Ep 80: Separating AI Discovery From Execution

Episode Summary

This solo episode focuses on a practical framework for getting better results from AI: separating discovery from execution. Dave explains that instead of treating AI like a one-line command tool, it is more useful to treat it like an intern, contractor, or agency partner that needs context, direction, and clear expectations before it can produce strong work.

Dave introduces the discovery phase as the setup work that happens before asking AI to create an asset. This includes explaining the goal, identifying missing information, defining tone and format, and building a reusable master prompt. He shares how this approach helped him create an AI project manager that could track projects, organize updates, draft stakeholder messages, and identify quick tasks when he had limited time.

A key tactic Dave recommends is using voice memos to quickly brain-dump context, then turning those transcripts into structured AI inputs. He also emphasizes the importance of telling AI not to summarize or truncate important details when the goal is to preserve project context. Once the context is built, the execution phase becomes much easier, whether the output is an email, Slack message, weekly update, or other communication.

The episode also applies this discovery-first approach to personal branding and creative work. Dave discusses using AI to develop brand standards, identify tone of voice, analyze transcripts of how you speak, and build a “craft layer” based on your preferences, influences, and standards. He uses examples from his own personal rebrand and creative writing process to show how AI becomes more useful when it understands your taste, constraints, and voice.

The larger takeaway is that generic AI output usually comes from insufficient context. By separating discovery from execution, building reusable prompts, and teaching AI what sounds and feels like you, users can create more useful, repeatable, and personalized AI workflows.

Ep 80: Separating AI Discovery From Execution Podcast and Video Transcript

Dave Dougherty: All right. Hello and welcome to Enterprising Minds. A little bit different episode for you today. Just a solo show with me and we're gonna do it kind of quick. A little shorter than normal, hopefully, but we will, uh, we'll see. I want to really focus in on some things that could be useful takeaways from what.

I've been noticing in my own playing with AI and the things that we've talked about on the previous episodes about how really chunking the activities between discovery and execution have been very, very helpful. And then also getting how to get your output to sound more like you, even if you don't have a whole bunch of.

Historical content that you can load up to get it to sound like yourself and less like an ai. So before we get into it, just a quick reminder, go to enterprising minds.com. It's enterprising-minds.com, where we have all the episode transcripts, links to things that we say in the episodes.

And, um. Ways that you can reach out and connect with us, we would love for you to connect with us and let us know what you are thinking and what has worked for you based on what we've talked about. Secondly, pathways I've been remiss in. Posting recently, I'm sorry. Um, getting that back up and running, please go subscribe on our new Substack version of it.

You can also find it on LinkedIn adding a bunch of new stuff. Have a whole content backlog to post on that. So, if you want to subscribe and get everything from subset, you can do that. You can get this podcast. You can get all the newsletters right there in your inbox or in your app. So, like, subscribe, share, all that good stuff.

And now let's get into it. So.

AI Overwhelm Mindset

Dave Dougherty: One of the common things that I've heard in working with coworkers or working with friends when we're talking about even just getting started in ai, right? Because it's really easy to get overwhelmed with implementing AI to the point of just being like, you know what? Screw it.

I don't want to deal with this. And I understand that it's really technical. It challenges all the ways that you've normally done things. Depending on what your job and industry are, but you know, it's, it's important that we get an understanding of, of how, uh, it can help you. When you are prompting, if you are just putting in a single sentence like, Hey, do this.

Go find this. That's okay. That's definitely where everybody starts out with, but ultimately, if you want the best projects and the best outputs from your ai, you should really think about the AI as if it were an intern or like a contract worker that's helping you do the work. So, they need all the context that you can provide so that they go and do a good job, right?

If you go to a person and just say, go make me 200 social posts, you're gonna be really surprised with what comes back because you didn't have any kind of direction around, I want it to look this way. My social media should target, you know, these channels with this kind of tone of voice, all of that.

But also creating those guidelines can be very daunting, right?

Discovery Versus Execution

Dave Dougherty: So, this is where what I'm calling the discovery and the execution piece of it really should be separated out. So, what I'm typically doing, and what I'm finding very helpful for my own things is to start out a chat with, I'm trying to do this.

Here's what I'm thinking from the start. Tell me what you would need to create a master prompt. What. Additional things should I give you in order to get this thing done. And then you can work with the AI to then create the project brief and the, the larger prompt for that activity. And then that way it has all the context and you have stress tested it a little bit to make sure that you get a lot of the output.

So how, what does that look like?

Project Manager AI Build

Dave Dougherty: For me, one of the more recent things that I worked on was a project manager, AI at work. I have way too many projects going on and I was finding that my normal way of keeping track of where all the projects were was too much and because it took too much. Of my own sort of administrative time to keep things up and going.

So for me, this is a perfect use case for ai, right? It's just a little bit of extra help and allows me to have sort of a running list of things where it can remember the details and then it can help me write the statuses to, whoever my stakeholders are that I need to reach out to. So. One of the things that I did was, Hey, here's my situation.

I'm running all of these projects. They fall into two different buckets, you know, new products, and then this priority project, and then I've got this third one that, does this, that and the other. These are sort of the prime primary metrics for each, and here's how I've been going about it so far.

And so I gave it all the context around all of those things. Now, mind you depending on what your projects are, if this is what you're doing for work, make sure that you are staying totally copacetic with the, uh, company guidelines around what you can share with the ais. So I was using an internal ai where.

I could share a little bit more than if I was using one of the outside company ones, third party ones. So make sure you keep that in mind as you're setting this up, because that's going to change how you do the outputs, right? So I started talking through it and then I realized, okay, I might not have a whole bunch of time to just type all of these things out, but if I go and I use my voice memo app.

You know, you can find that on get that on your iPhone. Android has one I use Mac laptops, so the voice memo app is on that as well. So I just turned on the voice memo app and then just started talking and then it automatically transcribes what I'm saying. And then I upload that into the AI as context saying, Hey, I just did this brain dump.

Organize it in this way. To match up with the projects that I've told you about and the sub-projects that I've told you about, and then make sure that you're capturing all of it. One important thing if you want to do this particular project manager AI piece is to make sure that you tell it to not summarize things to make sure that it keeps all of the information that you tell it.

Otherwise, it will try to truncate it. So that it optimizes for its own memory, and then you might get some weird outputs later. So keep that in mind. That was one of those iteration things that I learned uh, as we went through. But the nice thing about this project manager AI is after I did that like 40 minute brain dump transcript into it and had it organized, I could then look at it and say, okay.

These are the correct spellings of the people that I mentioned, you know, for their names. These elements were listed incorrectly. Please correct them this way moving forward. I need to do, these kinds of things later. And then I did that whole thing of, all right, now that you have all the context, help me write a project manager master prompt.

That I can copy and paste each day. 'cause the internal AI doesn't have a memory like a Claude or a uh, Google or, whatever else. So I have to do this each time, right? So it's important for me that you have this master prompt to understand what I need you to do. And then secondarily is all of this context.

Of all the different projects I'm working on. And so every day I load it up, I post I pro, I put in the master prompt for the context, then I update the do the second one that has all of the context in it. And then as I go about my day. And have my meetings and have all my different notes, I make sure that I'm updating that context and I'm saying, Hey, I just crossed off this activity.

Uh, I sent this email to somebody and yada. And then it tracks it, it tells me so that at the end of the week I can then send an email saying, Hey, here are all the things that we did this week moving forward into next week. This is what I'm looking to do. That has been massively helpful for me.

Right, and. That part of it is really just the discovery. It's setting all that up and making sure it has the context and has the prompt for it to work is gonna be, that is one activity in and of itself, the execution piece coming off of that project management thing. Yeah. Are some really nice things.

Right.

Daily Workflow Execution

Dave Dougherty: So if I find myself with a half an hour of where I don't have to do anything, which is often rare. But if I do, then I can just say to it, Hey, I got 20 minutes. What are some things that I could quickly take off the list of things to do? Who should I follow up with? Who, you know, where, what project needs, uh, a quick hit on something and then boom, I can have it, write a quick draft, put it in, go in, or go into, teams or Slack and say, Hey, just checking in on this.

What's going on? Tell me what's up. Uh, it has been massively helpful because then I've been able to actually keep things going at a relatively good, a relatively good clip. Also, it takes the pressure off of my memory which is not there with how much is going on. Now, with the execution piece of it, you're gonna have to understand like what you actually need to do, right?

So if you get the random requests on, Hey, what are you working on for this particular portfolio, or you know, this project that's going on what's the status of it? Well then, you know, you need to output some kind of email for your boss. Or, stakeholders and you as the individual who has dealt with those people will understand the context and the way in which that they prefer to be communicated, right?

So that's gonna flavor the way that these outputs come out. So, again, doing another one of those little discovery pieces, if you're unsure to be like, Hey, I have this project manager prompt. It's working great. This is what it's doing. I need you to be the execution piece of this. Write a prompt for that to understand how to do the outputs that I need to have from this.

And it can be a, it can be a wonderful thing. So whether or not that's a quick teams or Slack message, whether or not that's social posts or emails to your boss or emails to your stakeholders, whatever it is. Mark those down, write those out. And then have those in there and saved.

You can do this for literally anything going through and, chunking out the tasks that are involved like you would for any. Coworker or intern or contract worker that is helping you out, even your agency relationships, right? I mean, how many times do you do a kickoff meeting where you have to say, okay, here's the scope of the problem.

Here's what we're trying to do, this is how we usually go about it. Here are our brand standards. I mean, all this stuff is typically the kickoff meeting that you would do.

Brand Standards From Scratch

Dave Dougherty: But what if you don't have the brand standards? What if you don't have the examples of previous work that you've done, right?

Maybe you are starting a substack, like we are. Or you're finally going to launch your own, your own website, but you want it to look professional. So what do you do? This is where that discovery piece comes in and really focusing in on that. So one of the things, if you're really starting from zero, you, I would say go to Google Gemini, do a deep research project and just say, you know what?

I'm trying to do this. I need to know what are the best practices right now for brand standard internet brand standards. Okay what should I include in a media kit or in a personal brand standards document, photography document, et cetera, et cetera, right? Once you do that deep research, then at least you have some context you can load in and say, Hey.

To my knowledge, these are the brand standard or these are the best practices around brand standards for a personal brand. Right now I'm trying to create brand standards. Interview me. This is another approach that can be really great with your ais. Interview me so that you can get the information you need to help me create the output, which will be a word doc, a PowerPoint, A PD, F, whatever of.

Colors, typography, mission, voice, all that stuff, right? And then again, I find having the conversations again, either with the voice memo app or doing the voice activated things in the AI apps, I find talking it out is actually much easier than typing everything. Obviously you go do what? It feels best to you, but I'm finding that really helpful to me.

And I'm also finding that not a lot of people actually understand that you can do a lot with the voice settings already. Go and have a conversation with it, right? The being able to say, Hey, I like these authors or these thought leaders. These are the elements of their writing, their presentation, their podcast that I really enjoy.

You know, when I was doing my personal rebrand for dave doherty.com, I I started talking to him and said, all right, here's my current one. This is what I'm not liking about it. Here's what I do like about it though. And when it comes to design, I really like how. You know, Nike does the basketball logos for the custom shoes for basketball players, right?

So the KD logo or the Ja Morant logo or the Kobe logo, right? I like those a lot. So it became kind of a vision board thing. I also think F1 cars look awesome and they're sort of an example of, you know, using colors in a way. That just looks really good in my opinion. I also took, you know, one of the guitars in my collection and said, I actually really like this color palette.

How can we include that into the brand standards? What would that look like? What are these colors? And it was a black, a gray a raspberry ish color, and we saw a building from there. Now granted, I also have an aunt that was a graphic designer, so I called her up and said, Hey, help me out with the imagery for this.

So it's not all ai. But then I definitely took the work that she and I collaborated on and the designs that she built for me, which were wonderful. And then now I've put those into, Canva and other platforms to be able to say, here are my brand colors. Here's how I'm gonna go through. Here are the, here are the fonts that are acceptable.

And you can just slowly create this over time, right? But then also as you go and try to implement this, either on a website provider or out into either own, um, your own space, like maybe you're redesigning your LinkedIn profile to match your new personal brand standards. Well, then you're gonna have to be like, great, now I have to create this LinkedIn banner.

What, what do I think about that? What are the aspects for that that I need? What are the elements that I need on a LinkedIn banner to be taken, you know, in this way? Another thing with the voice that I really, really like is if you don't know how you come across when you're talking, going through and saying, Hey, here's a transcript of me talking.

Pull from it, the tone and the elements of my speech that you could then apply to social posts or other things to make the output sound more like me. Okay. And sometimes you may be happy with what it comes off out with. Other times you might be, you know, be less inclined to say yes to what it says about the way that you present yourself.

But all that's good information, right? So for me, I know I'm an overly direct human being, at least for Midwest standards, right? That is definitely part of my tone of voice standards, my brand standards. I'm gonna be direct, I'm gonna be straight to the point. But also professional. And, you know, you can just, you can list all those things out.

Obviously they're not top of mind because I put 'em in the brand standard so I wouldn't have to remember it. Yeah.

Quick Recap Framework

Dave Dougherty: Anyway, I threw a lot at you right away. With this, just a quick recap before I get into a more creative a more creative example. So I separating the discovery and the creation of the master prompt from the execution rather than just saying, here's what I'm trying to do.

Go create this asset. Take a little bit of time and set up the master prompt so that it could be a repeatable thing, especially if it's a repeatable thing. Take the time to create a master prompt that's going to give you exactly what you want in the way in which you'd expect the outputs to be. So that discovery piece is where you apply all the context, where you find out things that you might be missing or you didn't think of straight away, right? Take that time to build that out, to get the best possible outputs. And then once you have the outputs, iterate that again to make sure that you're getting exactly what you want.

When it does an output, and then you can go back and update your master prompt from the discovery piece so that it becomes the thing that you want. And then that, that whole process can be eventually automated with exactly what you want to have happen. Then getting the output to sound more like you.

If you don't have additional content, you can just have it interview you. Just sit down and talk for a while and see what comes out of it. Have it detect your tone of voice, the way in which you present your word choice whether or not your sentences are choppy or do you talk for a really long time and you do a lot of, you know, run on sentences.

It's all, it's all for you to find out. So that's gonna be my big suggestion. At least for this episode, and I'm gonna wrap up with one, an example, a more creative example than business for before we go. Oddly enough, as more and more people in my friend cohort have been getting into ai, the extra time that they have.

It has freed up either time or mental space for them to play around with creative writing. And as somebody who has a creative writing degree and as somebody who enjoys writing this has been awesome 'cause I've been able to have some conversations around character development and styles and all of these things that I really enjoy nerding out on.

And, you know, interestingly, I was not just one friend. It's like four, so something's in the water. But as I was talking to all these people, I've been writing a lot of poetry recently. But I wrote about seven chapters of a novel. In college for my capstone in order my undergraduate degree, and I have not touched it since then.

So I went back to it and I uploaded what I turned in for my capstone into Claude. I said, here's the context for this. I would like you to do this, that, and the other. Rate it, see where it's at. Tell me what's working, tell me what's not. Yada. Then I realized that I actually was trying to have it do way too much, right?

Like the first examples, I was trying to have it do discovery and output things and do logical work and do it. It was too much. So I had recently attended the Marketing AI Institute's AI for Writer Summit. So if you haven't attended one of these, it's a totally free thing. It's an awesome an awesome event done by them.

I have no affiliation other than I went there and enjoyed it. But one of the presenters for the AI for writers went through this five step process of how he wrote a. Like horror romance novel that he self-published on Amazon. And there was actually like a five step thing and these big master prompts around idea discovery, around character development, around editing, around publishing and summarizing and, all these different things.

So I took those screenshots that I had from that and I said, Hey, help me build out these things. But as I was building those things out, I realized that there was nothing about taste or craft of things that I prefer that make me more of a writer, right? So I started going through it and I said, you know what?

I need an extra layer of information for you to actually do what I want you to do more so than just the prompt. You needed, like an additional context layer of who I've been reading, what I like, why and where I see this story going. So I started going through and I said, Hey, I like Kurt Vonnegut.

I recently reread Hemingway and I really like Hemingway, but lately it's been like overly macho for me, and I don't really like that aspect of it, but I really like the way that he puts together sentences and the way he kind of describes things without describing things. I've been on a Russian literature kick for whatever reason, so, you know, a lot of Chekhov recently.

And what I really like about Chekhov is X, Y, and, and Z. I also, these are my favorite movies. These are my favorite rappers. Even though that has nothing to do with you know, writing a novel, it does show you at least the word play and, other forms of writing that I really enjoy. So it created all of this context around how I write, what kind of content I consume.

In the, the literature space and sort of who my favorite authors are from there. And then it was able to build out, here's the writer profile of Dave. He's already been published in these things for this kind of work, for novel writing, yada, yada, yada. But I also made sure to specify like, all right, if I, I don't want you to do any creative writing.

On my behalf, unless I specifically tell you that I'm going to publish it under a pen name, if it's going to have the Dave Dougherty name as on the author line, I'm going to be the one that writes everything. So I, in that case, I need you to just be an editor and tell me how good it is or where things are left off or whatever else.

So I spent a lot of time building all this stuff out and then towards the end I found myself completely stuck creatively. And I was reading through all the suggestions, you know, that we had and, and whatever. And I just, you know, I was like, man I'm really stuck right now. This is where I'm getting hung up on things.

And it just doesn't feel like anything is actually at stake for anybody. And so I started talking through about like, okay, if, you know, these are my favorite, movies where there's definitely something at stake. Now granted, it doesn't have to be as explicit as, you know, espionage or anything like that, but.

Help me understand in a different context, like what is at stake for each of these characters and what's at stake for them overall. Because that needs to be emphasized so that people are actually interested in, in reading. And that's where I was massively stuck, but because we had all the context of all the stuff that was written prior and then the iterations that we've done based around.

Who these characters are and what they would do, what they wouldn't do, sort of what are the swim lanes for each of these characters. I was able to have this really nice conversation around the context that all of them are operating in now.

Craft Layer And Constraints

Dave Dougherty: So why the hell would I be talking about this on a business podcast?

One of the important things from that lesson is that craft layer. Here are my preferences. This is how I can, how you can sound more like me or understand what it is I'm going to say yes to when you're pitching me ideas, right? Because whenever you're on LinkedIn or any of the other socials and they start talking about here, the 20 AI tells who cares ultimately?

If you, if it's easy to tell that AI wrote it, then you haven't done the work to make it sound like you. So having this context layer is really important, but in addition, also having a clear list of things that whatever it is you're trying to do are not, so this task would be considered a fail if you.

Create it as a T XT file, right? I only want it in a, a Word document file, those types of things. So not only what, what is the output I'm looking for and what is the project I'm looking to to accomplish, but also here are the things that it's not. I've found all of these things to be massively helpful in my prompting as well as ways of looking at scoping projects for the AI agents.

Wrap Up And Call To Action

Dave Dougherty: Hopefully this is helpful. I think it, it is definitely been helpful for me to, to put some context around these thoughts. Uh, like, subscribe, share as always, you know, you can find me on LinkedIn or dave doherty.com. And, and keep up with me at any of those places. Please let anybody else know about this.

It's helpful for us on the. On the algorithms and we love hearing from you guys on what it is you're working on and how things are useful for you. So go check these things out, try it out. Let me know what you think. If you have never tried separating that sort of discovery context work from the execution work.

So, um, thanks for sticking around and I'll see you in the next episode of Enterprising Minds. Take care.