is choosing a small stack that reliably improves one outcome.
A simple evaluation process
1. Write the job in one sentence: for example, “turn customer-call notes into approved follow-up tasks.”
2. Define success before testing: time saved, error rate, cost per task, or quality score.
3. Start with one tool category at a time: model, research, automation, transcription, or analytics.
4. Run the same five real examples through each option.
5. Check the unglamorous details: exportability, privacy, permissions, API limits, failure handling, and team adoption.
My rule of thumb: do not add a tool unless it replaces a manual step, connects cleanly to an existing system, or produces a measurable improvement. A good AI stack is often three dependable tools, not fifteen subscriptions.
What is one task in your work that you would most like to simplify, and how would you measure whether an AI tool actually helped?
A simple evaluation process
1. Write the job in one sentence: for example, “turn customer-call notes into approved follow-up tasks.”
2. Define success before testing: time saved, error rate, cost per task, or quality score.
3. Start with one tool category at a time: model, research, automation, transcription, or analytics.
4. Run the same five real examples through each option.
5. Check the unglamorous details: exportability, privacy, permissions, API limits, failure handling, and team adoption.
My rule of thumb: do not add a tool unless it replaces a manual step, connects cleanly to an existing system, or produces a measurable improvement. A good AI stack is often three dependable tools, not fifteen subscriptions.
What is one task in your work that you would most like to simplify, and how would you measure whether an AI tool actually helped?
Building small AI tools and automations in Bengaluru. Notes on what ships, what fails, and what I'm learning along the way.