Hey Herman, I was just looking at the calendar and realized how many meetings we have scheduled for the next two weeks. It is one of those stretches where it feels like the talking never stops, you know?
Herman Poppleberry here, and yes, I know exactly what you mean, Corn. It is the classic consultant's dilemma. You spend so much time in the meetings themselves that finding the time to actually document what happened and turn those conversations into progress feels like a secondary full time job.
It really does. And it is perfect timing because our housemate Daniel sent us a prompt about this very thing. He has been diving deep into voice technology and automatic speech recognition lately. He was asking about best practices for writing agendas and, more importantly, how to handle those contemporaneous meeting notes when you are using AI to format them.
Daniel is really onto something there. He mentioned that he prefers dictating his impressions of a meeting rather than just letting a bot record the whole thing and transcribe it. I think that is a sophisticated distinction that a lot of people miss. They think the AI should just be a fly on the wall, but Daniel is arguing for a more active, human led approach where the AI acts as the editor and formatter.
Exactly. He is looking for ways to keep client documentation clear and effective, especially when you are juggling multiple moving projects. So, I want to start there, Herman. Why do you think he is finding that dictating his own impressions is better than just using one of those automated meeting recorders that joins the Zoom call?
Well, it comes down to the signal to noise ratio. If you record an hour long meeting, you might have ten thousand words of transcript. A lot of that is small talk, people talking over each other, or half baked ideas that get discarded two minutes later. An AI can summarize that, sure, but it does not know what was important to the client’s emotional state or what the unspoken subtext was. When Daniel dictates his impressions right after the call, he is performing a high level data compression. He is filtering the raw data through his professional expertise before the AI even touches it.
That makes sense. It is like the difference between a raw security camera feed and a detective's report. The camera sees everything but understands nothing. The detective knows that the way the client hesitated when discussing the budget is more important than the five minutes they spent talking about the weather.
Precisely. And that leads us into the first part of his question, which is the agenda. You cannot have good notes without a good agenda. In my research into organizational psychology, the agenda is not just a list of topics. It is a contract. If you do not set the parameters before the meeting, your notes will inevitably be a mess because the meeting itself was a mess.
So, how do we use AI to make that agenda better? If I am preparing for a meeting on a complex project, say a software rollout or a marketing campaign, how should I be thinking about that pre meeting documentation?
I think the biggest misconception is that an agenda should be a list of nouns. People write things like, Project Update, Budget, Timeline. Those are useless. A high performance agenda should be a list of questions or outcomes. Instead of Project Update, it should be, What are the three blockers preventing the phase two launch? Instead of Budget, it should be, Do we need to reallocate funds from the media spend to the creative production?
I love that. It forces everyone in the room to be an active participant rather than a passive listener. And if you are using an AI tool to help write that agenda, you can feed it the notes from the previous meeting and ask it, Based on our last discussion, what are the five most critical unresolved questions we need to answer today?
Right. And you can even take it a step further. You can ask the AI to play devil's advocate. You could say, Here is my proposed agenda for the client. What am I missing that might cause a delay in six months? It helps you move from first order thinking, which is just what we need to talk about today, to second order thinking, which is what the implications of today's decisions will be down the road.
That is a great point. Now, let us get into the meat of Daniel's question, the contemporaneous notes. He mentioned dictating his impressions right after the meeting. I have tried this, and sometimes I find myself just rambling. What is the best way to structure that dictation so the AI can turn it into something professional for the client?
This is where a lot of people struggle. They just hit record and talk. But if you want the AI to give you a clean, multi project report, you need to provide some structural anchors in your speech. I like to use what I call the bucket method. As I am dictating, I will literally say the name of the bucket before I give the information. I will say, Project Alpha, Status, Green. Action Items for Project Alpha, Herman needs to review the contract by Friday. Then I will say, Project Beta, Risk, High. The client is worried about the supply chain delay.
So you are basically tagging your speech in real time. Does that not feel a bit unnatural when you are just talking to your phone in your car or at your desk?
It does for the first three minutes, but then it becomes a mental framework. It actually helps you think more clearly. It prevents that rambling you mentioned because you are forcing yourself to categorize the information as you recall it. And the beauty of modern large language models, especially the newer reasoning models we are seeing in early twenty twenty six, is that they are incredibly good at recognizing those verbal tags. You do not have to be perfect. You can say, Oh, wait, back to Project Alpha for a second, I forgot to mention the deadline changed. The AI will see that and move that piece of data to the right section automatically.
That is the magic of it. It is not just transcription, it is synthesis. But what about the tone? Daniel is a consultant. His client documentation needs to be professional, but it also needs to be honest. Sometimes his impressions might be, The client seemed really frustrated today, even though they said everything was fine. How do you handle that sensitive information?
That is a crucial distinction. I think you should maintain two separate outputs from the same dictation. One is the client facing summary, which is professional, objective, and focused on deliverables. The other is the internal project memory. In your dictation, you can say, Client Facing Note, the timeline has been adjusted to account for additional testing. Then you say, Internal Note, the client is clearly feeling pressure from their boss, and we need to be extra proactive with communication next week to manage their anxiety.
And you can tell the AI, Generate two documents from this transcript. One is a formal summary for the client, and the other is a strategic briefing for my internal team.
Exactly. And that solves the problem of keeping track of multiple moving projects. If you are doing this for five different clients, you can have the AI append these notes to a master project tracker. In twenty twenty six, we can even use agentic workflows where the AI takes the action items from this dictation and automatically updates the status in a project management tool like Notion or Linear.
You know, what strikes me about this is the speed. Back in the day, or even just a few years ago, you would have to sit down, open a blank document, and try to remember everything. By the time you got around to it, you would have lost twenty or thirty percent of the nuance. By dictating immediately, you are capturing the information while your brain is still in that high resolution state.
There is actually some fascinating research on this. The Ebbinghaus Forgetting Curve suggests that we lose the majority of new information within twenty four hours if it is not reinforced. But it is even more dramatic with complex social interactions like meetings. Within an hour, you start to lose the why behind the what. You remember that someone said no to a proposal, but you might forget the specific nuance of their objection. Dictating within ten minutes of the meeting ending is a superpower.
It really is. I want to talk about the AI formatting part specifically. Most people just say, Summarize this. But we are past that now, aren't we? If Daniel wants his documentation to be truly effective, what kind of specific instructions should he be giving the AI once he has finished his dictation?
He should be using structured prompts. Instead of summarize, he should say, Extract all dates and deadlines and format them in bold. Identify any mentions of budget and compare them to the previous meeting's notes. List every person mentioned and what they are responsible for. And most importantly, ask the AI to identify contradictions.
Contradictions? Tell me more about that.
Well, if in your dictation you say, The client agreed to the new price, but then later you say, They are still worried about the overall cost, a good AI prompt can flag that. It can say, You mentioned an agreement on price, but also noted ongoing concern about cost. Would you like to clarify if this is a firm commitment or a tentative one? It acts as a cognitive coach, pointing out where your own impressions might be fuzzy.
That is incredible. It is like having a junior consultant sitting next to you, reviewing your work in real time. But let us look at the other side of this. What are the risks? If Daniel starts relying entirely on this AI formatted dictation, what could go wrong?
The biggest risk is what I call the illusion of completeness. Because the AI output looks so polished and professional, you might assume it is one hundred percent accurate. But AI can still hallucinate, or it might misinterpret a technical term. If you are talking about a very specific software architecture, like a microservices mesh, and the AI hears it as something else, it might build a whole paragraph around a mistake. You still have to do a two minute proofread. You cannot just hit send.
Right, the human is still the final gatekeeper. And I think there is also a risk of losing your own voice. If every client update is processed through the same AI filter, do they start to sound a bit robotic or generic after a while?
That is a great point. To counter that, I recommend giving the AI a style guide. You can tell it, My tone is direct, empathetic, and slightly informal. Use active verbs and avoid corporate jargon like synergy or circle back. You can actually train the AI on your past successful emails or reports so it learns how you specifically communicate.
I love that. So it is not just formatting the notes, it is ghostwriting in your specific style. Let us talk about the multiple projects aspect. Daniel mentioned keeping track of many moving parts. How can he use this workflow to see the big picture across his entire client base?
This is where it gets really powerful. If he is using a consistent structure for his dictations, he can ask the AI at the end of the week to create a meta summary. He could say, Look at all my meeting notes from this week across all five projects. What is the biggest common theme? Or, Which project is at the highest risk of missing a deadline based on these conversations? It allows him to move from being reactive, just handling one meeting at a time, to being proactive across his entire business.
It is basically building a personal knowledge base that is searchable and queryable. You could ask, When was the first time Project X mentioned the delay in the API integration? And the AI could scan months of notes and give you the exact date and context.
Exactly. And for a solo consultant like Daniel, that is a massive competitive advantage. Larger agencies have teams of people to do this kind of tracking. He can do it by himself with twenty minutes of dictation a day and a few well crafted prompts.
You know, we have talked a lot about the technology, but I want to go back to the human element of the dictation itself. Do you have any tips for how to actually speak so that the AI understands you better? I know the tech is getting better, but there must be some best practices for the speaker.
Definitely. First, slow down. We tend to talk faster when we are excited or stressed after a meeting. A consistent, steady pace helps the ASR engine significantly. Second, enunciate technical terms. If you are using an acronym like GDPR or a specific brand name, say it clearly. And third, use punctuation commands. Even though modern AI can infer punctuation, saying, Period, New Paragraph, or Bullet Point, gives the AI a much clearer roadmap of how you want the document to look.
That is a good tip. I also find that if I am in a noisy environment, like walking down a busy street in Jerusalem, using a dedicated microphone or even just the wired earbuds with the mic close to my mouth makes a world of difference compared to just holding the phone out in the air.
Absolutely. Background noise is the enemy of accuracy. And one more thing on the dictation side, be honest with the microphone. If you are not sure about something, say it. Say, I am not entirely sure if the client meant the fifteenth of the month or the thirtieth. The AI will capture that uncertainty, which is a valuable piece of data in itself. It prevents you from accidentally turning a guess into a fact in the final report.
That is a really deep insight, Herman. Capturing the uncertainty is just as important as capturing the facts. It reminds me of how we have talked about AI in past episodes, how it is often better at helping us think than it is at doing the thinking for us.
Precisely. If any of our listeners want to dive deeper into how we have explored AI as a cognitive tool, they can always check out the searchable archive at myweirdprompts.com. We have covered everything from prompt engineering to the ethics of AI in the workplace.
Yeah, there is a lot of history there. Now, let us pivot to the practical takeaways for Daniel. If he wants to start implementing this tomorrow, what is the step by step process he should follow?
Step one, before the meeting, write a question based agenda. Use an AI to help you identify the critical outcomes you need. Step two, within fifteen minutes of the meeting ending, find a quiet spot and dictate your impressions for five to ten minutes. Use those verbal buckets we talked about, like Action Items, Risks, and Client Sentiment. Step three, use a structured prompt to have the AI format those notes into two versions, a client facing summary and an internal strategic note.
And step four, review the output for two minutes to catch any hallucinations or technical errors before you save it to the project folder.
Perfect. And for the multi project tracking, once a week, have the AI synthesize all those notes into a high level dashboard. It will save him hours of manual work and, more importantly, it will reduce the mental load of trying to keep everything in his head.
I think that mental load part is the biggest benefit. As a consultant, your brain is your primary asset. If you are using it to store mundane details like who promised to send which email by Tuesday, you have less room for the high level strategic thinking that your clients are actually paying you for.
That is exactly right. You are offloading the storage and formatting to the machine so you can focus on the analysis and the relationship. And speaking of relationships, the way you document things for a client really impacts how they perceive your value. Clear, consistent, and insightful notes make you look incredibly organized and professional. It builds trust.
It really does. It shows that you were listening, that you understood their concerns, and that you have a plan to move forward. It turns a conversation into a tangible asset. Now, Herman, where do you see this going in the next year or two? Are we going to see even more integration?
Oh, absolutely. We are already seeing real time, multi modal synthesis. Imagine a system that is not just listening to the audio but also looking at the screen share and the facial expressions of the participants. It could note that when the budget slide was shown, three people on the client side frowned simultaneously. That is a level of data that even a human note taker might miss.
That sounds like the next frontier for our research. But for now, I think Daniel's approach of human led dictation is the sweet spot. It keeps the human in the loop while leveraging the AI's strength in organization and formatting.
I agree. It is the centaur approach, the combination of human intuition and machine efficiency. It is how you stay relevant in an increasingly automated world. You do not compete with the AI, you use it to amplify your own expertise.
Well said. I think we have given Daniel a lot to chew on here. It is about the agenda as a contract, the dictation as data compression, and the AI as the master editor.
And don't forget the verbal buckets. Those are the key to making the whole thing work without getting overwhelmed.
Right, the buckets. I am going to start using those in my own life, not just for work. Maybe I should dictate my impressions after our brotherly dinners to see if we can improve our own communication.
Ha! I can see the report now. Corn was seventy percent curious but thirty percent skeptical about the new dessert recipe. Action Item, Herman needs to find a better chocolate to flour ratio.
Exactly! It would certainly make our family disputes more data driven. But in all seriousness, this is a game changer for anyone in a client facing role. It is about professionalizing the mundane parts of the job so you can focus on the meaningful parts.
Absolutely. And if our listeners have found this helpful, we would really appreciate it if you could leave us a review on Spotify or your favorite podcast app. It genuinely helps other people find the show and join our weird little community of explorers.
Yeah, it really does make a difference. We love hearing from you all. You can find all our past episodes and a contact form at myweirdprompts.com. If you have a question or a topic you want us to tackle, send it our way.
Thanks again to Daniel for the great prompt. It is always fun to explore how these new tools are changing the way we work and live together here in Jerusalem.
Well, I think that wraps it up for today. I am going to go try out some of these dictation techniques right now. I have a meeting in ten minutes, and I want to see if I can get my buckets sorted.
Good luck with that! Just don't forget to enunciate those technical terms.
I will do my best. Thanks for listening to My Weird Prompts. I am Corn.
And I am Herman Poppleberry. We will catch you in the next episode.
Take care, everyone. Peace out.
Bye for now.
