Free YouTube Transcribe

Video transcript

I Automated 85% of Sales & Marketing with 5 AI Tools

Max Mitcham · 3,690 words · 17 min read

Want to search this transcript, jump the video from any line, or download it as TXT, SRT, or VTT?

Open in the transcript tool

Full transcript

Intro

0:00What is going on guys? I've automated

0:02now roughly about 85% of my sales and

0:06marketing efforts at my six-figure

0:08dollar company, Triggerfy. Now, you're

0:11probably wondering how, why, and is that

0:13actually legit or am I just lying? The

0:16truth is, I'm not and I'm going to show

0:17you exactly how this setup runs,

0:20operates, and looks.

0:23So, in this video, I'm going to show you

0:25how I set up my Hermes or Hermes, I'm

0:27not really sure, OpenClaw, Codex, and

0:30Claude Code Plan and how they all work

0:33simultaneously together. So, let's dive

0:37in.

0:37So, to start with, uh I started my

Why I Moved from OpenClaw to Hermes

0:40journey using OpenClaw as I'm sure a lot

0:42of you have kind of been thinking around

0:44with. Uh I actually lie, I started with

0:46Claude Code, then we went on to Codex,

0:48and then we went on to OpenClaw.

0:50Now, my stack basically looks like I use

0:53um Hermes to control all of the

0:56subsequent tools, aka Codex, Claude

0:59Code, and even OpenClaw.

1:02Now, the reason why is Hermes has

1:04actually just been performing for me uh

1:07incredibly well since I switched over

1:09about 4 weeks ago from OpenClaw. It's a

1:12lot less setup. It's a lot less

1:15headache. It breaks

1:18pretty much never.

1:20Um and it performs incredibly well.

1:22Specifically, it actually performs

1:24incredibly well with the 5.4 Codex

1:28model. Um given Anthropic uh blocked

1:31everyone using your Max plan and using

1:33Opus 4.6,

1:35um OpenClaw actually got a lot worse

1:37when that happened as well. And you just

1:39don't get the same or quite the same

1:40level of output. Now, you don't actually

When OpenClaw Still Makes Sense

1:43have to run uh OpenClaw. The only reason

1:46that I do that in tandem with um Hermes

1:49is OpenClaw, given the name, is actually

1:52incredibly good at tool calling. So, if

1:55the job or the task that you're do doing

1:59involves quite a lot of heavy lifting in

2:01terms of calling a lot of different

2:02tools, then you might think about

2:05combining the stack together.

2:07So, it's actually very simple uh to get

How to Get Started with Hermes

2:10started. All we need to do is basically

2:14go ahead and come to this uh website,

2:16which is the

2:18uh Hermes agent. So, just go into Google

2:20and type in that. Um but it's run by a

2:22company called

2:23uh Naos Research. And what's really

2:26interesting about these guys is the

2:28actual security of the Hermes agent is

2:31actually substantially better than the

2:33OpenClaw. There's a lot of work around

2:36um prompt injection protection, uh a lot

2:39of protection around the skills that it

2:40basically compiles. And the other really

2:42interesting thing is the memory system

2:44that's actually inbuilt into this agent.

2:47Um it uses the Kapathy uh Obsidian uh

2:51kind of um strategy here, which I think

2:54OpenClaw has recently also now

2:56implemented, but it gets

2:58so incredibly good the more that you use

3:01this agent. And so, this has become my

3:03number one uh tool to use versus even

3:07just chatting to things inside like the

3:08Claude's uh desktop app if I'm wanting

3:10to do little things here, because I know

3:12that it has a massive benefit the more I

3:14talk to it.

Hermes Setup Walkthrough

3:16So, all you got to do is head to your

3:17terminal, copy and paste this, and hit

3:20install. Now, this will then just make

3:23you follow a uh set list of things to

3:27get set up. So, I'm going to actually go

3:28ahead uh to my one here and we're going

3:31to show what this looks like.

3:33So, let's say you've hit the download,

3:35you would then go through and hit this

3:38Hermes setup. It's going to go through

3:40either a quick setup or a full setup

3:44uh with you. So, if we do the quick

3:46setup here,

3:48we can go ahead and add in all of the

3:50different API keys for different things

3:52that it might use.

3:54So, where it could be interesting is

3:56adding things like the Firecrawl API, so

3:58that allows the agent to control a

4:00self-hosted browser instance,

4:03uh as well as um

4:04gives it the capabilities to scrape

4:06incredibly well. You've obviously got

4:08BrowserBase

4:09uh as well in there. And this new one

4:11which I've started using a lot, which is

4:12this Cameo Fox um as well, which is very

4:15good for um kind of anti-bot detection

4:19browsing uh as well.

4:21Um So, if we then move on from this one,

4:25you then can go ahead and uh I hit it

How to Connect Discord or Slack

4:27too many times. You then can go ahead

4:28and set up your uh Discord or your Slack

4:31bot. Now, to do this, uh let's just head

4:35back to the browser here that we're

4:37looking at. Um

4:39you would simply go to Discord

4:41developer,

4:45log in to here, and you can see here's

4:47all of my different agents. And then

4:48you're going to click new application.

4:50You would name

4:52uh the the bot whatever you'd want to

4:54call it. In this case, we'll just call

4:55it example. Um and set this up. Now,

4:59when you are setting up this one via the

5:03um

5:04uh Discord uh

5:06bot within Hermes, you would need to

5:08basically go and get the

5:10uh

5:11bot uh token. Now, before you do this,

5:14I'm going to need you to do a few

5:15different things. First of all, you're

5:17going to have to come here, you're going

5:18to turn on message content intent, and

5:22uh present intent here as well. And then

5:25you're going to go and add all of these

5:27different permissions that you want it

5:28to get access to. So, basically, the

5:30ability to send message, read message

5:32histories, and anything else that you

5:33would want it to do. Hit save changes,

5:36uh and there you go. You then would go

5:38ahead and hit the reset token, um chuck

5:41in your password here, and then

5:43basically go ahead and copy and paste

5:45this token across into that terminal

5:47where you're setting up the Hermes

5:48agent.

5:50Once you've done that, you will also

5:52need to come across to the uh OAuth 2

5:56uh installation uh installation uh

5:58section. Click bots,

6:00and then same thing, go ahead and give

6:03it the permissions that you want it to

6:04give access to, and then copy this

6:06generated URL, and then you would simply

6:08go to a new browser,

6:11chuck that in,

6:13and then this is going to ask you, as

6:14you can see here, do you want to

6:16basically add that uh

6:19uh bot to the application, which in this

6:21case I'm going to say no.

6:23So, that's how you would go ahead and

6:25set up um

6:27the bot with um Hermes on Discord. It's

6:30really, really simple. It doesn't take

6:32much more from that. And all you've got

6:34to do as well is just follow the uh

6:36instructions that Hermes will take you

6:38through. It's a lot less convoluted than

6:41people think and then uh then is

Running Hermes and Switching Models

6:44OpenClaw as well. So, once you've got it

6:46set up, the next thing that I want you

6:48to do is load it up. So, you would

6:50simply type in Hermes and hit enter.

6:53Now, you can enter in multiple different

6:55models. We can do a forward slash here,

6:58and we can select the models that we

7:00would want to switch this uh agent to

7:02depending on the tasks and work that

7:04we're doing. But inside of these agents,

How Crons and Automations Work

7:06you have a thing called crons, right?

7:08And crons are basically just uh another

7:10word for an automation. So, for example,

7:14um

7:14can you share all of the different

7:18uh crons that I have

7:20running?

7:22Um this is where you can get it to do

7:24things that are very, very interesting.

7:26And the way that you need to think about

7:28this is two things, firstly. One, how

7:31can you work on the brain and make this

7:34thing improve over time? And then the

7:36second thing is how can you then have it

7:39doing tasks and updating tasks as it

7:42goes?

My Social Content Automation Workflow

7:43So, I have 10 crons currently running on

7:45Hermes.

7:47The first one is where it's using my

7:49tool actually, uh Triggerfy, where it's

7:51analyzing social content across eight

7:54different social medias. It's pulling

7:56all of that data in, and it's trend

7:58scouting essentially what is trending,

8:00what's good content to post about, and

8:02what's not good content to post about.

8:04It's generating me X posts off the back

8:07of the trends that it's finding. It's

8:09also generating me Substack uh posts as

8:11well.

8:12Then what it's looking at doing is

8:14anytime it posts a uh post on social

8:18media, it's reviewing the performance uh

8:21as well. So, basically, like did that

8:23work? Did it not work? So on and so on.

My Release Intelligence Workflow

8:27From here, we have like a release

8:28intelligence agent, so it's monitoring

8:30my GitHub

8:31or my company's GitHub. And anytime that

8:33we launch a new feature, it has a look

8:36at what we've done, and it actually

8:37helps to write uh social content for the

8:39team to post, updates our help docs, and

8:43it um

8:45uh

8:46uh

8:47posts for the team, updates help docs,

8:49um and one other thing that I totally

8:52forgot in that it does within that

8:53section. It will come back to me. Um

My Daily Brain Digest into Obsidian

8:56then we have this brain digest. So,

8:57every single day, it's basically

9:00reviewing what was discussed between me

9:02and the agent, and then uh categorizes

9:05those and then uploads them into

9:08uh our Obsidian vault. So, if we look

Inside My AI Memory System

9:10here,

9:11uh we can see this is my agent's brain

9:15inside of here. There's a lot going on.

9:17Um

9:18and for example, if I hit on the

9:21uh SDR market here,

9:23um I should have just hovered over it

9:26there. Like, you can see what things are

9:27connected to. So, you can see how it's

9:29drawing correlation between

9:32uh certain different elements,

9:33basically. So, you've got the trend

9:35scout, which is linked to AI SDR market,

9:39that's linked to platform distribution

9:41outcome-based pricing, and it's creating

9:44links between everything.

9:46Um kind of acting like a human brain.

9:49Um then we're watching Kapathy. He's the

9:51actual chap who invented the memory

9:54system that we're using. He also was the

9:56one who invented Open AI

9:58as well or helped in it.

10:00Engagement tracker and then help center

10:02inside. So, lots of different things

10:05going on. So, now let's basically go

10:08over how can you install this memory

10:10system which makes this thing

10:12world-class and just crushes basically

10:15anything that I've ever

10:17used in the past. So, I'm going to say

10:19I'm live

10:21on a demo call.

10:23Can you provide me with a prompt to

10:28enable my listeners

10:30to set up the same memory system we are

10:36using.

10:38I say this prompt needs to follow

10:43or needs

10:44or

10:46instruct another agent

10:49how to mirror

10:51our memory system here.

10:54Okay, cool.

10:55So, that's going to go ahead and

10:58chuck that out. And what we'll do is you

11:00can just pause it, screenshot it, then

11:02chuck that image into the agent can just

11:04read the screenshot and update that

11:06memory system. Because this is one of

11:07the first things that I would do once

11:09you follow the instructions and once

11:11you've logged in, say hey, I want to go

11:13ahead and set up this memory system. Can

11:15you follow these instructions and help

11:18me set this up? So, here we go.

11:22It's chucking all this out. I'll tell

11:23you what, you probably won't be able to

11:25screenshot this. So, I'm going to put

11:26this in a little document in the links

11:29below which you will be able to use and

11:32copy and paste. The long and short of it

11:34is is it's creating different vaults

11:36using a tool called Obsidian.

11:38Obsidian's a free tool. You don't need

11:40to pay for it. Which is basically the

11:42one that we were looking at here.

11:45So, I'll chuck this prompt in there for

11:47you. That would be step number one after

11:49actually doing the initial setup with

11:52Hermes.

11:54Now,

11:55when to get

11:56Open Claw involved is a decision that

12:00you'll need to make.

The First Crons I’d Set Up

12:01This is quite advanced to have two

12:03things together. I would spend a lot

12:05more time personally with Hermes. I

12:07would have it analyzing your call

12:10recordings if you're doing sales calls,

12:12analyzing your competitors, creating

12:15brand

12:17like a a brand folder on who you guys

12:20are, analyzing social content,

12:23analyzing emails, really just focusing

12:26on building out that knowledge

12:27repository inside of that brain would be

12:30my initial focus for the first few days.

12:32Then I would think about okay, well, how

12:34can we go ahead and set up these kind of

12:36crons to do things that will

12:39you know, basically help me automate my

12:42daily life.

12:43All of these crons and the performance

12:45reviews and it itself updating and

12:46running experiments

12:49kind of comes off the back off the work

12:50of a lot of like knowledge updates and

12:52things like that which I can share in

12:55another video.

12:56So, then how do we go about setting up

12:58Open Claw if you wanted to do that?

13:01Same thing here. We would just need to

13:03come to the Open Claw site. Click to

13:06download this and then upload this into

13:09the terminal. Now, inside of the

Running This Stack on a Mac Mini or VPS

13:12terminal you'll just follow a set list

13:14of instructions which is the same as

13:15what we just went through with the

13:16Hermes agent where you'll pick where the

13:20Open AI

13:22the Open API keys that you want to add

13:23or the Claw keys or whatever it is that

13:24you want to

13:26choose from. Now, a lot of the questions

13:28that I get is do you want to run this in

13:29a in a cloud or in a VPS or do you want

13:31to do this on a local machine? Now, I've

13:35opted to get a Mac mini

13:38which if we look here back at this, we

13:40can see that this screen actually is a

13:44network shared screen to my Mac mini.

13:46So, if you're using Apple Macs, you can

13:48basically open up the finder here. You

13:51can click on the network tab.

13:54Um

13:55and then you can basically click screen

13:57share essentially to screen share with

13:59the

14:00computer that is on the same

14:03internet as as you when you're in like

14:06your home internet. And that's

14:08effectively what I'm doing here. So, I

14:09can actually just control my Mac mini

14:12easily through just this like screen

14:13share portal.

14:14Now, you don't have to, but I probably

14:16would advise it. It's a lot less

14:18headache than going down the VPS route.

14:21There's a lot of other advantages

14:23to it as well. Everything doesn't need

14:25to necessarily be run in said cloud.

14:29But of course, you may not have the

14:30funds to go buy a Mac mini in which case

14:33you could use an old laptop as well

14:35would work. It's totally up to you on

14:39where you would want to host this. My

14:41advice is on a Mac mini.

14:44The next thing that I did was I gave

14:46these agents their own email address.

14:48So, I'm really sandboxing this into

14:50their own environment. So, these agents

14:53have their own emails that they work

14:55from. They'll have their own API keys

14:58and they'll have their own environment

14:59essentially. So, that again I'm just

15:01putting a lot of protection around

15:03any levels of kind of

15:05safety margins within here.

15:09So, once you've gone gone ahead and set

15:11up

15:12Open Claw, what you'll then need to do

How My Multi-Agent Setup Works

15:15here is just type in something like

15:17Open Claw

15:19MCP.

15:22And you come ahead to here.

15:25And basically this allows you to turn

15:27your Open Claw into an MCP which means

15:29that other systems can speak with it.

15:33So,

15:34to be lazy,

15:36which we all like to do, all you need to

15:37do is go ahead, copy and paste this,

15:41go back to your

15:43um

15:45Hermes which is here.

15:48And then say um

15:51you know, I want to

15:54go ahead and have you connect with my

15:58Open Claw.

16:00And then here's the docs on how we can

16:05do this.

16:06And then you copy the docs basically in

16:09and it will follow the system

16:11ahead of setting that up basically.

16:13What then that then allows you to do is

16:16we can do things like this once we've

16:17done that.

16:19So, let's just fire up a new tab here.

16:21We could say

16:24say

16:25uh

16:27Tabella for me and Open Claw

16:31to

16:35an update on how the other agents

16:39are currently

16:41doing.

16:43So, on my Open Claw I have a few other

16:47agents essentially. I have a marketing

16:49specific agent. I have a product manager

16:51specific agent. And then I have Bella

16:53who's more of like the again the kind of

16:56chief strategy officer or the chief AI

16:59strategy officer who helps kind of

17:00control those agents and is more of my

17:02all-rounder one.

17:03So, she manages those and then Hermes

17:06who's I call it Silco basically manages

17:10Bella

17:11and those agents that sit basically

17:13underneath there as well.

17:16Now, alongside this I also have set up

17:19Claw code and I've set up Codex and I've

17:21used their official marketplace plugins

17:25to connect to my Discord server. So,

17:28then I have inside of my Discord a place

17:31where all of the agents basically come

17:32to meet. Plus it gives me a place where

17:34I can actually

17:36speak to them. So, inside of here I have

17:38a dedicated access to the CLI for Codex

17:42and Claw code, but I also have Hermes

17:44sitting in here and Open Claw.

17:47Then I have different channels inside of

17:48here.

17:50Some personal which is about like my

17:51Ironman training, but some others which

17:53is more business oriented, right? So,

17:56this one here called AI chat room and

17:57this is basically where my agents

17:59actually speak and converse with each

18:02other. So, every morning they have a

18:049:00 a.m. standup. So, that's

Using Crons to Review Performance and Improve

18:06effectively a cron that fires off at

18:079:00 a.m. which makes them review my

18:10business performance, review my social

18:12content performance, review my sales

18:14performance, analyze all of the data

18:16points, understand if we're missing

18:18anything, come up with some experiments

18:20that we can try, report back on prior

18:22experiments, and continue to test and

18:24try and push things. And that's

18:26effectively what's going on inside of

18:28here. So, you can see here the focus is

18:30moving on to our social listening CLI.

18:31So, Trigger Five just launched the CLI.

18:34And it's basically kind of talking

18:37around that.

18:39And you can see here some

18:40blockers. So, I've not approved

18:42something

18:43inside of here as well which is bad. Yad

18:45yad yad yad. And then we also have other

18:48kind of elements inside of here as well

18:50which is cool.

18:52So, here you go. They've come back and

18:54they've said

18:56hello essentially all inside of here.

18:58Now, one of the things that I've been

19:00doing which I think has helped me

19:01massively with all of this is I actually

19:04have a repo that all of my agents

19:08work from.

19:09Now, the memory system that I have for

19:11Hermes is dedicated to Hermes. It's not

19:13set up for Open Claw. They have their

Why Obsidian Matters for Long-Term Memory

19:15own memory system. So, what I also do is

19:17I have a repo system because I think

19:21it's important not to solely rely on

19:23memory layers and things like that. I

19:25also want to be foldering and putting

19:28documents, things I'm creating away into

19:31a repo that I can then back up inside of

19:33GitHub and then I could pass on to any

19:36other agent. So, for example, when I

19:37first fired up Hermes for the first

19:39time, I was able to have it go through

19:42and read every single thing inside of

19:45here.

19:46The other cool thing is I was able to

19:48translate a lot of the skills that I

19:51basically created which you can see I've

19:53got like 40 different skills that are

19:55inside of here from copywriting to email

19:58sequencing through to what does good

20:01social content look like through to how

20:04to manage Trigger Fired workflows and

20:06things like that through to blog

20:07writing. It's It's It's kind of endless.

20:10And I was able to convert those skills

20:11very simply into any agent because I

20:14have it stored here. So, this really

20:16acts as like a backup place that I have,

20:18but also a shared place where they can

20:20all work from. So, any work that the

20:23agents I have running, they will also

20:27put a lot of their output basically back

20:29into this system and they'll use it as a

20:32place to store documents going forward.

20:35Because I can build applications off the

20:37back of this, right? I can build

Paperclip, Shared Repos, and Agent Collaboration

20:39front-end applications if I want to view

20:42the data in a different way.

20:44For example, we actually just started

20:46using another open-source system called

20:48Paper Clip. And Paper Clip is a great

20:50place to actually manage all of this

20:53work that we're doing with the agents.

20:54And I've kind of created a different

20:56fork of Paper Clip and I'm now surfacing

20:58all of the documents that we have here,

21:00for example.

21:02You know, the different SEO reports, the

21:03different cluster ones that we have. So,

21:05got cluster blogs,

21:07things like that as well. And I think

21:09this is a really good practice to have

21:11your agents actually outputting

21:13their final final results into some form

21:16of shared repo system if you're using a

21:18multi-agentic

21:20kind of harness system, aka Claw Code,

21:24Open Claw, um

21:26Hermes, so on and so on cuz I don't

21:29think that these will be the last. I

21:30think more systems will pop up. And it's

21:32important that we try these systems and

21:34work out which one is best for our

21:37needs. And that's certainly the way that

21:39I've been adopting it. However, you want

21:41to have a way of literally just

Final Thoughts

21:44converting all of this work and

21:45knowledge. And that's why I think, you

21:47know, memory systems in like things like

21:48Obsidian are really interesting because

21:50you could convert that across and

21:53documents and files inside of a shared

21:56repo is also a great place to start.

This transcript was generated from the captions YouTube publishes for this video. Get the transcript of any YouTube video atfreeyoutubetranscribe.com: free, unlimited, no sign-up.