← All posts

Nov 14, 2025

Using AI Carefully in Engineering Workflows

Where AI helps engineers, where it creates noise, and how to keep it grounded in reality.

AIEngineeringWorkflow

AI is useful in engineering, but only when it is treated like an assistant, not an authority.

Right now there are two common mistakes:

  • people dismiss AI completely because some generated output is sloppy
  • people trust AI too much and let it produce plausible garbage at scale

Both are bad engineering habits.

Where AI actually helps me

AI is strongest when the task has:

  • a lot of mechanical structure
  • obvious output format
  • high drafting cost
  • human review still required

Good examples:

  • writing first-pass documentation
  • generating repetitive test cases
  • turning rough notes into clearer issue templates
  • creating draft postmortems
  • summarizing config risks
  • converting project notes into project pages or case studies

Where AI becomes dangerous

The most dangerous zone is where the output looks competent but is wrong in ways that matter:

  • infra configuration
  • security recommendations
  • migration scripts
  • concurrency-sensitive code
  • cost assumptions
  • production troubleshooting steps

Final thought

Use AI where it sharpens the engineer. Avoid it where it starts replacing engineering thought.