Build a Simple Custom AI Agent to Write Emails Faster!

Transcript

Today we’re exploring how AI can change the way that we communicate with others through written communications. G’day. My name is Mike. Welcome to Lone Wolf Unleashed.

So I have a participant in my program. One of the things that we’re working through with her is communication to politicians. And she’s running a regional not for profit, which is around the childcare and education sector.

And I gotta say, I’m right behind her with some of the challenges that are happening at the moment, particularly with working parents and access to childcare and things like that. She needs to get in front of a lot of politicians to be able to get funding so that she can do the work out there.

One of the things that she has to do with the emails is make sure that she communicates in a way that the politician or the politician’s office will respond positively.

And so I had an idea in that the way that we communicate is based on different styles, has been done a lot of research and things like this to understand that. And now we have the ability to search at scale and do research on particular people, particularly with a public profile, and that is politicians.

So what we did is in the working hours this Wednesday, I got to work and started to work through what that would look like to have an agent do that for us. So I’m going to talk now about how that has come together.

Now, when you’re listening to this episode, I want you to not only think about the very specific use case that we have here, but just think about the concepts, about how this comes together.

Because once you crack that, the framework that goes into building an agent or a custom project in your AI tool of choice, that’s where you’re really going to start to get some productivity increases, because it means that you’re going to be able to go away and apply those to how you do things in your own business.

So first up, how do we get started with this? Well, first I kicked off Claude and I had made sure it was on the thinking model, Opus 4.5 at the time of recording. And I basically gave it what we’re trying to do.

We’re trying to communicate to people, to politicians, through email, to have them understand the problem that is being dealt with, what the funding opportunities are, things like that. And that’s at the moment, in a standard template. Start off with systemize in increments.

Okay, first of all, we need a standard template that we can work off. If that doesn’t exist, then we need to create one. So that’s step one, right? Let’s assume that we have a standard template.

Now we want to go and we want to figure out how to tailor this to their comm style. Now I’m also a registered practitioner in DISC and in extended DISC. What DISC does is it covers off four different primary profiles of behavior.

People call it a personality profile. It’s not really a personality, it’s just a behavioral profile.

And so I fed it some information on DISC and basically the workflow of how we would want to interact with this project. We need to have a politician and a name that we want to go after and that we want to communicate with.

We would want to put that into the chat with the LLM and say in this case Anthony Albanese, you know, Prime Minister of Australia. What it will then do is it will then go and search the Internet for content that Anthony Albanese has done about his particular profile.

Now this can be a little bit tricky with politicians because they rehearse and they communicate and they have decided what sort of messaging there’s going to be before certain press conferences and things like that.

But basically what you want to be able to do is in the interviews and things like that, have a research agent or a research component of that agent go out on the net and find interviews and things that this person has done.

And then once that comes back, we then have it benchmark whatever those communications are against a DISC profile so that we can understand what the communication style is like. We tend to communicate with people the way that we like to be communicated with.

So for an example, I am a strong dominant type, the D type. That means that I’m very action oriented, I’m very direct, I’m very task focused. So I don’t typically like to communicate or be communicated with in a flowery way. Right. I like to be very direct.

And so that can influence the way that I consume information because I just want to get to the point. Whereas other styles, like the steadiness style, the S style, likes to have that little bit more time to think through things.

You know, they’re more people oriented rather than task oriented. They’re going to want to have a little bit more flowery stuff in there, right? They’re much more feelings based.

The idea is that the agent will then give us a profile of that against the politician. After that, what we want is for it to ask for the communication that we want to give them.

So that’s where we take our standard template that we’ve drafted and we will feed that to the LLM in this chat and then it will return to us a modified communication based on that template, tailored to the communication style of the person we are addressing.

Now, why is this important? Well, in this case, we want to maximize the chance that someone is going to read this and take action on it, that they’re going to consume it, they’re going to think it’s a good idea and they’re going to respond positively. Right. It could lead to a meeting, it could lead to funding, that sort of thing.

The same could be for people in your industry. It could be leads that you’re looking to reach out to, or prospects that you’ve already sort of canvassed and you sort of have an understanding of what they’re doing.

That could apply as well because you could feed them social content they have done and things to try to understand their communication style and then you can tailor your communication style to that.

That’s how the agents sort of come together and how we’ve sort of built it around, right? Putting in all of that about how we want it to act, how we want it to behave into this initial chat and then instruct it to do some custom instructions for a project.

Okay, so ChatGPT has its custom GPTs, Claude has its Claude projects. You know, those types of things where you can have a project or a GPT exactly for one very specific thing. Right. And that’s what we’ve done in this case.

There are things that I’m looking into soon about how to do this in Claude skills, which is more innate within the Claude product itself. I don’t have to go to a custom project. It will assess each of my inputs against the different skills that I’ve got set up for it.

But that will be an episode for another day. Now we have a runway. We’ve got the workflow, we’ve got our custom instructions that we’re going to give the project, we’ve got our workflow about how we want it to behave, in what order, and then we’ve got an output of a custom, customized, personalized email template to send to that person.

Why does this matter? This is going to increase this participant’s in my program, this person’s productivity immensely because they spend so many hours a week writing emails. Okay. Oftentimes writing emails from scratch each time, it’s a lot of time.

So if you’re able to palm a lot of this research material off to be able to come back with a customized template, essentially that you can copy, paste into an email and then only have to modify the 1, 2, 3 things in our email rather than, you know, all 17 of those things that really does change the game in terms of how many of these politicians or how many of these people you can email a day to try to get what it is that you’re after. In this case, funding.

So have a think about the concepts there. There’s custom instructions. We want it to do a very specific thing.

Okay, the way that I talk to an LLM or the way that I describe things to an LLM is just as I’ve done in this episode today, any supporting documents?

So in this case, I’ve given it some supporting documents around what the DISC assessments are like and the methodology, what the different profiles are, things like that. There’s heaps of information online about that as well. And so it has been able to build something against that.

Give as much detail as possible. It’s okay to sit down and, you know, I use Whisper Flow and I literally just talk to it and it’ll input the text as a transcript into the text box. You don’t have to worry then about typing and then trying to format your text and stuff like that. It doesn’t really matter.

Now I’ve got a full custom instruction here about what this looks like that I can give to a project and they can communicate with basically anyone who they can do some sort of profiling from their online profile, which is really, really cool.

You know, we have the four different styles, what their core drives are, how they communicate, what irritates them, what their fears might be, what their motivations might be, those types of things. All really, really important for understanding how people operate and how they think and how they behave.

So, yeah, cool little use case. I enjoyed putting this one together. I’m going to have a link on my website. It’s going to be lonewolfunleashed.com/my-first-agent. That’s forward slash, my first agent.

And it’s going to walk through basically the components of what I’ve talked about in this episode about breaking it down into a custom instruction and then getting it to assess against that custom instruction what some of the gaps might be and then getting those custom instructions into the workflow of that project so you can start optimizing and streamlining that activity for you.

Thank you so much for listening today. That’s going to do us for today. Thank you so much for your time.

I really appreciate you hanging out with me today. You could have been doing so many other things, but you decided to hang out with me and learn more about how you can make your first custom project, your first agent in your AI tool of choice.

That’s going to do it for this week. Thank you so much. I’ll see you next week.