A common approcah to using AI in projects is to provide detailed prompts, requirements and instructions, asking AI to generate a solution. While effective, this approach assumes we already understand the problem and know exactly what we need. But do we actually?

What if, instead of telling AI what to build, we give it a basic idea and ask it to question us, challenge our assumptions, and clarify our requirements?

In this approach AI is not a builder yet. It is a smart client or a QA, who understands the problem but needs us to explore the requirements and edge cases. AI is an interviewer here, asking us about business objectives, users, existing process, constraints, exceptions, and dependencies. This approach improves requirement quality, reduces misunderstanding, and avoids assumptions. We think critically on the problem rather than directly jumping to the implementation, just because we have access to AI.

Here is the same prompt in the two approaches.

Execution oriented approach: Design an automated employee onboarding workflow. Include document collection, IT access provisioning, HR approvals, and status notifications. Suggest the workflow and implementation approach.

Question oriented approach: I want to improve our employee onboarding process. Act as a business client and ask me questions to understand the problem, current workflow, user expectations, exceptions, and business constraints. Ask one question at a time, challenge my assumptions, and identify gaps. Do not propose a solution until we have sufficiently clarified the requirements. Then summarize the requirements for my validation.