I am seeing a lot of resentment from leaders / managers / team members that their coworkers are using AI, which I can understand, but I feel that putting blame on individuals is wrong.
There are three issues primarily
- One, people have a hard time keeping up with AI's output. AI can generate output much faster than humans can consume it. PRs are produced at a rate which makes human review hard. A one-page PRD is now a 10-page PRD with diagrams, tables, and references. A single PR can now have 20–30 test cases instead of 2–3.
- People feel that the person generating the document would not have made an effort on the task and just gave it to AI, yielding a suboptimal result.
- AI produces output in a style which does not match your reading style / team standards. Example, your codebase follows object-oriented principles, but the AI produces functional-style code.
It is not correct to expect people not to use AI, but rather it is the responsibility of the organisation to set standards on how to use AI and make sure that every team member is aware of these standards.
An organisation without these standards is bound to produce slop which will create a lot of debt.
Every team should work on
- Setting up standards on how to use AI.
- Figuring out how these can be enforced.
- Implementing a feedback loop so that AI can review its own work.
- Building agents for each use case, for example: a PRD agent, a test-case agent, a code-review agent, etc. If anyone in the company wants to use AI, they should use these agents, which will ensure that the output is in line with the standards set by the organisation.
AI brings speed. How to channelise this speed is the responsibility of the organisation.
