Subscribe
Share
Share
Embed
Use structured interview questions and check the resulting draft for accuracy, context, and voice.
You already explain your best ideas to clients, colleagues, and audiences. Join Mike Giannulis, founder of Dictate, for practical 10- to 15-minute episodes on turning that spoken expertise into a published book through conversations with an AI interviewer. Each episode gives experts, coaches, and business owners one clear step to capture their ideas, shape their message, and move their book forward.
Episodes of this show are narrated by an AI-generated voice.
What if the conversation you're already having with clients is the raw material for your book? Today we're looking at what an AI interviewer actually does, and where you still have to show up yourself.
Welcome to Talk Your Book Into Existence with Mike Giannulis, founder of Dictate. Learn how to turn the ideas you already share out loud into a published book, one conversation at a time. Today, you will leave with one practical step you can use on your own book.
A quick note before we begin: this episode is narrated by an AI-generated voice.
The Questions That Pull Knowledge Out of Your Head.
Imagine an expert struggling with a blank document, then explaining the same idea clearly during a conversation. The explanation may contain a useful example, a distinction, or a sequence of steps that never made it onto the page. That conversation is potential source material. It still needs checking and shaping before it becomes part of a book.
An interview-based approach gives you another starting format. You can speak through an idea, then work with the recording or transcript. This may help if conversation comes more easily than drafting, but it does not remove the need for writing and revision. The goal is to capture material you can evaluate.
In my book, I describe three layers of an interview process. Use them as a way to assess what an interviewer is doing, whether that interviewer is a person or an AI system.
The first layer is discovery. Before any chapter gets drafted, you work through a set of foundational questions about your path, your methodology, your ideal reader, and what makes your perspective different from everyone else in your space. These aren't warm-up questions. They become the spine of the entire manuscript. These answers provide a working direction. You can revisit them if later conversations show that the reader or scope needs to change.
The second layer is content extraction, chapter by chapter. This is where the real work happens. And the key is that not every chapter needs the same thing. A chapter might be rich with stories but have no framework. Another might have a solid methodology but nothing the reader can actually do today. A good interview system tracks what you're getting against what you need, category by category: stories and examples, named frameworks, contrarian points, evidence that anchors your claims, practical tools the reader can use, and responses to the objections your ideas will face. When you know what's missing, the next conversation has a specific target. You're not asking open-ended questions and hoping something useful comes out. You're asking because the chapter needs one more concrete example, or because there's a claim that isn't supported yet.
Here's what that looks like in practice. Say you're working on a chapter about how you diagnose a client's situation in the first meeting. You've talked through the broad strokes. You have a framework with three steps. But when you look at what you've got, there are no contrarian points and nothing the reader can use immediately. The next conversation opens with two questions: "What do most people in your field do in that first meeting that you think is actually counterproductive?" and "If someone wanted to practice this on their own before they ever hire anyone, what would you tell them to do?". Those are targeted questions, not generic ones. They fill a specific gap.
The third layer is dynamic follow-up: responding to what an answer reveals. Suppose you mention a project that did not go as expected. A useful follow-up asks what happened, what you changed, and what you learned. That may uncover an example worth exploring. It may also reveal something irrelevant or too private to include. The question opens a possibility; it does not establish that the story belongs in the book.
A system with access to material across chapters may also flag a missing explanation or an apparent contradiction. Treat those flags as questions to investigate. Do not assume it will notice every gap, or that every difference is a contradiction. Ask what material the system can actually compare, and keep a separate list of unresolved points during review.
A structured question process can help organize what you know into usable raw material. Its usefulness depends on the questions, your answers, and the checks afterward. You can assess those parts directly instead of assuming the technology will capture everything that matters.
If you want to work through this process in more detail, my book, Talk Your Book Into Existence, is at startdictate.com/talk-your-book. The link is also in the episode notes.
A Worked Example: What Good Follow-Up Actually Sounds Like.
Here is a hypothetical example of how follow-up can improve an explanation. Imagine a team leader working on a chapter about handing projects to colleagues. The opening question is simple: walk me through how you prepare a handoff. The leader describes a checklist, a meeting, and a follow-up message. That gives us a starting outline, but a reader may still have questions.
The first follow-up asks for a definition: what do you mean by a complete handoff? The leader might say that sending a folder is not enough. The colleague needs to understand the next action, the deadline, the decisions already made, and the questions still open. That definition makes the advice more concrete. The interviewer should not invent it; the author needs to explain what the term means in this method.
The second follow-up asks for an example: describe a handoff that went well and what made it useful. If the author uses a real workplace situation, they should consider permission and confidentiality before including it. Changing a name alone may not prevent someone from recognizing a situation. A clearly labeled hypothetical example is another option. What matters is that the draft does not present an invented story as an event that actually happened.
In our hypothetical example, the team leader sends a short brief before the meeting. The colleague reads it and notices that the delivery date conflicts with another task. During the meeting they clarify the priority, record who will resolve the conflict, and agree on the next update. The example illustrates a specific part of the method: making open questions visible. It is not evidence that the checklist will solve every handoff problem.
The third follow-up asks about limits: when would this approach need to change? A routine handoff between experienced colleagues may require less explanation than a complex project moving to a new team. An urgent situation may need a different sequence. Asking about those differences helps the author avoid turning a useful practice into a universal rule. If the author does not know the answer, that uncertainty should remain visible until they can investigate it.
The fourth follow-up asks for a practical step: what could the reader try in an appropriate situation? For this example, the reader might review a routine handoff brief and identify the next action, owner, deadline, and unresolved question. They should use their workplace's requirements and permissions. The interview is helping the author make the explanation usable; it is not replacing the rules of the setting where someone will apply it.
Definition, example, limits, and a practical step are four useful prompts. You do not have to ask them mechanically in the same order. If an answer already contains a clear definition, move to what is missing. If a story reveals a new problem, follow that thread, then return to the chapter's purpose. A checklist can help you remember the questions without making the conversation rigid.
Notice the difference between asking for detail and supplying it. If the interviewer asks whether the colleague misunderstood the deadline, that suggests a possible event. A more open question is: what, if anything, was unclear? The author may answer that everything was clear, or identify an entirely different problem. Good follow-up makes room for the actual answer rather than pushing the conversation toward a tidy story.
This process produces raw material. A recorded answer may repeat itself, use shorthand, or rely on context that a reader does not have. Turning it into prose involves choices about order, emphasis, and wording. Keep the original material so you can check those choices. Mark gaps for follow-up instead of filling them with plausible details.
The useful distinction is between a draft grounded in your explanations and a draft produced with little information from you. Your own material provides a stronger basis for representing your thinking, but the result still needs review. An interview can capture what you said. It cannot by itself prove that what you said was correct, complete, or appropriate to publish.
Why You Still Have to Review Everything.
Now consider the review that follows the interview.
The interview system does a lot. But it doesn't replace your judgment. The draft that comes out of the process is a starting point, not a finished product, and treating it as anything else is how you end up with a book that technically contains your ideas but doesn't quite sound like you said them, or quietly gets something wrong.
Let me tell you what can go wrong, specifically.
AI can misstate facts. If you gave an answer that was ambiguous, the system will make a choice about what you meant. That choice might be wrong. A name gets slightly off. A number from a case study is transposed. A sequence of steps gets reordered in a way that's grammatically clean but logically backwards. The consequences depend on the subject and the error. Check them before publication rather than assuming a small change is harmless.
AI can omit context. You know why a particular step in your framework matters. You've seen what happens when people skip it. If you did not explain it during the interview, the draft may omit it or fill the gap incorrectly. A review pass is where you notice those gaps and fill them.
AI can smooth away your distinctive language. This one is subtle, and it's worth paying attention to. You might have a phrase you use consistently with clients, a specific way of framing a problem that your audience recognizes as yours. In the process of turning conversational answers into polished prose, that phrase might get replaced with something more generic that means roughly the same thing but doesn't carry the same weight. That's a loss worth catching.
So what does a useful review actually look like? There are five things worth checking, and none of them require you to read like an editor.
First, check every proper noun. Names of clients, companies, books, people you've referenced. Compare these with reliable records rather than relying on the spelling in a transcript.
Second, check every number. Not just statistics. Any number. Percentages, timeframes, quantities from case studies. If a number appears in the draft, confirm you said it and that it's accurate.
Third, check every attributed quote or claim. If the draft says someone said something or a study found something, make sure that's actually what happened. The system can conflate two things you said in different conversations or misremember something you mentioned in passing.
Fourth, read the reasoning. Not just the sentences, but the argument. Does each section actually say what you meant? A paragraph can be grammatically correct and logically broken. Read it as if you're the skeptical client who's going to push back.
Fifth, read it aloud. Not to perform it. To catch the places where the voice isn't yours. If you find yourself thinking "I would never say it that way," that's a signal. Revise it until it sounds like you actually said it, because your readers who know you will notice.
This review step is where the author's responsibility is clearest and most non-negotiable. The interview system extracts. The AI drafts. You remain responsible for confirming that it represents your thinking. Editors and appropriate subject specialists can help check it. You do not skip it because the draft looks polished. Polished and correct are different things, and a book with your name on it needs to be both.
The good news is that reviewing a structured draft is a different job than writing from scratch. When you have organized, categorized content that came from your own answers, you have something concrete to check and refine. You may still need new explanations, research, or substantial rewriting. Plan time for that work. But it is an ask, and meeting it is part of what makes the book genuinely yours.
Before your next interview, choose one idea you explain regularly. Prepare a definition, an example you can appropriately share, and a situation where the idea needs adjustment. Label invented examples clearly. During the conversation, notice which follow-up question makes your answer more specific. Afterward, compare a short draft with the recording: what was preserved, what changed, and what still needs checking? That small exercise gives you a practical way to evaluate an AI interviewer before relying on it for a whole chapter.
Take the next small step on your book today. For more guidance, find Talk Your Book Into Existence at startdictate.com/talk-your-book.
Visit startdictate dot com slash talk your book.
Next time on Talk Your Book Into Existence: Find the Method Inside Your Everyday Work. Turn a process you regularly explain into a sequence a reader can understand and try. Follow the show so the next episode is ready when you are.
Thanks for listening to Talk Your Book Into Existence. Follow the show for your next practical step, and share this episode with someone who has a book in them. Your next conversation could be the start of a chapter.