When we need an LLM to tackle a complex task, we usually assume it must be fed a mountain of carefully crafted context. But what if the LLM could extract that context itself—by interviewing the human expert? This technique, often called an interrogatory LLM, flips the traditional workflow. Instead of the human writing pages of instructions, the LLM asks targeted questions to gather the necessary information. This approach not only saves time but also helps people who struggle with writing or reviewing long documents. Below are six key tactics and insights for deploying this method effectively.
1. Let the LLM Build the Context for Complex Tasks
When designing a new software feature, you typically need to provide the LLM with descriptions of user interface, implementation guidelines, details of external systems, and more. Rather than composing this context yourself, prompt the LLM to interview you. Instruct it to ask all the questions it needs to generate a comprehensive context report. You can feed it high-level information and point it to external sources it should consult. Once the interview concludes, the LLM produces a thorough context document that another session (or a different model) can use to carry out the next step. This technique shifts the burden of organization from human to machine, making the process more efficient.
2. Enforce a Single-Question Rule for Better Focus
Harper Reed, in his insightful blog post, emphasized a crucial rule when using an interrogatory LLM: the model should ask only one question at a time. This prevents information overload and keeps the conversation structured. When I tried it, I found the LLM frequently tried to bundle multiple inquiries together, so I needed to remind it repeatedly of the constraint. The one-question rule forces the model to prioritize what it deems most important, leading to a more coherent and manageable interview process. It also makes it easier for the human to respond thoughtfully, without feeling overwhelmed by a list of simultaneous queries.
3. Use the LLM to Verify Documents Through Interview
Another powerful application is giving the LLM an existing document—such as a software specification—and asking it to interview a human expert to check its accuracy. This is an alternative to having the expert read and review the document directly. Many people find reviewing hard, especially if the document is poorly written or dense. A conversational approach with an LLM can be more fruitful; the model asks targeted questions about each section, and the expert confirms, corrects, or expands upon the content. This method can uncover inconsistencies or missing details that might be overlooked in a traditional review, and it often feels more natural for the expert.
4. Combine Two Interrogatory LLMs for Robust Context
You can chain interrogatory LLMs to achieve a thorough process. First, use one LLM to interview a domain expert and build an initial context document. Then, employ a second interrogatory LLM to interview a different expert (or the same one) to review the document. This double-checking mechanism ensures that the final context is both accurate and comprehensive. If the two interviews yield conflicting information, you can resolve it through further questioning. This layered approach is especially valuable when the stakes are high, such as in critical software systems or safety‑related documentation.
5. Help Non‑Writers Capture Their Knowledge
Writing is a process that many people find difficult, even painful. Yet capturing expertise from subject matter experts is essential. An interrogatory LLM can serve as a bridge: it interviews the expert, asking the right questions to draw out their knowledge, and then compiles a written document. While the output may have a distinct AI‑generated style that some find off‑putting, it is far better than having no written record at all—or a rushed, incomplete one. This technique democratizes knowledge documentation, making it accessible to individuals who are not natural writers but possess valuable insights.
6. Apply the Technique Beyond LLM Workflows
The interrogatory LLM method is not limited to preparing context for other AI tasks. It can be used in any situation where you need to extract and structure information from a human expert. For example, you could have an LLM interview a project manager to create a detailed project plan, or interview a historian to capture oral history. The same principle applies: the LLM asks questions, synthesizes answers, and produces a structured output. As LLMs become more adept at conversation, this pattern will become a standard tool for knowledge acquisition, especially in fields where documentation is scarce or experts are overburdened.
Conclusion: The interrogatory LLM is a clever twist on how we interact with AI—turning the model from a passive responder into an active interviewer. Whether you're building context for a complex task, verifying a document, or helping a colleague share their expertise, this technique can save time and improve quality. Start with the single‑question rule, experiment with chaining interviews, and remember that even imperfect AI‑generated text is better than missing knowledge. Try it next time you need to get information out of someone's head and into a reusable format.