Unleashed · 2026-08-31 · 46 min
Key moments - from our scoring
Substance score
60 / 100
Five dimensions, 20 points each
Waziri Garuba walks through G.R.I.O.T (inspired by Black Panther's character), a sophisticated personal AI agent that he's built by combining n8n's workflow automation, Claude for the LLM, ElevenLabs for voice, and Vercel for frontend integration. Unlike traditional voice assistants like Siri, G.R.I.O.T functions as a true nervous system that routes queries to specialized sub-agents based on classification - direct answers, tool-based tasks, calendar operations, or complex questions. Garuba demonstrates a fully operational system managing email automation (auto-drafting responses, filtering by contact status), health tracking integrated with Notion databases, calendar event creation, and document retrieval from Google Drive. The architecture relies on sub-agents for Airtable, Monday, Google Drive, Notion, Gmail, and search tools like Tavily and Perplexity. He started small with an email auto-responder and evolved into a multi-platform system where different 'appendages' (agents) handle specific functions. This is particularly valuable for independent professionals, consultants, or operators managing multiple tools and data sources who want AI to intelligently dispatch tasks rather than simply answering questions.
n8n is a workflow automation platform that allows you to connect hundreds of applications and create complex automations. Like Zapier, it provides pre-built integrations for platforms like Gmail, Airtable, and Notion, but n8n is designed more for developers and builders who want to create custom, agentic workflows and has better support for advanced logic and sub-workflows.
Start small with a single automation like an email auto-responder using n8n's Gmail trigger and Claude prompts. Begin by classifying incoming emails and drafting responses, then log results to a spreadsheet to iterate. Once you understand the flow, expand to other platforms and sub-agents.
Both Tavily and Perplexity are AI-driven search tools that provide factual, current information when an agent needs to answer questions like weather or real-time events. G.R.I.O.T's agent can choose either one based on context; Tavily is developer-friendly and affordable.
Each incoming query is first classified as direct, tools-based, schedule-related, or complex. The brain agent then reads the available sub-agent instructions and selects the best match - for example, if you ask about files, it routes to the Google Drive agent; if you ask to create a calendar event, it routes to the calendar agent.
Yes. G.R.I.O.T has access to your databases in Notion, Airtable, and other platforms where you store personal data (health goals, contact preferences, writing style examples). The system uses this stored context to personalize responses and decisions, such as designing age-appropriate workouts based on logged health data.
Our reviewer’s read on each dimension, with quotes from the episode.
The episode provides moderate technical insight into agentic workflow architecture and specific tool integration (N8N, Claude, Telegram, Airtable, Notion), but spends significant time on demonstrations that viewers must see visually, reducing information density for audio listeners. Novel ideas are present (routing logic, sub-agents as specialized appendages) but heavily interspersed with explanatory padding and repetition of the same conceptual frameworks.
When you speak to the platform, that being Grio, whether it be on your phone, on a screen, you're really just sending ones and zeros, right? Because all the data, that's ultimately what it is to a web hook
I like to think about it like a nervous system, right? You hear a message, hey, it's time to eat. And your brain doesn't like take your face and shove it in the food. It sends a message to the part that's best equipped to get the food in your mouth.
The episode presents an interesting application of agents-as-specialized-appendages metaphor and the governance-before-offense approach to automation, which is somewhat contrarian. However, the core concepts (prompt-driven agents, multi-tool orchestration, routing logic) are now standard in AI implementation discourse. The personification of an AI system called Griot is creative but doesn't constitute fundamental new thinking about AI architecture or strategy.
I like to think about it like a nervous system, right? You hear a message, hey, it's time to eat. And your brain doesn't like take your face and shove it in the food. It sends a message to the part that's best equipped to get the food in your mouth.
let's focus on governance. Let's move into what I call defense versus offense, um, understanding how to save time, how to block your time. Instead of jumping into let's, let's bury people in emails and sell a lot.
Waziri Garuba is a legitimate operator with hands-on experience building and deploying AI automation systems for himself and clients. He's a CEO of Harlem Labs and teaches operators/decision-makers, demonstrating practical implementation at scale. However, he is not a marquee figure in AI or a founder of a well-known AI product, limiting seniority perception relative to VC-backed founders or major AI company leaders.
I serve operators. So that's the decision makers who really have got to get the work done. Uh, because we oftentimes are the bottleneck.
I teach and I build or co build. So I don't want to build things for people outright and say here you are because I'm not there to maintain. Uh, I am here to say this is how it works
The episode provides specific tool names (N8N, Claude 4.6, Tavily, Telegram, Airtable, Notion, QuickBooks, ElevenLabs, Vercel) and specific use cases (email auto-response, health tracking with workout scheduling, debt management on $150k student loans at 6.875%). However, most concrete examples are restricted to the speaker's personal workflows rather than client case studies, and quantitative impact is limited (one reference to 2.49x productivity claim with minimal detail on measurement methodology).
6.875% on 150,000 my wife owes, you know, ah, a bit for law school. So we just decided to keep an eye on that number and have every additional dollar go to like just beat it down.
it was logging how many hours I spent on my keyboard, you know, and how many terminals I had running. And it was getting, uh, I was getting about 102 hours of work done.
The host asks clarifying questions and requests screen navigation, demonstrating some follow-up discipline. However, questions tend to be descriptive clarifications rather than challenging or probing deeper into assumptions, limitations, or competitive positioning. The host does not push back on claims (e.g., the 2.49x productivity metric) or explore failure modes, cost-benefit tradeoffs, or alternative approaches. The conversation reads as guided product demonstration rather than rigorous investigation.
What is Tavily?
Why Telegram, by the way, Instead of, Ah, WhatsApp or something else?
Computed from the transcript - who did the talking, and the words that came up most.
Waziri Garuba, creator of G.R.I.O.T and CEO of HarlemLabs, talks about how he was inspired by the limitations of Siri and where the name for his custom AI persona, G.R.I.O.T, comes from. He integrates several platforms including n8n.io, 11 Labs, Vercel, and it can be integrated with an LLM of choice, including Claude. The system operates using an agentic workflow where a central brain assesses inputs via webhooks. Incoming queries are classified into four categories: direct, tools, complex, or schedule-based to determine the appropriate routing. He explains how his agent can talk to the platforms. The First Automation Waziri describes how he built his first automation which was an auto email response maker designed to monitor Gmail every minute. The agent reviews unread messages and drafts responses in his specific writing style while logging actions in a spreadsheet. His current evolved system uses a contact map to identify known contacts and manages inbox organization through classification and labeling. Tracking Health and Fitness Waziri talks about how he uses G.R.I.O.T in his personal life. He calls it G.R.I.O.T in the Garage.
Transcribed and scored by The B2B Podcast Index.
Speaker A: Hello and welcome to Unleashed. I'm your host, Will Bachman, and I'm delighted to be here today with Waziri Garuba, based in Austin, Texas. Woz, welcome to the show.
Speaker B: Thank you very much. Glad to be here, Will.
Speaker A: So, Waz, I know that you're doing a ton of work related to AI. You're, uh, kind of a power user of a lot of different AI tools. You've presented on several cases led sessions for the Umbrex community and Veritux community. And today you've kindly offered to or agreed to show me how you have been using AI agents. And I know that you've built a whole kind of tool on top of innate in that you refer to as Griot. And let's show it to me. And for listeners, we are going to put a video version of this on the Umbrex website. So if you're listening on the podcast platforms, you'll, uh, have the audio version, but in the show notes, we'll include a link where you can go to the permalink for this and see some screenshots as well as we'll have the link to go and see the video, which will be just hosted on the Umbrex website. Okay, Woz, so, uh, tell me a bit about, uh, what, what you've built here and how you're automating a lot of aspects of your, of your life and practice.
Speaker B: Yeah, so I think I'll start. The inspiration for me was, you know, Siri actually, um, utilizing Siri and just kind of witnessing all of the limits with all the things that Siri could not do. So I believe I'm sharing my screen now we are.
Speaker A: Now we are.
Speaker B: Okay, excellent. So what, what happened was, in the process of witnessing that, I thought, how could this be more useful? And of course, I'm a Marvel fan, so Iron Man's Tony Stark came to mind. I dug a bit deeper and actually realized that one of my favorite characters, Black Panther, um, used a character called Grio. And so I kind of built my own Griot, which, as I'm going to show you here, is really a Persona, uh, that stretches across different platforms. Uh, those four platforms are N8N IO ElevenLabs, which gives Griot the personality and the voice and a lot of the things that you encounter. And then there's Vercel, which is where all the things tie together. And then finally there's the LLM of choice. I use Claude. And so what you're looking at, uh, here is essentially Griot's Persona written out and this enables the voice that Griot ultimately kind of expresses. And so there's some instructions here. The most important instruction is anything that you can't do or think through. Route to N8N. So just imagine that the way Siri, you speak to Siri and she doesn't know who you are, you're just a generic person. When you say route to N8N. What happens is, and I think I'll show you when we jump into Grio, this brain is essentially like the main dashboard. And when you hear about agentic and agentic workflow, you're really looking at an agentic workflow here. Whenever you speak to the platform, that being Grio, whether it be on your phone, on a screen, you're really just sending ones and zeros, right? Because all the data, that's ultimately what it is to a web hook, right? Uh, and that web hook is sitting right here. And what it does is it takes that data in and it assesses. Is this a question? So as you could see here, I'd asked when is the next game between the Knicks and the Spurs? And also who's playing in the World cup tonight? So this is a question I'd asked a while back. I have it pinned and what happens is very logically no different than your email that information gets sorted into. Is this a question? Classify it. And classify it in one of four ways. It's direct, it's tools, its schedule is complex and based on its classification, it winds up here, which is in the real first agent. And this agent's entire task, this is GRIO on the back end. It's find the tool that will help us resolve this. So I like to think about it like a nervous system, right? You hear a message, hey, it's time to eat. And your brain doesn't like take your face and shove it in the food. It sends a message to the part that's best equipped to get the food in your mouth. And that's what this agent does. It identifies the best tool and it looks a little hairy, but. But as you can see here, the beauty of agents and agentic is that they're. I won't say it's unlimited, I hate that word. But there's a lot of things one agent can orchestrate, let's put it that way, because what you will see and you'll hear terms like MCP and, you know, APIs and webhooks is I can have this agent talk to any of many platforms. It can talk to airtable, it can talk to notion, it could talk to Monday. So all of the different subscriptions I have and all the different kind of, um, places where my data is kept, I have an agent that is designed to talk to that platform and do whatever it is I need. So I want to pause there because people see this and go, whoa, that looks scary. But it's again, that analogy of like, your brain knows that if you need to walk, it's going to use your feet. And if you need to, you know, um, do something with your hands, you know, you'll need your thumb and your finger. That's what this is really like illustrating.
Speaker A: Okay, so I'm new to N8N. Tell us, just help us. Help me understand what I'm looking at here. And you know, is everything in N8N or are some of these agents like, do they kind of live elsewhere and they're just referenced here? Explain, uh, to me the different systems you have connected to one another.
Speaker B: And absolutely. What I think I'll do is let me show you the first thing that I built and I, I literally remember the first, um, tool I built was an auto email response maker. So I'm going to find it and I think it's right here. Now to really walk you through what Agentic is, right, let's just move up here. If you want something to happen out in the Internet world, you need to open up a port that allows you to see what's happening. And so in this case, let's just kind of build it. I want to keep an eye on my emails, right? I want to know what's going on in my emails. And so I want to, Let me see if I can find a trigger. This is a, uh, this is actually a trigger. So this is a Gmail trigger. What I did first was I want to be able to have emails that come in every minute, get reviewed and go through a process. So anything is the platform where you say, I want to look at what's happening in my Google Sheets or on my Amazon platform. Um, that's what this platform is for. In this case, it's Gmail. But I could also say any of these apps that you're going to see here, right? You could say in Amazon, right? Or in Google, I want to have a sense of what's happening in Google Sheets or Google Drive. This is its superpower. This is what makes Zapier and a lot of these tools powerful is that they're already integrated. And so there's probably tools you use, HubSpot or other where you want something to happen without you Having to turn on and turn off perhaps. And so this allows you to authenticate so that access to that thing is essentially given to you here. And when you start thinking about the fact that there's hundreds of applications out there, it starts to follow a bit of like the Moore's uh, law, like the more things you have access to, the more you can do between them and take yourself out of the mix. So does that make sense?
Speaker A: Okay, yeah.
Speaker B: So in this case I set up the first one that said you have access to my Gmail. Every minute that a message is received in my inbox and is unread, I want you to capture it and I want to look at it. Right. And so that's what this does. This is essentially on and it's active. When you see it in publish mode, it's active and it's watching what's going on. Now when that email comes in, what it's going to do is it's going to run through this agent. And agents are prompt driven. So when we think about people sitting down and writing a prompt and being impressed at the outcome, this is like a 247 prompt. And so it's sitting there and it takes that email and it says review the email and draft response in my name. Here's some examples. Or in my style of writing, here's some examples and here's what I want you to do. I want you to draft a response to the email using these parameters, the subject, who it's from, the message, et cetera. Now if this email is not worth responding to, just say no response necessary. Now this is my very first one, so it wasn't like phenomenal but it worked. I also said create a spreadsheet so I can log exactly, you know, whatever it is response I gave so I could kind of error check and then finally create a draft and put that draft in my inbox. And so what this did was um, I'll show you how it's evolved. But this thing essentially saved me a lot of time. This is the first real time saver because a lot of emails that like I didn't need to pay attention to, I open up and it would just say no response necessary. No response necessary. I delete those. And then some emails that I wanted to respond to, the draft would be done and I could copy and paste it and just plop it in there and send it. And so it started acting as my first kind of introduction to automations. And we understand, okay, the information comes in here. Now my job is, is to modify it And I do that with prompts. I tell it what to do, I tell it, hey, if you get this kind of email, junk it because it has these words. And so what has happened over time has been this evolution to let me see if I can find it here. This is my current email auto response. It's a little bit more complicated but what it does now is it reads contacts. So I have a contact map. If somebody's already approved in my contact map, the email is going to go through otherwise it's going to kind of flag it. And it's just constantly logging. Right. It knows who it should respond to. It shouldn't. It also is a shadow log, but it's regularly helping me manage my inbox. If it's shadow mode, it logs it. If not, as you can see here is the agent. It classifies, it labels, it organizes it for me. So you want to start small with these tools. I started with that simple automation but over the years I've gotten to the point where like my inbox is kind of the easiest part of my day. Right. Things are auto drafted, they're ready. But this platform gives me access to Notion, Gmail, Outlook, all of them. So now they can all play well together in the sandbox.
Speaker A: Mhm. Okay, so these, these sort of connections to your Gmail or to the Google Sheets, those are sort of existing in this innate in environment and you connect them and plug them together and then some of them plug into one of the dozens of tools that you have login for and then they can go and they can do some action on that site.
Speaker B: Yes, exactly, exactly. So you've got this first online layer. I will say that you know Griot, when you look at Grio you can see that I'm talking to him. He's got a brain online. But that's where Claude code really makes it way more powerful. So now we're going to jump into this is not Grio online. This is Grio in my personal life. I actually have a story on my website. Um, this is, these are my field notes. This is just how I like log things that are happening that are interesting in m my life. I call Griot in the garage. I got like, you know, some blood work done and the reports that came in were a bit concerning. So I realized, okay, I need to keep a better track on my health and I also want to set some goals right before I hit 50. I want to, you know, be able to run X amount of miles in this time and so on and so forth. So what I did Was I took all that information and I put it in notion, right? So in my notion, there is a, uh, database that allows me to kind of just check in on my health, right? Like, like my health report. And Grio also has access to that database. So what's good about it is that because he understands quite a bit about me and my health goals, right? And all this information is, like, in my health dashboard. I can say to Griot at the gym, um, hey, Griot, look at my reports and design a workout for me for Saturday morning. And he comes in, understands, you know, realizes where my goals are and literally puts them on my calendar. Like, puts the workout on my calendar for me and then checks in to see if I got it done. So this is what my calendar works out. Like, looks like you can see that, you know, I had a workout on Tuesday, uh, that involved running and stretching. I did not. I skipped my workout on Wednesday. And today I've got to work out. Um, I'll probably have to get to it later, but involves a run, doing some Bulgarian split squats and then push ups, right? Stuff that's age appropriate. But it literally is, again, me being like, oh, I need to move that workout. And because Griot is again, plugged into my calendar, there's this ability for it to take things that I'm doing on a terminal and communicate them, right? Put it on the calendar and send me an email reminder and check in these other areas that you're connected to, such that my health matters are being addressed. And so any one of the databases that you plug into Notion or that are on your hard drive become sources of insight that your agents can use to help you, um, with your kids, right? In one case, my son is now building video games using, uh, Claude, and he's always telling me his ideas. So he built him, um, a game site and his friends are playing the games. All of this with Claude, right? But it's me saying, hey, Grio, how do we make this work? And it's literally there sitting, talking with my son about how to develop his own HTML3 files and then have Claude spin them up on Vercel. He doesn't understand a lot of it, but he knows how to give me the HTML3 file, right? He knows how to make it. So this is where it's making personal life a little bit easier, um, by allowing me to hand off these types of tasks and challenges.
Speaker A: That's pretty cool. Maybe we could go Back to that N8N overview of Griot with the map, and we can kind of Zoom around a bit on the map. Let's start on the very left. So there's some kind of input that sort of starts it on the very left hand side. And is it kicked off by you speaking to it or is it also have things that are just continually running?
Speaker B: Yes, great question. So that's a great one. So this is very, I'd say reactive. Right. Um, so this one is kicked off by me sending a message in and then it does what it does. Right. But let me see if I can find an old one. Uh, I'll see if I can find a kind of simple one. So this one has an execution and you might, uh. It's not showing any of them. So in these cases or with this one that's live. Yes. A message comes in and that message actually gets kind of driven through the process. So let's go ahead and just, we'll run it. Right. So let's, let's actually send a message the way I would if. Imagine that I was on writing this. So this is the query. It would take whatever I said. Let's ask it a question. Um, what is the weather going to be in Austin, Texas today? Right. Uh, tomorrow. Can you put a calendar event on my calendar tomorrow from 12pm to 12:30pm saying run? So I'm asking it two things here and what we're going to do is we're going to walk it down the chain so you can see every single step. So imagine I, um, asked those two questions as I'm walking down the street. What N8N does that's really great is you see it, it takes that information. Now I'm going to run it.
Speaker A: And would you normally input the query by typing in that HTML there or would you speak it into your phone?
Speaker B: I would usually speak it into my phone just because we're here. It's kind of, uh, awkward. So let's see if that goes through. So let's just see. So you can see that that query just came through here. So this, this basically is saying, was that a question or not? And if it is a question, then it's not going to get dumped. And here it's going to classify the question that I just asked. So every time I hit execute, imagine that it kind of went on here. What it's saying is this was a text, so it recognized that I typed that in. And to show you what grio looks like, this is griot. I can also type in a question here and you'll see that end. So I can talk to it or type In a question. So now imagine whatever, whatever I talked it in. It tells me that it's typed and it also tells me this is one that requires tools. Remember what I said about complex? No, there's probably a tool for this. So we're going to move it down the line and say, okay, we know that it's one of these things. So when you execute this step, you're going to see it, it dumped it, moved it down here to asking if it's a schedule. It's not a schedule matter. So we execute again. It's going to move it down here to Tools. And once we execute it here, it's going to move it into the brain. So you see that these measures are again, because I could say, I could cough and it'll be like, is this a query? No. And it dumps it. So it doesn't start working on it. But if I ask a question that it can classify, it knows that sometimes the questions are about what's on my schedule or what am I supposed to do. And it doesn't send that to tools, it sends that to my calendar. But if I'm like, I need you to buy that building, that's a complex question. Uh, you know, I don't know if we're using this tool. Like, what does he mean by this question? So that gets sent to, you know, you've heard of openclaw, that gets handed off to a really smart system that can figure it out on its own. Because I already know what I genuinely generally want. I have all tools aligned to make sense of it. So it's going to send it here now. And as you'll see, this key agent is your job is just to find the right tool, right? What is the tool that can help resolve this issue? So now when I hit execute step, two things happen. One is now it's thinking, it's using Claude, right? Claude 4.6 and saying, what's the best tool for this? And as you can see, it's working through it. It's gone to Tavily, all right? And in Tavle, it's asked the question and Tavlee's like, here, Austin, Texas, there's all this.
Speaker A: What is Tavily?
Speaker B: Tavle is like Perplexity or Google. It's like a search kind of tool that is great for developers because they don't like bop you over the head for every search. It's pretty affordable, but it's like an AI driven search kind of thing. So as you can see, I asked that question and what my agent did was it turned on the LLM engine. I'm going to burn some of your credits, Woz. That's okay. It logged a little bit in memory for M me and then it went out and found the tools. And the two things it did was it used tablet to answer the first question. And then it went to my calendar agent and I'll show you, um, what it did. It went to my calendar agent and said, create a calendar event on Aug. 1 from 12am to 12:30 toddled run. And it sensed it to the agent and the agent said the calendar event titled Run has been created. If you'd like to add attendees or make any changes, let me know. So now I'm going to jump into why. You see, like, remember I said like the brain. Imagine these are all appendages. Now I want to show you the fingers. Right? Because what happened when it sent that, when that agent sent it off to that calendar agent was this. Let me find my M. Yeah, maybe
Speaker A: just back up a step. Quick look at the overall. That, that graph again. That map of all the. Go back to that.
Speaker B: This one.
Speaker A: Yeah, let's. Let's just uh, help me understand the landscape a little bit here. So let's just drives and brains. Explain to me what are those four different things, uh, with the yellow background?
Speaker B: These right here. Okay, great question. So I started very small with like the email agent. And then I realized very quickly I keep a lot of information in airtable. And so I said I want my griot to be able to access airtable, right? And so to do that I had to set up a sub workflow which was I had to map all of the different things in airtable that I wanted it to access, right. And then give it the key to get to those things. So what you're seeing here is I disconnected it for a second, but I'm going to type in airtable. Um, I have a airtable sub agent, right. That I connect it to and then it's able to read all the things within that sub workflow. Similarly, I've got Google Drive. This is a great example. I've got a Google Drive agent. And that Google Drive agent is a credentialed agent that can go and do things in my Google Drive, um, copy, delete, move, whatever it is. So if I say to Grio, can you upload that to my Google Drive? Grio can't do it, but it knows that there is an agent. And that agent basically says, call this tool to find, read or reference files in Google Drive Documents Decks folders or any attachment use when was asks about a file that might be in drive. So that's where the agent instructions are saying, you've got all these tools. Look at what wiseri wants. Read the headers for each of the tools and then select the tool that makes the most sense and send it the task. Right. Hm. So these tend to be places where I'm keeping a lot of valuable information like, hey, my Notion health database or information that relates to my CRM. And those things sometimes relate to elements such as my calendar or email, but they tend to kind of cluster together in. These are the drives I access data in or put files in. But they're not the places that I use to like contact people. Right.
Speaker A: Yeah. Just to read it. There's Airtable Agent, Monday Agent, Drive Agent, I guess that's Google Drive and Notion Agent. Okay, all right.
Speaker B: So you're thinking, and this is where, I mean, like if everybody's going to do it differently, but like again, the. I was pre med, but I'm um. Like, these are my arms. They're very good for doing things. Not so good for holding weight. Right? That's what legs are for. These are so like everything's broken up in the way that I work. And you, for example, see these right here? These are my searches, Perplexity and Tavli. Uh, I'm kind of like, you have both of them. Pick whichever one is best. You can see in this case it picked Tavli. But I'd be fine if you use Perplexity. I don't care. Just answer the question, right? What's the weather going to be tomorrow? Um. And so I give it enough tools to get done, get the job done, right. And when I am giving it the tools back to where I was kind of going. You see these sub workflows right here? I showed you Google Drive. This is the Google Drive. Uh, where is that? That's missing. I'll find one. That is version 2.0. Sorry. So this is the Google Drive. So when it sends that message, that agent sends it here. So what it was pointing at really was to this one. Search for any documents related to the operator's map. So it send the message here and that message would then come to this agent. And this agent says, look through Google Drive, right? Go through the folder, try to match whatever this person's asking for as well as possible and send them back the information. And so what happens is what I said to Griot is getting routed through all those chains coming here and then this thing is going to go in and search and if necessary, download for me whatever it is from Google Drive. So this is not, uh, like now saying, hey, you two fingers, this is what you do, right? Do it and get me back the file. And every sub, every single circle that you see there has one of these. So this is how my email works. That sub agent right here that says email. Hm. Right. It sends the message here. This is actually what's happening. The message is coming to this email agent and it's saying, does Woz want to send an email reply label, create a draft, get email, get labels or mark on red. Right. So all of these are just stacking and what you see is this. Right. But when you're looking at Siri, right, your, your Siri could have this level of complexity. Right. With your information, I want my calendar, my, you know, sales coach, etc. So all of these things are, they are all driven by agents, but like one agent is the one that's picking the tool, sending it in, and then identifying if the answer is appropriate.
Speaker A: Okay, so in terms of the key tools, we have Calendar agent, Task agent, Email agent, Contact agent, and calculator. Okay, so like what, what are, uh. And then under finance tools we have Debt agent, Finance log, Finance query, Finance report agent, uh, QuickBooks, full purse, log, Income Agent, and Purse, Confirm Set Aside agent. Okay. Creative tools, Rober Agent, Content Creator Agent, and Blog post X. Okay, and then you have sales workflows, Sales coach, Network agent, Intro agent, Reference agent, and health.
Speaker B: Okay, that relates, Remember the, the nurturing spoke about? Uh, that's doing that. Okay, there goes.
Speaker A: Right. Okay, maybe we can talk about some of these particularly, like, let's go through the key tools like your calendar agent. What are the types of things the calendar agent is doing for you?
Speaker B: Absolutely. So remember I asked it, I asked Grio, I was like, hey, can you book a run for me tomorrow? Um, let me see if it. I don't know where I put it, but it just, it should have logged the run. I gotta find where. I don't know what date I put it on. But it, it allows me to basically check what's on my calendar. Right. So speak to my calendar, what's on there. And generally like log things like, can you please go ahead and just put a workout for me at this time or move this appointment from this time to the other. Right. On my behalf. And I really find that helpful. When I am working in terminal and you know, I'm working through prospect a project and it says, you need to do this thing. I'm like can you put a reminder for me on Saturday to do it? And by the way, put the prompt in the description so we can just pick up from where we left off. And so by having it MCP in or having a griot in, the things that I need to kind of access on my calendar are kind of always there. And I think more importantly, because it's able to look at my calendar in terminal, it's able to take things directly from any one of the platforms it's connected to and move them entirely. So like I don't have to like actually manually put something. My counter, I could say, can you just add that and put everything that relates to it in the calendar so when I see it on Sunday I know what needs to happen.
Speaker A: Now when you say, uh, you can just ask it, are you typically typing those words or are you just speaking to your computer all the time?
Speaker B: It's a blend. So one of the things that I found is, I don't know if you use, um, is it terminal? It's in. It's like a telegram. Sorry. So with Telegram I've got an instance of Grio on Telegram and I talk to that a lot. That's when I'm usually at the gym or on the road. And so that is probably the most productive just because like I find like when I'm on the treadmill doing like a three mile run, I'll have an idea be like, hey, you know what, put something on the counter for me to follow up with this. Or like all those ideas that happen when I'm like working out are like literally placed, um, or the email is sent or the draft is created because I can very quickly like pick up the telegram app, hit the microphone button and talk to my entire brain. Right. Check notion and tell me if there is this thing I'm supposed to do. And then you know, send Jim an email that I won't make today's 9 o' clock calendar meeting. Is there better available time? Every single one of these things becomes accessible to me while I'm on the go. Um, and that is where for me it's beneficial. Like who haven't I followed up with recently in my nurture engine? Can you send them an email? Let's and let them know what days I have available. Okay. It can figure out on its own.
Speaker A: So you're sending effectively a voice memo as if you were sending it to a person?
Speaker B: Yeah, I'm sending a voice memo to my agent and then the agent is executing uh, for me, whatever it is I Want it to within.
Speaker A: But your agent on the Telegram app is like another contact that you have. Just.
Speaker B: No, it wires into Griot. So it's like I'm talking to Griot.
Speaker A: Uh, okay, but like Grio is a contact on your.
Speaker B: Yes, yes. It's literally like my first. Well, it's. It becomes like a chain. You basically, you set it up in Telegram.
Speaker A: Yeah.
Speaker B: And why.
Speaker A: Why Telegram, by the way, Instead of, Ah, WhatsApp or something else?
Speaker B: Great point. So what I've learned is that, um, platforms are evolving and they're doing their best, but many are not yet really serving in the best interest of the user. So it's really difficult, for example, to set up your WhatsApp account. Just kind of easily integrate. At least. I found it to be similar with Twilio. The most difficult was I showed you QuickBooks. That took me like three to four hours. And for good reason. Right. That's your financials. You should be very like, uh, Even as a developer, I had a difficult time getting wired up because they don't really want you kind of like, kind of having your information out there. And so I just found that Telegram really makes it just like N8N versus Zapier. They really make it easy for you to figure out what you need to do, so you save time. And so I don't want to spend six hours figuring out what's up in order to save two hours kind of texting. Right. I want to spend an hour, figure it out, and then I can get that time back, um, with N8N. And the reason I really love this tool is that you could say, hey, Woz, I want to copy ralph. And I could say, hey, I'm going to download it and give it to you. Right? You can't do that with Zapier. They want you to pay your $49 a month and blunder your way through building it. And I'm like, here's the code. Go and replicate this and then you can start saving time. So the tools that are really, it's like the early Internet days and the, you know, the, like the coffee, uh, clubs or whatever in San Francisco, like the folks who make it easy for us to adopt and use are the ones who are going to get adopted. And so Telegram makes it very simple to set it up so that you are all of a sudden communicating with your LLM and that is getting work done for you. Right? And I think I shared with Megan that, um, one of the really interesting things it had done for me recently was helped me log my Productivity and it gave me a pretty great kind of 2.49x. So it was logging how many hours I spent on my keyboard, you know, and how many terminals I had running. And it was getting, uh, I was getting about 102 hours of work done. So, oh, I need this thing done. Hey, you know, start working on it. Download this, you turn to the next one and by the end of the day you've gotten twice as much done because you had a clean slate of things to do. And then while you're at the gym it says, hey, would you like me to go ahead and do that research? And you hit yes. And you're still working, but you're not. Right. It's, it's a, uh, it's, it's delegating it, uh, to systems. So think that my agent can keep delegating. I think that that's where the productivity we've been promised comes from. And it's really. Now how do we track it? So as you can see here, this actually logs and I spent most of my time on systems and meta work, followed by client engagements. And that systems and meta work starts paying off over time because next Wednesday it's going to do the thing and save me time.
Speaker A: Okay, let's go back to the overall map there. Okay, so we talked about the calendar agent. Uh, tell us about the task agent. Like that next one over to the right.
Speaker B: Oh yeah. So the task agent is, as you can see, sub workflow. And so whenever I say, hey, do this thing for me next week, set this up, etc. The task agent sends that message to a sub workflow that I believe this one's connected to a, it's connected to a Google sheet and. Or. No, it's, it's in my notion. What it does is it creates that notion task inside like this table I have that basically, um, logs all of my, whatever tasks need to be done. And so that way it's like tracking them, letting me know when things are done and I can kind of get rid of them. And so this just winds up being a. Let me see if I can find it in my, in my notion. So, so just like you would send this out to Google, uh, Google sheets. It's all just relational databases, right. So I could have my task tracker and that task. True. Yeah. So it feeds into here, it keeps track of it. If I add something, it's going to ask me the necessary questions. It's tracking the task id. So now when I'm speaking in gre, I'm like What was that thing I was supposed to do that relates to xyz? This is where it's coming. And then it's going to identify or find in this database the thing of relevance. It's going to mark off right when it's done. You could see that I did something and, um, it marked it as done on these dates. But then there are things that are pending. So if I said, what's pending? It'll just tell me. And so, and that's why I say, like, kind of you want to be agnostic on the. Which platform is it Google Sheets? Or in this case, I use Notion, because Notion's a real good central brain. But as you can imagine, you can see that there's a Notion database here. And the task agent is pointing at that or pointing at a specific page in Notion. This will give me access to all of Notion. This right here is going just to that task table. And so I, I'm not dealing with, uh, um, too many different, you know, places of where my stuff is. It's just kind of buried in there.
Speaker A: In terms of the email agent, do you have it set up so that right now it's sending out any emails just automatically without your intervention, or is it always human in the loop for you?
Speaker B: Oh, that's a great question. And yes, it does send me emails without intervention. Right. Um, a lot. Um, and then one of my clients, sometimes I have it do research and just send them information. Now I've had this client for six years, so, like, very comfortable with, you know, the information being sent and, uh, that ongoing relationship. But for the most part, it's mostly drafts. So I'll just look, show you my inbox right now. These are all drafts. Okay. Oh, so things are kind of, oh, wow, replace this or send this to this person, my nurture engine, my sales, everything is just, it's just pending so I don't have to come in and follow up. So, for example, Megan actually introduced me to somebody and so it's drafted. The email for once that goes through. Right. And so I don't need to, like, follow up whenever that goes off. I can come in here and the email is sitting there ready. And so it allows me to also keep my inbox clean. Uh, one of the key things I found is I call it the, uh, guilt pile. I get a lot of really good newsletters, including yours, uh, that I really love. And I'm like, oh, I don't have time. And what it does is it regularly goes in there and it'll clean it up. And say, hey, here's some really great topics. Let me show you what Will has done in the last two weeks. So instead of having to read every single one, I can kind of go into my knowledge base and say, hey, what's Will been writing about? And oh, I like this PDF. This is kind of in line what I wanted. It helps me manage just the overwhelm that comes with like having an email period. Right? Mhm.
Speaker A: Okay, tell us about some of the finance tools that you've got.
Speaker B: So the finance tools are the really interesting. There's a story that I share I started reading the other day, the uh, the Richest man in Babylon. And I love the book. I was like, okay, I want to start adhering to some of this. So I basically said I want to create. So this is actually the blog story, something that helps me work through stage one, right? Start that person fattening. A tenth of everything is going to go into this account first. And so this is one kind of test run where you call, you see the purse log income and purse confirm. I now have a system that's watching my payments and it's like, guaj, you got paid this much, let's confirm and move that money aside. So these are kind of sub workflows that I've given access to. Track funds as they come in and then kind of be annoying and say, hey, let's move that 10% here and you're going to have to cut expenses in other places. But like we're sticking to this, right? And so this is a behavioral kind of agent that I am. I'm tracking the story of like, what if I automated away the task of, you know, you know, I'm getting paid these four or five times a month. Let's put the money in here. And then there is um, a very interesting setup I have here that looks at all of my debt, student loans, credit cards, and regularly gives me a report on like what I'm bleeding out in interest and which payments or debts I should take care of first. So all of these are in the finance tools department because I have access to my um, my finance logs. That's a um, Google Sheet Type QuickBooks thing. So it knows what I'm spending and if I, you know, do any of the things that I am unhappy with myself about, it's like there's a bit of accountability. That's the best way to put it. So this is a savings mechanism. These ones are just basic, they give me reporting. And then this is really looking at how I can, you know, cut Expenses to pay down those really onerous debts faster. And Brio is the one that's sending me all these reports regularly.
Speaker A: Okay. And I guess you're using it for QuickBooks to add expenses or add uh, revenue.
Speaker B: Yeah. Any questions? I have that big ugly QuickBooks, uh, thing that I showed you. It was me wanting to explore it. I had a client bookkeeper and we, we dove down the rabbit hole of how do we get QuickBooks to work for us. But yeah, it's good at, it's good at letting me know where I'm being irresponsible with my money and uh, at the same time giving me alerts like you don't really think about maybe student loans or something. When you own, when you owe six figures, it's like, that's like 600 bucks a month. Yeah, it starts to really add up. So you, you, it's 6.875% on 150,000 my wife owes, you know, ah, a bit for law school. So we just decided to keep an eye on that number and have every additional dollar go to like just beat it down. And that is the mechanism that we've leveraged to do it. Like we're aware, very rare. And then credit cards come in with a score rating. Right. Like this one's bleeding you dry too. Let's get rid of it. And so it's just kind of following it all the way down. But yeah, this quickbook one is a great example of can you access every relevant feature in the tools you use and not have to log in every time you want to get information? Yes, you just have to set it and then I could share this with you, for example, and save you the time. Cool.
Speaker A: And uh, just give us one minute. Was on telling uh, us a little bit about your practice. Are you setting up systems like this for your customers or what kind of clients do you serve and sort of projects do you do?
Speaker B: Yeah, I serve operators. So that's the decision makers who really have got to get the work done. Uh, because we oftentimes are the bottleneck. And so I really service time sensitive folks. And that of course is CEOs to, you know, high level executives. I, I teach and I build or co build. So I don't want to build things for people outright and say here you are because I'm not there to maintain. Uh, I am here to say this is how it works and if you want to own the solution and truly be the person that is able to le AI in your operation as it should be moving forward, then we have engagements of retainers that are monthly or kind of a six week with the operator's map, eight week course that enables people to understand those ones and zeros, see and develop the map that they use to actually move forward into leveraging AI. I think people are too excited about what they can do without mapping what they need to do because as you can see, that's where the value comes from. Knowing I need to pay off debt, knowing that I don't want to log into 10 different tools is why I've done all of this. And so I help people understand what they can do with it and then bring that into reality by spinning up the tools that they manage.
Speaker A: Okay, and I think you have a course coming up that you're offering.
Speaker B: Yes I do. It's uh, the operator's map, uh, that's@the operators map.com. it's really me wanting to dive in. I've served several different businesses successfully but I really want to start working with you know, groups, uh, getting more exposure and really testing the pathway I've discovered which is let's focus on governance. Let's move into what I call defense versus offense, um, understanding how to save time, how to block your time. Instead of jumping into let's, let's bury people in emails and sell a lot. I like to focus on again the time that you can get back by having automations running good parts of your business or helping you with your personal and professional goals. Because I feel that the people who get that now and have ownership of it will be the ones who can successfully turn that time into money in the future when it becomes kind of more clear how AI can create jobs and can be used in sales. Uh, and so the operator's map is eight weeks of digging into and helping people understand how to map out what they need and then how to design the tech stack. As you can see I use zapiermonday.com I use make N8N. I just prefer N8N but I'm familiar with all the tools because I've spent years working with pretty much every single thing that came out. Uh, it was my job for a while to kind of track which tools uh, really mattered. And so this is the outcome of all of that time.
Speaker A: Amazing. Fantastic. Well, where can listeners find you online if they wanted to follow up with your firm?
Speaker B: Yeah, harlemlabs. Ah.com. uh, that's H A R L E M M L A B s dot com. Um, long story behind it but got uh, my MBA in New York and used to work a lot of small business owners in Harlem. And so my little lab has followed me across the country. Um, I also have a. I call my field notes, uh, waziri gariba.com if you just want to see what I'm experimenting and playing with, building like websites with my kids, with AI. I put a lot of my experiences on there. And a lot of my business stuff is you can find@harlemlabs.com Amazing wise.
Speaker A: Thank you so much for joining us today.
Speaker B: Thank you very much. Was a pleasure speaking. Looking forward to speaking more.
Other episodes covering the same guests and topics, from across The B2B Podcast Index.