ChatGPT agent mode: The “little helper” that transformed recruiting, crafted user personas, and solved parking nightmares | Michal Peled (Honeybook)

[00:00] We're going to start with something that we haven't actually seen on How I AI yet, which is agent mode in ChatGPT. My use case was with our hiring team. Part of their workflow is to browse through many LinkedIn profile and search for relevant candidates. It takes a lot of time. Let's talk about the prompt. I'd love for you to go through how you thought about structuring it to make it effective with the agent. [00:30] if you're an ID recruiter. And then I described what I wanted to do. I love that you called it your little helper because don't we all want an AI little helper? Welcome back to How I AI. I'm Claire Vem, product leader and AI obsessive here on a mission to help you build better with these new tools. Today, I have Michal Pled from HoneyBook. [00:53] They're a technical operations engineer who's building tons of internal tools and automations [00:59] to make their team's life easier and reduce friction. [01:02] Mihal is going to show us some advanced features of ChatGPT, including Agent Mode, [01:08] replicate not one but five of their personas as AI identities and [01:14] save me a lot of time on my commute using ChatGPT. I'm really excited about this episode. Let's get to it. This episode is brought to you by Brex. If you're listening to this show, you already know AI is changing how we work in real, practical ways. Brex is bringing that same power to finance.

[01:33] Brex is the intelligent finance platform built for founders. With autonomous agents running in the background, your finance stack basically runs itself. Cards are issues, expenses are filed, and fraud is stopped in real time without you having to think about it. Add Brex's banking solution with a high yield treasury account and you've got a system that helps you spend smarter, [02:03] runs on Brex. You can too at brex.com slash how I AI. [02:11] Michal, thank you so much for joining How I AI. I'm excited to see what you have to share. [02:17] Thank you so much for having me. We're going to start with something that we haven't actually seen. [02:22] on how I AI yet, which is agent mode, [02:25] in chat gbt and so i'm wondering if you can just go ahead and dive into what was the problem that you were trying to solve and [02:34] Why was this agent mode, this agentic browsing, the solution to the problem we were having? [02:40] Our problem was, you know, same as our customers are having. You have to do your job. You have a job that you really love doing and you have your... [02:52] proficiencies and expertise [02:56] However, you spend a lot of your time doing the mundane, thoughtless, [03:03] manual repeating work in order to do

[03:06] to get the information that you need. So my use case was with our hiring team. [03:12] And as a recruiter, when you get a job description that you need, [03:18] to recruit to find candidates for [03:21] part of their workflow [03:24] is to browse through many LinkedIn profile and search for relevant candidates that may be relevant for the job descriptions. [03:37] And it takes a lot of time, can be hours of browsing through profiles and going through all of the [03:44] characteristics that they're looking for. [03:46] So I wanted to take that load off of them. [03:50] and ChatGPT agent mode came just in time. We all talk about [03:56] What... [03:57] agent is and what agents do and how we can use them. In ChatGPT it's very simple to understand. So you just open [04:07] a chat with ChatGPT, [04:09] but then you add an instruction [04:11] and turn it into an agent mode very simply. [04:15] from the toolbar. [04:16] And once... [04:18] goes into agent mode it means that it can take the prompt or you can actually use specific prompts [04:25] to tell it not just to search for information online, [04:30] but also to perform actions for you. [04:33] And why did I need it in this case?

[04:36] because I needed to log in to LinkedIn. I don't want it to just search for profiles on LinkedIn [04:43] just [04:44] just profiles that are [04:46] publicly accessible. [04:48] That's not the information that I need. [04:50] So I needed it to log in into LinkedIn and I needed it to perform search and I needed it to go through the profiles and look for the restrictions that I... [05:01] want to give it. And those restrictions were provided by the actual [05:06] hiring team that they actually use it as [05:11] requirements for potential candidates that they find. [05:15] Yeah, let's talk about the prompt really quickly, because I think this prompt is interesting as I'm reading it. And I'd love for you to go through how you thought about structuring it to make it effective with the agent. [05:28] Of course. So I usually start my prompts, begin my prompts with telling the GPT [05:35] its role. [05:37] And so here I told it you are an IT recruiter. [05:40] I want a little helper, right? I'm a recruiter. I want someone who is like me that will assist me with my job. So I started by telling it you're an ID recruiter. [05:51] And then I described what I wanted to do, what the task is. [05:54] Log in to LinkedIn using my account. If not already logged in, let me take control and log in. It is something that is possible. [06:04] find up to five LinkedIn profiles where the current title and job description match the attached job description.

[06:13] And here is the part. [06:15] where I just uploaded [06:18] job description. In this case it's for an engineering role, okay? [06:23] So I have the job description and I have the [06:28] You are an IT recruiter. This is your job. [06:31] this is the task, and I [06:33] provide like a full description of the task. [06:37] Actually, [06:38] actually describing what an actual IT recruiter would do. [06:42] And then I added restrictions or special instructions, it doesn't matter how you call them. [06:49] But these are important because this gives [06:52] don't just search for something that matches the description. [06:56] Do it the way that we do it. And when our hiring team goes into a search, they have specific criteria that they go for. So I collected these and [07:09] and I added it as a list. [07:11] as a restriction, I could call it instructions. It would have been the same. [07:16] So candidates must be from Israel because the job is being filled up in Israel or currently working at an Israeli company. And they must be active in LinkedIn within the last three months because that's something that our hiring team does. [07:33] is looking for. [07:34] uh and the current job job role must be close enough [07:39] to the open role in title and seniority.

[07:44] And also, something that is special, the candidates must either [07:49] work in their current workspace more than a year, [07:53] Or they can be unemployed but no more than a year. [07:57] and have worked in their last workplace [08:00] for over a year. [08:02] These are all things that I didn't invent them. [08:06] They were taken. [08:08] specifically from our iron process. [08:10] What I love about this is exactly what you said, which is, first, I love that you called it your little helper because don't we all want an AI little helper? That is my goal. Maybe I'll rebrand my product to just little helper. But what I like about this is... [08:26] You know, when you're building a tool like this or a prompt like this, the simplest way to get to a good outcome is simply interview somebody and say, [08:37] Step by step, just tell me what you do. Like, tell me what you do. And if you can codify what a person's step by step workflow is, and you can put that into a pretty simple prompt, which here it's only a paragraph and three or four bullet points. [08:53] You can replicate and automate that at scale and typically [08:59] This is not the highest order thinking you want your recruiter or sourcer to do. You don't want them just to build a list and be looking, is this person here a year or not? That is an input to what you hope is a great recruiting process, great outreach, all that kind of stuff. So one thing. [09:14] I think it's just really great to interview your colleagues and say,

[09:17] how do you do your job and what parts do you hate? And let me automate them. [09:21] The second piece that I think is really interesting here is you're actually pretty specific about a couple outcomes. You're specific about the number of candidates that you want. So I think that's really helpful. You're specific about a kind of threshold of matching your criteria. So you say 70%. And I often find these LLMs are very, you know, what is 70% matching? [09:51] So I think that is really interesting. And then the last piece I wanna call out, which we'll maybe see in the demo, [09:57] is [09:59] Agent, while it can be agentic and independent, can also be a co-pilot and collaborator. And so you actually instruct the agent when you're going to take over and when they're going to take over. And so I think that is also really interesting things for folks to know is you don't just have to like press the button and walk away and let the agent run. [10:18] You can press the button and say, hey, wait, when you get to this point, let me take the next step. And then you can go on from there. So that's really interesting. [10:26] Exactly. Or if you encounter a problem with this and that, stop and ask for... [10:32] my assistant. [10:33] And that's exactly the agent mode. Thinking about it as a little helper will really help you come up with good prompts for it.

[10:47] Okay, I think you gave us our show title will be... [10:51] ChadGBT agent mode. [10:53] your little helper. So let's see how it runs. Exactly. [10:56] Exactly. So once you start running it, and this is something that is mind-blowing for anyone who... [11:05] or tries it for the first time [11:08] even the ones who are, you know, very proficient with using AI tools, suddenly you get like a computer opening up in your, you know, [11:20] You see your little helper actually doing things in the computer. [11:26] So it started by reading my job description, and then you can see it goes... [11:33] It will try to go to LinkedIn. It will probably be already logged in because I logged in beforehand. And, uh, [11:40] The thing that I like the most [11:43] you see it's logged in and then you can see like the arrow goes and clicks on things and searches on things and go through the list and [11:51] The thing I like the most is that during all of this time, you can see the thoughts. [11:57] Poverty agent. [11:58] Now I will go to the feed page loaded again. I plan to click on [12:04] First, I need to make sure you can see inside the brain of your agent, [12:11] while it is thinking. [12:13] And so, and all of this is live.

[12:17] So I will let it run here, but we can go and see results. [12:22] Yeah, you know, I want to call out a couple of things because I know that often on How I AI, we have highly technical use cases for highly technical folks, but we also have a lot of people that are actually quite new. [12:34] to using generative AI tools and have probably been pretty familiar with ChatGBT and the direct experience. But I know if I showed my mom this or even some of my friends that maybe don't work in tech and said, hey, did you know that [12:48] ChadGBT can open up a magic computer and navigate it and narrate its thoughts and look for things for you. They'd be pretty, pretty surprised. And I think. [12:58] You know, as you watch this, what I hope our listeners and viewers are taking away is you don't just have to rely on text prompts and chats when you're using these tools. Now that the next kind of like evolution of these LLMs, especially the more like consumer focused general purpose ones like ChatGPT have evolved now. [13:23] you actually have a lot more tools. And so I just wanna call out, so for some of those folks out there, I'm thinking a little bit of my parents and some people in an older generation who are like, [13:35] how do I get from here to here? I need help searching for flights, or I need to do a certain kind of research on a niche site. [13:43] having sort of this expert computer operator on hand, I actually think is going to make information more accessible to folks, but it's also going to make UX and websites more accessible to folks that don't have the time to figure out how do I use the best filters on LinkedIn or those sorts of things. And so I just, I want to make sure that people that have not experienced agent mode, and I know we're all on the edge now.

[14:09] So maybe all of us have. Just, you know, take a minute to appreciate the kind of use cases that this... [14:14] opens up also it's just fascinating to watch [14:17] Yeah. AI operating. [14:20] Of course, first time I tried it, I just sat and watched the thoughts of the agent while it was thinking. Going to our more technical audience, though, a couple of things that I want to call out is one, props to the OpenAI ChatGPT team. What a great... [14:37] user experience design here. [14:40] It could feel very strange to watch an agent browse a website. It could either be boring or weird. And I think this user experience of being able to see where the cursor is, showing the reasoning and thoughts, watching it navigate is actually pretty entertaining. That's a hard thing to pull off for a consumer product. And so for anybody designing AI products, it's worth thinking about some of these things. [15:04] interaction patterns here and then again [15:08] I just think about how long this would take someone to do. We're watching it because I'm trying to narrate some of the features, but... [15:16] You know, you could walk away, you could go to a meeting while this happens or you could do something else. If you have ChatGPT on your phone, you will get notification and notification when it's done to tell you that it's done and here are the results for you. We said little helper.

[15:46] executive leader in an organization, I was constantly... [15:50] looking for like [15:51] who's a senior director of DevOps and Platform Infra, who is either in San Francisco or works for a San Francisco based company, who has experience in DevTools. [16:04] That, you know, one change I would maybe make on your prompt is like, is one or two connections away so I can actually... [16:10] message them or get a back channel reference on that. I did this all the time. I had my hiring managers do this all the time. So even if you're not in recruiting and you're just somebody who does hiring, I think the specific workflow is really... [16:23] really useful. [16:24] Yeah, definitely. That's an excellent use case that you just mentioned. Well, let's actually... [16:32] Let's look at the output. We'll let this run, but I know you have an example output for us. [16:36] Yeah, yeah, Franz. I have an output that actually worked for 10 minutes. [16:42] just that and [16:43] Within these 10 minutes, I got a list of five candidates, as I requested for. And you can see the match score. [16:53] Having a match score or rank for results is something that I really love doing. It's not a must. [17:01] But... [17:02] If you give specific, if you provide specific requirements and ask for a match score, it is easier to understand [17:11] What results are more [17:14] have more quality for you.

[17:16] I mean, otherwise you could just get a table of like, these are the five results, but you [17:21] Is someone of them better or... [17:25] may be a better match for what I need. [17:29] Then the others, I won't be able to see it unless I instruct the GPT [17:35] to provide me with some score. [17:37] So it's not an exact science, but it does give you [17:42] some kind, a way to compare between the results. [17:47] So I will say someone who got 90% match [17:52] is probably as like probably... [17:56] will be a better match than the 78% match. [18:00] And I will have to go deeper and understand why. [18:03] Yeah, I'll call it a couple of things here that I think are interesting for people to look like. I know we were talking before the show, you actually made this anonymized data just so we weren't showing people's profiles or showing how person A versus person B fit a specific job you're hiring for. [18:20] But I will say anonymizing candidate profiles is actually a pretty standard practice and a lot of recruiting flows just to make sure [18:29] You're not biased. This school, that school, this person, that person, this name, that name. And so I actually kind of like this flow where you're really just comparing the qualifications against your stated objections or objectives. And so I think that is a really interesting kind of meta flow that you're showing here.

[18:49] The second thing that I wanted to say as I was reflecting back on agent mode is it's almost exactly like a recruiter or sourcer would navigate LinkedIn. [19:00] Except for one thing. [19:01] When I log into LinkedIn, I don't go straight to the job to be done. I don't go straight to that search bar and search for like VPs of engineering. No, no. [19:09] I get distracted by the notification. I start reading the feed. I'm responding to comments. I go through my inbox. And so I think like, [19:19] Why is it 10 minutes? It's 10 minutes because it's like pretty hyper, hyper focused and efficient, but it's also 10 minutes because you're not getting so distracted with all the other things in, in the application. And you can really just get the agent to focus on the task, task at hand. So maybe it's a way for us to all break. I saw, sorry, LinkedIn, LinkedIn growth PMs. I apologize. But a good way for us to still get the value of these platforms without getting our time sucked into the, um, [19:49] less value generating aspects of them, maybe. [19:52] Yeah, that is correct. [19:55] So tell me a little bit about how this was received by the team. I'd love to know the kind of outcomes here. [20:00] Yeah, well, [20:02] I, I... [20:03] I will be real and say that on [20:07] Like the first result that I got, I was very skeptic about. [20:11] So I just took the table, [20:13] and I sent it to our hiring manager. [20:16] And I told her, this is the job description. This is what AI found for me.

[20:21] uh in linkedin if you can go through the results and let me give me feedback [20:28] Are there good results? Are we familiar with these candidates? Did we try to reach out to them? Or [20:35] you're looking at them and say, "Oh no, that's a terrible fit. I don't know why this person is even in this table." [20:43] And so she went over them. You can see that the table as [20:49] link. [20:50] to the LinkedIn profile per candidate. [20:54] So she scanned those five profile. [20:57] And she came back to me and she said, "You know what? Out of these five, [21:02] Four of them were never found by us manually, and they really fit the description. So we would want to approach and... [21:13] you know, try to get them to come for an interview. And the fifth one was actually one that we caught manually [21:21] and is already coming for an interview. [21:25] And so to me, it was a great sign for quality. [21:30] I mean, it's not just a list of names. [21:33] Those are actual... [21:35] real quality candidates that we can work with. [21:38] And so now they want the agent to run on... [21:44] a lot more job description, many more job descriptions that we have. [21:49] provide them with more than five candidates. I wanted just five to see if it's worth

[21:55] Something? [21:56] - Cool. [21:57] Yeah, but now it's going to be a real part of their hiring process. [22:02] freeing their time to do other things that they love and appreciate a lot more. [22:09] Yeah, and I can't emphasize what you're saying enough because so many people push back on AI saying, yeah, you may get speed and you may get efficiency. [22:19] but you're not getting quality. And my experience has been the opposite of that. You get speed and you get quality. And again, it's that last mile, those edge cases, those ones that are just a little hard to find, a little hard to research. [22:34] where I think AI can increase the quality of that last bit. And so it's amazing to see that this work for your recruiting has given me so many ideas. Not just the recruiting use cases, but just in general, one people finding use cases. I was thinking about how you could find great candidates for [22:51] or customers on like X or LinkedIn. And then the other thing that I think is is really great here. [22:57] is just showing we don't get a lot of gna functions we don't get a lot of people functions getting love in how i ai i feel like all the noise is about like product design engineering support and so i just love seeing the recruiters get some love here because you're the people that bring in great talent and fun colleagues to work with so [23:19] Thank you for showing this.

[23:49] bed. [23:50] Let's zip to your second use case, which I think is really, we're going from finding real people. [23:57] to creating fake people. So I'm excited about this. That's an excellent description of it. So let me ask you that. Imagine you're a business owner and imagine being able to [24:12] to talk to thousands of your potential customers all at once and gather their insights on your [24:19] planned ad campaigns, planned features, product experience, all from your phone or tablet, 24-hour a day with one click of a button. [24:30] actually talk to them. I thought it's mind-blowing. And so it started with [24:38] HoneyBook invested in comprehensive customer research with a third party provider, [24:45] we interviewed hundreds of our target small business owners. [24:49] and they created five detailed buyer personas. [24:54] But the research was trapped in documents, hundreds of [24:58] pages of insights that themes rarely referenced because it was too time consuming to extract actionable [25:06] information, [25:07] when making product or marketing decisions. [25:10] So the end goal was we have five personas. [25:14] We want to talk to them. [25:16] Let's create a chat GPT.

[25:19] that is that person, that actual person. [25:23] And so I started with [25:25] And here that [25:27] There are some technical takeaways, but here I want to put the spotlight on the thinking process. [25:34] Because it's very easy to go to chat GPT and everyone with a subscription tier of... [25:42] Um. [25:43] It's not... [25:44] a plus subscription tier and above can create their own custom GPT. [25:51] And so you go to create a custom JetGPT, [25:54] It's quite simple process. [25:59] You add a name to your GPT, a description, but the most important parts are the instructions you're providing it with. [26:06] and the files that you can upload as a knowledge. [26:10] for that chat that you're talking to. So I needed [26:15] five like them. But first I thought, okay, this is all I know [26:20] about custom chair GPTs. [26:22] I can basically take all of the documentation from the interview and just [26:26] upload all of the files, text files, presentations, whatever into this custom chat bot. [26:32] provided with some instructions on how to answer what the research was about, what to say, what not to say, and ask it questions about [26:42] the research itself, but that's not what I wanted. [26:45] So I was like, okay, if I'm taking [26:48] just the files.

[26:50] per persona so i'll concentrate in one persona take the files related to that persona upload it to the chat [26:58] instruct the chat what that persona is, how to read the files, what's included in the files, then I'll be able to maybe ask about that persona, [27:10] And I will probably get answers like: [27:13] That persona would have done this, or that persona would prefer that, but it's not like talking... [27:20] to the persona. I'm still talking about the persona. [27:24] and not with it. [27:27] And so I realized I'm not going to rely on uploaded files. [27:33] for the chat GPT. I need the instructions, [27:36] to be [27:37] the main and most important part [27:42] of what consists of that persona. And the instructions here, [27:46] will not be what's in the file or talking about someone else. [27:51] there will be exact instructions that go like, you are that person. [28:00] and this is your belief system and this is how you run your business and this is how you deal with media. [28:06] and this is how you deal with technology. [28:09] And everything has to be super, super, super tight. [28:12] So that's ChatGPT once running live. [28:16] will actually become that persona [28:19] that can answer based on what it knows about itself, not files.

[28:25] that are attached to it. [28:27] And so I needed to bridge that gap. [28:30] I have all of the research there. [28:32] I need to make a person out of it, or five. [28:37] Five of them, actually. Okay, so you heard it here first. This is our first Hawaii AI where we are manufacturing not one, not two, not three, not four, but five people. So show us the process. [28:51] Yeah, so... [28:52] I thought about it and then I decided to go to another tool that I really like. This is Notebook LM. [29:00] It's a Google store. And Notebook.lm [29:05] The thing I like about it the most and the reason I picked that tool in order to construct the instructions or the prompt per persona is actually there are some. [29:17] several reasons. But one of them is, Notebook LM allows you to upload your own sources and can answer only based on these sources. [29:28] and not things that it goes and finds online or thinks or knows or filling in the gap. [29:35] This is the information you ask me about something, I will answer based on the knowledge that you [29:41] provided alone. Also, it allows you to check and uncheck the sources that you want to rely on. So I can [29:49] ask a question without relying on the buyer journeys, for example, and then [29:54] the answer will not include that part of the knowledge, things that I cannot do in Chajiquiti or anything else.

[30:02] Um, [30:03] And then there's this check part within the notebook [30:08] where it's Gemini based. Gemini is Google. [30:13] um, [30:14] chat model okay and and in this uh chat window [30:20] I prompted [30:22] the chat, [30:23] Again, you are an expert prompt engineer, again, with the role, what you are, specializing in creating custom GPTs by providing strong AI prompts. [30:35] And then the mission, what your task or mission is. [30:38] Your mission is to create AI prompts for custom GPTs, representing entrepreneurs and small business owners. [30:46] where the decision makers ends on runs out, you will craft highly detailed, nuanced, and authentic chat GPT prompts for five distinct buyer personas based on your sources. [30:59] I never told it what [31:01] were the personas. It had to get it from the sources. And that's the most important part, again, is guidelines. [31:10] I mean, prompts are nice. [31:12] They usually should come with some guidelines, instructions, [31:18] um [31:19] Anything that you want to be [31:22] the chat to take specific care of. So in my case the guidelines were [31:29] to ensure that the prompt correctly and fully described the core identity, mindset, decision-making style. I didn't want the chat to decide...

[31:39] on itself [31:41] what I care for about those personas, because I knew what I care for. So I wanted their mindset and decision-making style and tone and communication style. [31:53] And then the business needs. [31:55] and the technology stack. [31:57] And the journey map, social media preferences, I pointed the chat to exactly what I needed to get out of this. [32:05] Research. [32:07] And then that's another important thing. [32:10] One. [32:11] I think one should not go on without that instruction. Don't add or modify text that is not written or implied in the text. I know you're creating... [32:21] I'm turning you down. The text describes a specific persona must remain true to the original persona. [32:29] Yeah, I'm laughing because yesterday I literally wrote a prompt that was like, do not make up any links. I had a thing that was making up. Do not make any up any links. [32:41] that are not in your source of links like and it's so funny we get so used to operating these chatbots as if they have human reasoning and sometimes they have [32:51] kind of like superhuman reasoning. And sometimes [32:54] They just do stuff that a human would never do, like just make up. [32:58] something. And so [32:59] I think this third prompt, we'll zoom in on the show, is probably applicable to a lot of things, especially when you're trying to constrain [33:08] an LLM space to a specific set of data and inputs. So it's a good prompt. Everybody should use the

[33:17] don't make up stuff prompt your hallucination rate will drop by as yeah yeah exactly wow it it is very [33:26] It is crucial to add those things and to think about them. [33:31] Yeah. Um, [33:33] So the result was, and another thing that Notebook LM is good at doing is you can save the responses of the chat as notes and those notes are saved here for you to [33:48] look up later. [33:49] So I can show you an example of the notes that it created, but mainly I just took the prompts. [33:58] I went over them. [34:00] uh the important thing about the prompts is [34:03] that's another strength of using notebook.lm. It uses [34:10] It uses... [34:13] citations. [34:14] So you can actually go over... [34:17] Um... [34:18] a piece of information that the chat decided this persona is and see [34:23] Where did you take it from? [34:26] And just make sure and verify that it went through all of your... [34:30] data information and didn't invent anything. [34:35] I'm going to laugh because I'm from Austin and I'm pretty sure I know people pleaser, whatever Parker. So it's like a very accurate Austin entrepreneur persona. If folks are wondering if this is creating high quality, high quality persona problems. Yeah, yeah, yeah.

[34:54] So eventually, [34:56] Yeah, I what I did is just I took the prompts. I did need to refine them a little. Okay, because [35:06] Even with all of my instructions, Gemini didn't [35:10] exactly realized what kind of prompt it needs to create. So it was missing some guardrails. It was a little too long. [35:20] The custom GPT instructions are limited to 8000 characters and it created [35:29] Some of the prompts being longer than that. [35:32] So I did need to do a little refine and create stronger prompts. [35:37] So for example, and I'll show you in the demo, I needed to add, I used chatgpk itself or sometimes cloud because I like working with cloud. [35:47] um [35:48] I use them just to tighten the instructions a little and make it more robust. [35:55] And add guardrails. [35:57] So I added, for example, [36:01] You do not act as a general purpose assistant. You do not ask follow-up questions. You avoid slang, bad language, or distasteful content. And keep communication respectful and inspiring. You avoid political, religious, gender, or racial commentary. [36:21] And I really wanted to edit. That's another key point for...

[36:26] creating custom GPTs that need to talk as a person, [36:30] Because believe me, [36:33] Those people work with you. They are your friends. The first thing they will try with a TedGPT-like persona, [36:41] is to tackle them with swear words or their ideas about political things or, I don't know, recipes for food. This maybe should be a default prompt wrapper on all enterprise GPTs, and it would save us all a lot of heartache. [37:11] business needs are... [37:13] technology stack whatever, and then what you get is, then it's the time to actually [37:20] test them. So we created those five. [37:23] And I can go to BalancedBlake, you can see she's one of the most talked to internal chats. So we can go to BalancedBlake and I can ask her, [37:36] What kind... [37:39] of Ed. [37:40] Hebbline, [37:44] with catch [37:46] your [37:48] Attention, maybe I'll move it. [37:51] would catch your attention, [37:53] during, [37:55] a busy.

[37:57] Workday. Don't you? Wouldn't you want to know that thing about your ideal or prospect? [38:05] client. [38:06] And I can send her the question, [38:10] Thank you. [38:11] And then, [38:13] from scanning quickly between meetings or juggling a few... [38:19] would catch my eye, save 10 hours a week with this tool, no tech skills needed. Or from chaos to clarity, one dashboard to run it all. [38:29] Your clients don't need another email, they need this. [38:33] This persona actually explains why [38:37] even why every one of these headlines [38:40] will catch their attention. [38:43] and I can take the same thing [38:46] and tried on a completely different persona like Aiden, [38:51] and Aidan will give me a complete different answer. [38:55] Aiden will say: [38:58] I need. [39:01] One that respects my time and speaks directly to the pain I'm feeling in that moment. [39:07] Still doing admin during edit days, ears are to reclaim 5 hours a week. [39:12] other variations that might grade me. [39:15] win better clients without burning out and so on and so on so each persona actually [39:21] Answers. [39:22] based on the persona that we got from research and that single persona

[39:28] represents thousands of potential [39:32] Customers. [39:33] And so you can try add [39:35] headlines or you can try a product journey what would [39:41] be your best first impression when you get into a new CRM. [39:46] um [39:48] What would be the feature... [39:50] that will convince you not to churn a CRM. [39:55] So you can try it on them. They are 24-7 ready to talk to you on whatever. And I really like them. I mean, personalizing those personas has changed the way we work with them. [40:09] I just love this workflow. And to recap it for folks, you took a bunch of, I'm presuming pretty expensive research, [40:15] that probably sat in a bunch of pdfs and docs where you know we occasionally said head down hayden but otherwise did not use these personas [40:24] You use notebook LLM to create a prompt that embodied the personality of the persona. You put those personas in GBTs, and now you can see that dozens of times your colleagues have gone to them to brainstorm. [40:35] With. [40:36] the persona which i think is really interesting and it's giving me a lot of ideas so many people go to just plain chat gp it's like give me five headlines for an ad campaign as opposed to going and sitting with you know sitting with your fake persona and saying what what what ad campaigns would work on you so [40:55] I love this flow. We learned a little bit about the strengths of notebook LM, GPTs,

[41:01] And flipping these like sort of personas on their head. Let's go to workflow three, which I will tell you, I personally, I have a personal connection to. So people in San Francisco. [41:13] Listen up. Here's the use case for you. Let's jump to your last use case and then we can get you out of here. Well, this one is actually, yeah, a really big pink one. [41:27] Favorite is Seoul. [41:29] Well, imagine getting ready in the morning, driving to work. [41:34] already planning ads for your busy schedule and morning routine, [41:39] only to discover that parking in your favorite parking lot now costs [41:46] $40 an hour instead of the usual $50 for the entire day that you paid so far. So this can ruin your entire day for sure. [41:57] So the thing is, [41:59] Anibook's office is right next to Oracle Park, where the San Francisco Giants play. [42:05] And on game days, especially those taking place in the morning or afternoon, [42:11] Parking rates spike from $50 a day [42:15] to a 40 plus dollar per hour. [42:19] Our team was constantly getting caught off guard, showing up to expensive parking or scrambling to find remote, cheaper alternatives. [42:29] We needed a way to know in advance when to take public questions.

[42:32] public transit instead of driving to work. [42:36] And so the solution was, I was thinking, okay, I think... [42:42] Let's share. [42:43] A calendar, like a joint calendar that just show you on which days [42:48] parking lot prices. [42:51] are likely... [42:52] to search. [42:54] I needed two things for that. I needed to figure out when games are taking place in the ballpark, [43:00] and I needed to create a colander file. I had no idea how to do. Colander file is ICS file. This is the type. I have no idea how to create one. [43:11] Okay, whatever. Let's go to ChatGPT. [43:14] So while you're getting this up, I am just smiling and laughing because my LaunchDarkly office was right behind Oracle Park. And I got there. I found a $20 a day parking. [43:26] And I still have, like, I texted my friend the day I had to pay like $100 to park. I was already down there. [43:34] ready for a meeting. And so San Francisco downtown is we're coming back. People are [43:40] But don't forget that the summer baseball season, and sometimes they have two games a day. They have double headers. [43:47] Yeah, that is correct. And then you have, yeah, walk over there. Just don't use your car. Don't go. Just don't go. If you can avoid it, avoid it. [44:01] Yeah, so I was like, let's try chat GPT. I mean, this should be a simple one. Hopefully.

[44:06] So I tried a naive one. Okay, as you can see, this prompt doesn't tell the chat you are this or that. I was like, I have a simple question. [44:17] Find [44:18] all home games that take place in Oracle Park in San Francisco during the next six months. I use six months because I knew it's the end of the season coming soon. So you can ask for the next year, whatever. [44:32] Filter out only the games that start anywhere between morning to 2:[redacted address] in the evening, when we arrive in the morning, the prices are still the usual sane one. So using these dates, create an ICS file for Google Calendar. That's the calendar that we're using at work. [44:54] that will show these dates as an all-day event I wanted. [44:59] I want to just to see... [45:01] very clearly of potential dates [45:05] days in my week [45:06] where I rather avoid driving to the office. [45:11] And a key point was availability free. [45:15] Otherwise, this all-day event [45:18] will just block my entire day, show me as busy, just because the giants are playing. [45:24] Also, the event description should contain the game details and time. I wanted to add that [45:30] so I can verify that the game is the one that I... [45:35] thinking about that it's actually one that is taking place there. I like to add those extra verification point, validation point, just to make sure

[45:46] that we know what we're talking about. [45:49] And then I also added an instruction other than just... [45:55] output the ICS file that I need, the calendar file, [46:00] I want a textual list of all the dates, times, and events included in the created calendar. [46:07] Now, basically, if it was human, [46:10] they may have been a little, you know... [46:12] offended by me. Why don't you trust me? But Sanctuary doesn't care. So it's thought for [46:20] 36 seconds, provided me with the file and also with a list of all the remaining games, because the season is about to end. All the remaining games. [46:30] that are taking place in oracle park with their [46:33] dates and times. [46:35] And so I know all of these are included in this file. I just took the file. I [46:41] installed it or added it to my personal calendar or work calendar. I also shared it with all of my team members and then you upload it and then you can see, for example, that on September 10th, there is a game. Arizona Diamondbacks are playing the Giants. [47:03] at first pitch is 12:45. So better avoid driving to work that day. [47:12] I love this so much because, again, I have hit this problem.

[47:18] so many times and you don't want a calendar that has the game in the middle of your work [47:24] day right you want to customize you probably could have filed like an uh sf giant schedule calendar that's not exactly what you wanted and it would have had all these games weekend games night games all these things away games as well away games exactly so you can have this really filtered to what you want i'm going to give you one [47:42] one improvement that you can make to this, which is... [47:46] for the night games. [47:48] You should put an alert because if you park in the morning... [47:52] And you're still there for the night game. You should just go to the game. It's a great stadium. That's an excellent suggestion. To watch the games. Good view. It's finally warm in San Francisco. So it won't be freezing. Yeah. And you parked cheap. And you parked cheap. I've done that once or twice where I parked. I'm like, I'm not leaving. Honestly, I don't want to deal with the traffic. I'm just going to go to the game. This is a great little workflow. I think like a very good little helper personal workflow that helped you and also your team. [48:22] So just, again, to recap your use cases, first one we did, oh, agents for recruiting. Loved it. [48:29] So straightforward. I'm going to use that right away. Two, generated persona GPTs. And three, make your daily life a little bit easier by giving you ambient information that can help with your commute. So we're going to wrap our episode with a couple lightning round questions and then get you out of here.

[48:49] The first one I have to say is... [48:51] You are the little helper. You seem to be all over HoneyBook, just helping recruiting, helping the product team, helping the whole team. You know, tell us a little bit about your role and what you think is. [49:06] this role will look like. Do people need a dedicated person or a dedicated team [49:12] towards these automations? What do you think the future of this inside companies is going to look like? [49:16] Okay, for sure. [49:18] I love nothing more than talking about myself. [49:23] So... [49:24] My title is Technical Operations Engineer. [49:28] but it encapsulates a lot of other things. So I do, I research and integrate [49:35] paid tools, but a lot of the times you don't find the exact paid tool that you want, so I build them. I build internal tools and processes. I'm using no-code solutions, automations, and also coded solutions. It can be an AI-powered Slack bot, [49:54] It can be an internal application. It could be integrations between two different applications that don't speak with one another. So I come in the middle and I connect them. [50:06] Um... [50:08] It's not just doing things for others, it's also teaching and enabling others to do for themselves. [50:16] I'm a great believer in enabling. So I do company-wide presentations, I do personal advisory, training classes, documentation.

[50:27] Actually, [50:29] And his own book is a [50:31] is a platform for small businesses. Okay. I see each team and department within Anibook. [50:38] as a small business of its own. [50:42] they provide services, they collaborate with other teams, other businesses, [50:47] They have their own goals, they have their own expertise, passion for different things. [50:53] and they all want to spend less time on manual, thoughtless, repetitive tasks. [50:59] and more time doing what they love. [51:01] So this is where I'm coming. [51:05] for, this is what I'm trying to do. This is what I'm trying to provide. [51:09] To take the... [51:11] to take the friction away and leave you to do what you love. [51:17] One thing I want to call out for folks is I've been in tech a long time and unfortunately, [51:23] Basically up until the last couple of years, I feel like internal tools teams were [51:27] were very starved for resources and occasionally starved for respect. It was like, [51:33] Oh, you got the product teams and their customer facing and they build all the cool products and like internal teams are always underfunded, not enough people, blah, blah, blah, blah, blah. And I think now what I love is this is the moment for internal tools teams to shine, to do legitimate, great, high impact product work. [51:52] I would recommend anybody who really wants to lean into AI find their way into this kind of role. Because honestly, a lot of times it's moving faster than you can even get some of these AI experiences into product, which have a lot of like customer impact and legal implications and blah, blah, blah. But if it's if it's all internal tooling, you can kind of let it rip. And so I just want to like shout out to all the internal tools teams out there that I know today have not got the love and respect that they deserve.

[52:22] This is your moment. You can have really high impact and do some pretty great work. And honestly, [52:26] do a lot to differentiate your career in this moment by taking advantage of the fact that you can build these tools. So I think you are a great model, and I'm excited to see you do it. [52:35] Okay, last question. [52:38] You're a very good prompter. In fact, you create prompters to create prompts. [52:44] when ai is not replying to you the way you want when it is frustrating when the agent [52:50] gets distracted by the notifications in its inbox. What is your prompting [52:55] technique? Are you... [52:58] All caps. D-L. [53:01] Wow, well... [53:04] I... [53:05] I love using all caps and no one can persuade me otherwise. I mean there will be people saying, [53:13] It's just a robot. It doesn't care if it's caps or not. But I'm like, no. It takes me a lot more seriously when it's all caps. [53:22] But... [53:25] But I will say, [53:27] My go-to technique would be to take my current prompt [53:32] and then "Tecl". [53:33] The Chad GPT. [53:35] with my prompt asking it. [53:38] to make it better. [53:39] And how do I do that? It's not just, this is not working. Make it better. Even as a person, I would... [53:48] I have no idea what you want from me. So I'm just going with...

[53:54] This is the prompt I'm using. [53:56] This is what's wrong with the output. Like I outlined, the output is inconsistent, contains too many hallucinations, invent things that are not there. [54:10] That's the second part. And then it's very important for me that the prompt will 1, 2, 3, 4. I list the things that not just what is wrong, but how I want it to be right. [54:27] And then... [54:29] I also add, um... [54:31] I also add it, I give it permission. [54:34] take away everything that doesn't work well. Yeah, you can delete things from my prompt, you can rewrite things that don't work well, and you can add [54:43] things that you feel are will do a better job. [54:47] And I feel like given permission, [54:50] to change, delete, remove, whatever, provides a better output. Because otherwise, and chat GPTs tend to be pleasers, [55:00] they may try to use your prompt and not move a lot out of it. Like, this is yours. This is so great. I'm not going to change it. No, I allow you to change it. I allow you to rewrite it completely. And I tried it several times. [55:18] on several prompts, not just my own. People are coming with me with, "Why does my custom chat GPT act so badly?" They're like, "Let's take your prompt and rewrite it using chat GPT." So I go with that template,

[55:34] And [55:35] And [55:36] Like first try, it's amazing. First try, you get a prompt that is much, much, much better. Usually, it only takes that one iteration. [55:46] for it to work exactly as you wanted it to. [55:49] So that's my team. [55:51] I love it. It's a very professional tip that I will use in my moments where instead I just want to write no in all caps. [55:59] which, you know, I try to pretend that I'm this very... [56:03] you know, patient and sophisticated. [56:06] and AI friendly prompter. But I think the more comfortable I get with it, the more ridiculous my prompts get. It's a good reminder that structure can help. Well, this has been so fun. Where can we find you and how can we be helpful to you? [56:20] You can find me in LinkedIn. [56:23] I am LinkedIn.com slash Michael dot. [56:29] Pellet. [56:30] And I work at HoneyBook, so you can just search for Michael Pellet HoneyBook and find me. I would really love to connect with anyone who is into automation, AI, technology. [56:45] new things, whatever. I want to see what you do. I want to see what you're working on. I get constant ideas from other people. [56:54] in LinkedIn, in Facebook, whatever. [56:58] I'm there. And... [57:01] uh from you um you saw something that i

[57:06] did here and [57:08] It's... [57:09] strike to you with a great idea, a way to improve it. You want to suggest things [57:17] that I can do better. [57:19] or even if you want more information from me, just feel free to reach out, ask questions. I'll be happy to answer. It's one of my favorite things to talk about my work. [57:32] So feel free to do that. [57:34] Well, you heard it. Drop questions in the comments. Connect on LinkedIn. And if HoneyBook is hiring, you now know how they search for great candidates. So if you're a great candidate. Yeah, we are hiring. [57:47] Make sure your profile is well, well structured. Well, well made. You know, maybe maybe use the agent to say, I'm applying for this job. Could you find me on LinkedIn as a good match? And what would I do to improve it? That's the last tip for how I AI professional AI girl right here. Well, it was so nice to have you. Thank you so much for showing us our workflows. They're really inspiring. [58:12] And we will see you soon. Thanks. [58:15] Thank you so much, Kim. [58:25] You can also find this podcast on Apple Podcasts, Spotify, or your favorite podcast app. Please consider leaving us a rating and review, which will help others find the show. You can see all our episodes and learn more about the show at howiaipod.com. See you next time.