Is The Most Valuable AI Function Asking Better Questions?

How the “Grill Me” method became a key part of Project Weaver’s approach to AI-assisted software development.

In previous articles, we’ve shared some of the thinking behind Project Weaver—the internal engineering framework we’ve developed at profiq to help our teams and AI work together more effectively. Rather than treating AI as a magic code generator, Weaver is built around a simple idea: the better the structure, context, and engineering discipline, the better the outcomes.

One of the influences behind our latest addition comes from the wider engineering community. The original “Grill Me” concept was popularized by Matt Pocock, who introduced the idea of asking AI to aggressively challenge and interrogate a problem rather than immediately trying to solve it. His approach resonated with a huge number of developers because it flipped the normal AI interaction on its head: instead of prompting for answers, you prompt for better questions. (If you haven’t seen his original work, it’s worth exploring the resources and examples he shared with the community: https://github.com/mattpocock/skills.)

As we’ve refined the framework across real client projects, this practice has consistently proven its value. Before implementation begins, or before a significant architectural decision is made, we ask the AI to do something other than generate code. We ask it to challenge us.

Not to solve the problem. To interrogate it.

Adding “Grill Me” to Weaver

One of the lessons we’ve learned building AI-assisted workflows is that generating code isn’t always the hard part. Most delays, rework, and project surprises stem from assumptions that were never questioned in the first place.

Requirements often arrive incomplete. Different stakeholders have different interpretations of success. Edge cases hide in the gaps between user stories. And experienced engineers, perhaps because they’ve solved similar problems before, can unconsciously fill in missing details without realizing they were never explicitly defined.

Traditionally, these issues are uncovered through architecture reviews, design workshops, or the invaluable habit of having senior engineers ask difficult questions. But those conversations are constrained by time, availability, and the natural tendency for teams to become aligned around a shared set of assumptions.

Grill Me was designed to introduce a fresh, systematic challenge process into that workflow.

Instead of prompting AI with:

“Build this feature.”

We first ask:

“What would an experienced architect, QA lead, security engineer, or operations specialist want to know before agreeing to build this?”

The result isn’t a specification, it’s a conversation.

The Value Goes Beyond the Answers. It’s About the Questions.

What we’ve found using a Grill Me approach within Weaver across client engagements is that its greatest strength isn’t uncovering obscure technical details. It’s exposing ambiguity that everyone had quietly accepted.

A typical Grill Me Weaver session may surface questions around:

• Hidden business rules that were never documented.
• User journeys that break under unexpected conditions.
• Data quality or integration assumptions.
• Security and compliance implications.
• Operational and monitoring requirements.
• Scaling constraints that won’t appear until success creates growth.

None of these are novel engineering concepts. The difference is consistency.

Grill Me provides our teams with a repeatable way to apply the kinds of critical thinking that experienced engineers naturally bring to projects, ensuring that those perspectives are considered before implementation accelerates.

A Better Input Process Creates Better AI Outputs

As AI becomes more integrated into software delivery, the quality of the prompts and context we provide increasingly determines the quality of what we get back. AI is exceptionally good at transforming clear intent into implementation. It is much less effective at guessing the intent that was never fully articulated.

That’s why Grill Me has become an important part of the Weaver methodology. By intentionally stress-testing project definitions early, we improve not only the human understanding of the work, but also the quality of every AI-assisted step that follows—whether that’s architecture exploration, implementation planning, code generation, or testing.

In many cases, the exercise doesn’t just improve the solution. It helps refine the problem itself.

AI as a Thinking Partner

A lot of the conversation around AI in software engineering focuses on productivity gains and code generation speed. Those benefits are real, but they may not be the most interesting part of the story.

As an AI engineering assistant, we see Project Weaver’s potential as a thinking partner, a tool that can challenge assumptions, test ideas from multiple perspectives, and help experienced engineers see around corners before they commit to a direction.

Sometimes the most valuable contribution AI can make isn’t writing the next thousand lines of code. It’s asking the key questions that save you from rewriting the code later.

Talk to us about how Weaver can help you. Contact us.