You already know outreach works. You also know what it costs you: a whole morning to send ten messages worth reading, a contact database on a monthly invoice, and an outreach platform priced for a sales team of twelve. So it slides down the list again, and the pipeline stays flat.
Taylor Logue, founder and CEO of Taylored Marketing, runs social and demand generation for founder-led businesses. She built her first outreach system on Claude and n8n with an Apollo subscription sitting behind it. Then she rebuilt the whole thing on a single computer-use agent costing thirty dollars a month, and dropped the database entirely. This episode is the rebuild, on screen.
What you get out of it: a daily prospect list that arrives before you start work, messages written against a brand voice you defined once, and a deliberate manual step that keeps your LinkedIn account safe. It is AI for social media pointed at the pipeline instead of the content calendar.
The short version
- A computer-use agent can source LinkedIn prospects on its own, which removes the need for a separate contact database like Apollo or ZoomInfo.
- Running twenty connection requests a day for a month did not exhaust the agent’s supply of matching prospects on a product-brand campaign.
- Having the agent produce a CSV instead of sending automatically protects the LinkedIn account and creates a natural review step.
- Claude Skills holding brand voice and ideal-client definitions stop you re-explaining your business every time you adopt a new tool.
- An agent that does not ask clarifying questions makes a thin brief expensive: a bad list plus a re-run costs roughly half an hour.
- Outreach data gives a social media manager an attribution story that impressions and reactions cannot provide.
Timestamp index
00:00 The prospect database she stopped paying for
01:00 Why most social pros never actually do outreach
04:00 What the off-the-shelf outreach tools get wrong
05:00 Why two good tools beat twenty
06:00 What a computer-use agent does on your screen
07:00 Scheduling the daily run so the list is waiting for you
08:00 Why she makes the agent stop at a CSV instead of sending
09:00 The contact database that dropped out of the stack
11:00 The Claude Skills holding her voice and her ideal client
12:00 Why an agent that never asks questions needs a fuller brief
15:00 Her gut check on whether a DM still reads human
17:00 What changed about why clients hire her
20:00 Where to find Taylor
The subscription that dropped out of the stack
The first build worked. Claude wrote the messages, n8n moved them between steps, and Apollo supplied the contacts. That last piece is the one most people never question, because a prospect database feels like infrastructure rather than a choice.
“When I was using Claude and n8n, I also still had to have my Apollo subscription,” Taylor says. “I was still using Apollo actually as a database, and with Grok Bot, I don’t really know how they’re finding all of these contact information, but they are just finding it somehow, and they’re doing it really, really well.”
The test that convinced her was a product brand reaching out to affiliates, ambassadors and content creators. “We’ve been running this for probably about a month, and we’re doing 20 per day, and they haven’t run out of people somehow,” she says. “I’m kind of waiting for the day where they’re like, ‘Hey, we don’t have anybody left.’” That day has not arrived.
Her read on who this suits is narrow and worth repeating. Solo operators and small teams, yes. Enterprises already committed to another assistant, no.
Why she will not let it press send
The agent is perfectly capable of sending the connection requests itself. She made it stop one step short.
“Basically what I told it is, ‘Hey, can you create an automation where you are sending out 30 messages basically on a daily basis in the morning, and can you put them in a CSV so that I can do it manually?’” The reason is risk, not taste. “I’d rather not risk my LinkedIn account or my client’s to have it done automatically.”
The CSV lands in Google Sheets with six columns: name, title, company, the niche signal explaining why this person surfaced, a suggested connection note, and the profile URL. “All I have to do is hit the connect and copy in basically the message,” she says. The automations are scheduled back to back in the morning, so the list is waiting before the working day starts.
What looks like an unfinished automation is doing two jobs at once. It keeps the account out of platform-risk territory, and it forces a human to look at every message before it reaches a real person.
The two Skills that hold the thinking
The agent sources and drafts. Claude holds the part that is actually hers.
“I built out skills in Claude for say like my voice, like as a business owner or even like my audience of like my ideal client,” she says, “so that Claude knows exactly what I’m looking for with how I wanna be portrayed, like what I represent, and that type of thing, and then also skills for my ideal clients and who I currently work with too, so that it has an understanding of what type of services I offer.”
For client work the same information lives in sheets instead: who the ideal target audience is, who the ideal content creator is, and five named personas to match against. Different container, same principle. The definition is written once and reused, rather than retyped into whatever tool is newest.
This is also her stated reason for keeping the stack small. “You can be using like 20 to 30 tools, and it just gets overwhelming having to move between all these different tools,” she says. “If I can use three or four different AI tools, maybe even like two, then I’d rather do that than be switching between screens constantly, and also having to teach these different platforms about my business, because that is the most time-consuming thing.”
What a brief has to carry when the agent will not ask
Here is the part that surprised her, and the part most worth knowing before you try this.
“One thing I’ve noticed about Grok Bot that is not good, compared to Claude and other AIs, it doesn’t really ask you questions,” she says. “If you prompt it to do something, it’s just gonna do it, and then it will wait for you to re-prompt it.”
The cost of a thin brief is time you cannot get back. “It’s probably gonna work for like 10 minutes to create you this list, and then you’re gonna get it, and you’re gonna be like, ‘All right, now I gotta wait another 10, 15 minutes, 20 minutes for it to go through all of this information and then reproduce a new list for me with new messaging too.’”
So the brief carries everything up front: who you are, who you want, why they qualify, and what the first message should sound like. Prompting is a conversation. Briefing is not.
The check on a message landing in a real inbox
Taylor is blunt about the state of the channel. “I’ll look at my LinkedIn and have like 10 messages, and I look at it basically every day. And most of them are AI, and I mean, these are too.”
Her standard is not whether a message could pass as human-written. It is whether it is doing something a human would do. “At this point, I do think that most people just assume they are at least somewhat AI-generated, but do they still sound like you’re trying to connect on some human level?”
The tell she looks for is evidence the profile was actually read. A good opener references a recent post or a step the business just took, the way a note might congratulate a founder on a launch announced that morning. “That already tells me that they looked at her LinkedIn profile,” she says of one example on screen. “You just wanted to make sure that it sounds like they’re looking at the profile somewhat and have something personal to add into the message.”
What it changed about the client relationship
Taylor started in organic social media management, added ads, then cold email. Outreach is now the centre of the work, and it changed what she is hired for.
“Now people see me as not only somebody who’s just posting on their social media accounts, but I’m also driving sales for their business,” she says. “They’re getting legit leads and traction.”
The second effect is measurement. The agent can check back on what happened to the people it reached: who replied, who filled out a form, who joined the email list. “It helps me show my work and my value and what I’m doing a lot better to my clients,” she says. For anyone who has tried to explain the business value of a strong month of reach, that is the more useful half of the story.
Grab the Episode 13 Blueprint
The Brief Pack: the voice Skill, the ideal-client Skill and the persona sheet to write once, plus the scheduled-run request and the six CSV columns to demand back. Download it here.
Connect with Taylor Logue
Full transcript
Taylor Logue: With this, you do not even have to have an Apollo, ZoomInfo, or whatever database you want to use or have been using to do your outreach, which is really cool. We have been running this for about a month, and we are doing 20 per day, and they have not run out of people somehow. I am kind of waiting for the day when they say they do not have anybody left or need another niche of who we want to reach out to, but that still has not happened. If I can use three or four different AI tools, maybe even just two, then I would rather do that than be switching between screens constantly and having to teach these different platforms about my business, because that is the most time-consuming thing.
Mike Allton: Welcome back to the show. I am Mike Allton. Most social professionals skip outreach. The software that does it is priced for sales teams, but today, Taylor Logue, founder of Tailored Marketing, shows us the Claude outreach system she built herself and then rebuilt cheaper. Hey, Taylor, welcome to the show.
Taylor Logue: Hi. Thank you for having me. I am super excited to go over this workflow that I built out with you.
Mike Allton: Yeah, I selfishly cannot wait to learn how you are doing this because I want to use it for myself. So those of you listening and watching get to learn too. Taylor, I want to start with something I think is universal for folks in our world. We all know we should be reaching out to people, sending messages, and starting conversations, but then we do not, because doing it well for even ten people takes up your whole morning. When did you look at that and decide it was a fixable problem for you?
Taylor Logue: When I first started out, I would just be doing organic social media and maybe also doing some ads. But I figured out that the more lead generation and connection tactics you can have, the more pipeline you are going to build over time, and the more effective your lead generation is going to be. The more things you can layer into it, the better your outreach and pipeline are going to be for you long-term. I have seen that with clients over and over again, anything from product-type businesses building out their ambassadors, influencers, and affiliate programs, to service-based businesses reaching out to people to build connections, build their network, and build their pipeline in turn. Even businesses that have membership-type programs building a corporate side of that and reach out to these businesses as well. I have seen it work in so many different ways that it really works for any type of business.
Mike Allton: If you are not doing outreach, you are missing potential opportunities. But there are also follow-ups, cadences, and sequences you should employ, because typically one touch is not enough to get most people to do anything. Then there are replies and follow-ups that have to happen on a timely basis, multi-threading different accounts, which gets really complicated. For a lot of people, the obvious answer is just go buy something off the shelf that knows a lot of this stuff, or at least says it does. There are plenty of tools out there that say they will do all this for you. You priced one out for a client, went and built your own version instead, and then weeks ago, you rebuilt that one too on something cheaper. So what do these other tools keep getting wrong?
What the Off-the-Shelf Outreach Tools Get Wrong
Taylor Logue: Yeah, so I have looked at a lot of different tools that are out there for LinkedIn outreach, Instagram outreach, and email outreach. There is a ton of stuff out there now, especially with AI. There are tools that do this for you, potentially the same way or better. It is just that a lot of those tools are built for larger teams that have a larger budget. Some of them are affordable, but if you are already paying for an AI platform, the one I am using to do this is $30 a month, and it also does not use a ton of cloud storage, so you are not having to pay extra for anything.
It basically avoids adding that extra cost, and it also helps with simplicity. Tech stacks out there these days can be overwhelming, where even a solo entrepreneur or a team of two to five people might be using twenty to thirty tools, and it gets exhausting having to move between all of them. For me, one of the biggest things that creates ease with AI is simplifying the amount of tools I use. If I can use three or four different AI tools, maybe even just two, I would rather do that than switch between screens constantly and teach different platforms about my business, because that is the most time-consuming thing.
Mike Allton: I could not agree more. I have been loving connecting all these different tools to Claude, just talking to Claude and letting it do a lot of the work. So, Taylor, show us at the very beginning of this. How does someone even end up in a system like this, and what are we looking at once they are in there?
Taylor Logue: This is actually a new tool that I just started using within the past month and a half. It can basically take control of your computer screen and do things for you, and if they need human intervention at a certain point, you can take control of the same screen they are operating out of. It is really unique in that aspect. I think all AI will eventually have some version of this, but they are not super common right now. The main two that I know of are Muse and Grokbot, which is the one I chose to use just because I found it first and now it has all my information, so I am locked in.
Mike Allton: Yeah, you are right. ChatGPT released something comparable, and I know it is just a matter of time, because this is definitely the direction the industry is going. But walk us through what you are doing inside of Grokbot.
Taylor Logue: The cool thing about Grokbot is that it is kind of like Claude; you can schedule things to run at a certain time. Typically, all of the automations I have running in Grokbot are scheduled for the morning, kind of back-to-back, so I can just go in and use them. Basically, what I told it was, “Hey, can you create an automation where you send out thirty messages on a daily basis in the morning, and can you put them in a CSV so that I can review them?”
Why the Agent Stops at a CSV
Taylor Logue: Because typically, with any of the social media platforms, you do not want to do it automatically. There are software options out there, and even Grokbot will do it for you automatically if you ask it to. But I would rather not risk my LinkedIn account or my client’s account by having it done automatically. So I basically just have it create a CSV, and then I will upload it into Google Sheets.
This is basically what it presents me with. It gives me their name, title, and company. It also gives me their niche signal, like why they are recommending that I reach out to this person, and then it gives me a note to send them when I send a connection request.
The coolest thing for me personally is that it literally gives me the URL to go to, so all I have to do is hit connect and copy in the message if I choose to do so.
Another cool aspect of this is that when I was using Claude and n8n, I still had to have my Apollo subscription. I was using Apollo as a database, but with Grokbot, I do not know how they are finding all of this contact information. They are just finding it somehow, and they are doing it really well. For another client of mine, a product-based company, we are reaching out to affiliates, ambassadors, and influencers to create content. We have been running this for about a month, doing 20 per day, and they have not run out of people. I am kind of waiting for the day when they say they do not have anybody left or need another niche of who we want to reach out to, but that still has not happened.
That is probably the most impressive thing to me and the reason I think this is better. An Apollo subscription is not crazy expensive, but it is also not cheap. With this, you do not even have to have Apollo, ZoomInfo, or whatever database you want to use to do your outreach, which is really cool. It is very good for solopreneurs or smaller businesses. I do not see enterprise-level companies switching over to Grokbot because they are already deeply entrenched in Claude or Copilot. Again, I think Claude will have a similar tool soon, so things are just changing fast.
The Claude Skills Behind the Messaging
Mike Allton: Okay, so you are using Grokbot primarily at this point. It is surfacing all these people that it thinks you should connect with. It is giving you a downloadable CSV that you pop into Google Sheets, and you can see their name, their URL for their profile that you can click through, suggested messaging, and so on. Now you are reaching out to dozens of new prospects every day. Help us understand how you are deciding, or how you have made Grokbot decide, what those messages actually say so it is not just making stuff up. Is there a particular prompt there, or how are you handling that part?
Taylor Logue: That is the piece that I am using Claude for a bit. I built out skills in Claude for my voice as a business owner and for my audience of ideal clients, so Claude knows exactly what I am looking for in how I want to be portrayed, what I represent, and the types of services I offer. I have done a similar setup for the client reaching out to ambassadors and influencers, creating sheets defining our ideal target audience, content creators, and different personas.
To make this process simple for Grokbot and to get good results, you have to give it good information. If you enter Grokbot as a new user with no saved context and ask for 20 people to reach out to on LinkedIn with a connection note, it will give you 20 people. But compared to Claude and other AIs, Grokbot does not really ask follow-up questions. If you prompt it, it will just execute, and then you have to re-prompt it if the results are not right. Providing full detail from the start saves time. If you do not give it initial context, it might take ten minutes to build a list, and then you have to wait another fifteen or twenty minutes for it to reprocess everything with updated messaging.
Mike Allton: I think it is partly a maturity situation. Claude and ChatGPT used to be the same way, and they have learned to ask for context. With newer tools like Grokbot, you have to provide that structure up front. You mentioned earlier that you can scroll through your feed and immediately spot when there was no human involved in a post, which makes you want to put your phone down. A direct message is even more personal than a public post. What is your gut check for whether an AI-generated outreach message actually sounds like a real person before you send it?
The Human Check Before Anything Sends
Taylor Logue: I think people are craving genuine human connection so much right now. On Instagram now, they have features calling out AI-generated content, which some people do not care about as much. But for certain businesses, that is a big deal. You do not want your content to come off like you are not putting time into it and are just letting AI generate everything.
With the messages you send out to people, now more than ever, email inboxes and LinkedIn inboxes are flooded. I look at my LinkedIn and have ten messages, and I check it basically every day. Most of them are AI-generated, and I mean, these are too. But the first thing I do is look through the messages and make sure they make sense and do not sound too robotic. At this point, I think most people assume outreach is at least somewhat AI-generated, but the real question is whether it still sounds like you are trying to connect on a human level.
Even this first note says, “Hi Michelle, congrats on taking Bare Beans direct to consumer today and on the Raven seed.” That tells me they looked at her LinkedIn profile. She probably had a post saying they were taking Bare Beans direct to consumer. That is really important when doing outreach: taking into consideration a recent post they made or a step the business has taken.
What Changed About the Client Relationship
Taylor Logue: You just want to make sure that it sounds like they are looking at the profile somewhat and have something personal to add into the message.
Mike Allton: Got it. Let’s talk about what this did for your business. You told me a few years ago you were editing video and managing social accounts, and you told me that outreach and actual sales conversations are now the core of your work. What did building this change about why people hire you?
Taylor Logue: When I first started out, I was just doing organic social media management, and then I got started into doing ads as well, and then I started doing cold email outreach. It is more of the sales side of things, and it works really well for some businesses. But I found that since I was already doing organic and ads, it made more sense to actually reach out to people via social media.
How this has changed my business is now people see me not only as somebody posting on their social media accounts, but also as someone driving sales for their business. They are getting legit leads and traction. Not that social media cannot bring that, but since you are having these conversations with people, it is more in their face and a lot easier to track. Another nice thing about this is that since Grokbot or whichever AI platform you are using has the data of who you have been reaching out to, it can log into LinkedIn and track who responded, who filled out a form, or who signed up for the email list. You can turn that data into various metrics to calculate how you are moving the needle for the businesses you work with.
That is one of the things AI does a really good job of. I will check the work, but there is not a ton of human intervention needed; it can do it basically on its own.
Mike Allton: I think your point is spot on because it is exactly the struggle every single social media professional goes through. You are struggling today to prove the value of your work because you cannot connect a dot between Facebook likes and reactions to actually moving the needle in the business. You need to have something else in mind. That is why years ago at Agorapulse, we invented a tracking algorithm where every link you share is automatically tracked and personalized so you know exactly how much traffic and sales came from a particular post or DM. This is next-level, elevating what you are doing for yourself and your clients. Taylor, this is amazing. For folks who want to know more or get help setting up a system like this, where should they go?
Taylor Logue: They can go to my website at tailoredmarketing.io. I am also on LinkedIn and on Instagram at @taylor_logue (with an underscore at the end). I am pretty active on both LinkedIn and Instagram.
Mike Allton: Fantastic. Thank you, Taylor. As always, we will have all of Taylor’s links down in the show notes below. That is a wrap for today’s lab session.
