Last week we covered how to take a more practical approach to leveraging custom GPT’s and AI agents in your day to day.

The framework for this was centered around:

Persona Matrix
Messaging Matrix
Replicating “links of a chain” vs the ENTIRE workflow chain
Creating your Custom GPT

I’ll briefly summarize how to put together both the persona and messaging matrix.

Most of today we’ll be covering how to create and build custom GPT’s for relevant messaging at scale, something that I’ve done this week.

Persona Matrix

This is the lowest hanging fruit that will provide immediate value once you’re able to implement. I’ve covered this in a previous post but the following prompts will help you to provide a more comprehensive persona matrix which you can reference, followed by a messaging matrix.

As mentioned I’ve done my own prompts before to get to a persona matrix:

Below I’m using Kyle Asay’s approach. I’ve found the prompts easier and malleable for most people. Won’t spill all the sauce since some of the prompts I’ve used today are from his paid product, along with his free LI posts and newsletter. You can check his AE + AI frameworks here (no affiliate links).

Worth noting this post will focus more on how to tailor messaging using custom GPT’s. It’s building upon the persona and messaging matrix frameworks.
That said, I won’t spend a ton of time on the persona and messaging matrixes since it’s been covered extensively by a lot of free resources.

Prompt:

I sell [define product] to companies. These are the outcomes we provide:

1) Help companies [insert value prop 1]

2) Help companies [insert value prop 2]

3) Help companies [insert value prop 3]

(You can add more if you like).

This will provide you with a breakdown of each department and how they will benefit from your product.

Messaging Matrix

Once you have a list of departments then provide the following prompt:

For each of those departments, please summarize below points. Return the data in tabular format, with the below points as column headers. Each department will have its own row.

Here are the points to summarize:

1) What individuals in each department care about as it relates to business outcomes
2) How individuals in each department spend their time to achieve business outcomes
3) What individuals in each department often struggle with to achieve those business outcomes
4) How the solution I sell may be able to help individuals in each department

I took this a step further by adding the following prompts:

Prompt:

Can you also add rows and columns which include titles within each department?
Can you also add personas to the matrix?
Can you also add a column of what potential fears or emotions are related to not achieving their objectives?

Some of this might be overkill, but I figure the more data I can give to my custom GPT, the better the output will be. Ultimately you’ll end up with the header that has the following info:

When you finish this matrix inside of ChatGPT, copy and paste with the following button. You’ll want to put this into a word doc and save it as a PDF so that you can upload into your custom GPT. We’ll get to this shortly.

Replicating “Links of the Chain” vs the ENTIRE Workflow Chain

This is where it starts to get fun. The goal here is to get this VERY good at one thing. Future state is chaining multiple GPT’s that are GREAT at one thing vs trying to shove everything into one. This was a mistake I was making.

For my application, I work on a limited number of accounts so the first thought that popped into my head was, why not focus on one account?

I’m aware that some of you reading this might have a wider net, that’s okay. Couple of things come to mind:

You can modify what I say below by focusing more on verticals or sub segments within a given vertical. For example, healthcare is broad. However if you’re talking about pharmaceutical vs say, medical device, you’ll get much more pointed messaging, personas, and emotions related to your persona and messaging matrix.

Upload that individually into ONE custom GPT. Don’t mix and match.

Additionally, we all have our tier 1 accounts, you can use what I’m working on below to be more pointed and relevant.

Moving forward I can take what works with a given account and then use that data for another company in a similar category.

Creating your Custom GPT

A couple of notes here before I dive into how to create your own GPT:

The goal isn’t to spit out a perfect campaign, I just want some working material that uses my voice and with messaging that resonates and is pointed.
You can plug and play the initial messaging with more company centric material (ie case studies, LI posts, interviews, etc.)
This is a building block to help me refine messaging over time. The first run took me an 30 - 45 minutes. Keep in mind that we’re front loading work. Next time you or I do this it’ll go MUCH faster. Now I have this set and can just copy and paste for other accounts while also training future GPT’s with my style and voice.
Remember, this isn’t a silver bullet. We’re just taking out 80% of the workload off our plates by using GPTs to help with ideation. The real value is HOW we refine it based off those judgements.
After you put your messaging in market, you’ll also have data around what worked and what didn’t which lets you easily upload revised messagings that does or doesn’t work.

In the example above, instead of using verticals or subsets of a vertical, I downloaded a copy of an annual report for a specific account.

To begin, you’ll need the premium version of ChatGPT which is roughly $20/mo. last I checked. Well worth the spend IMO. From there:

Go to the left hand side bar:

That will bring you to the following screen:

Click on “+create”. Which will take you to the configuration screen:

Make sure that you uncheck web browsing, image generation, and code interpreter.

Upload your persona matrix and annual report in .pdf format.

For conversation starters, write the following:

Prompt:

Write me a cadence for xyz team

You can rinse and repeat that up to four times.

Under “Instructions” write the following:

You are a cold outbound assistant helping sales executives at [ your company] to be able to create outbound email cadences with messaging that resonates with the departments, titles, and personas which you have been provided.

You will also incorporate documents that such as annual reports, 10k files, so that you can be relevant to the personas that you are reaching out to.

Create an email cadence with the following prompt:

I got this one from Jed Mahrle, on his recent post showcasing how to use AI effectively for outbound sequences. Giving credit where it’s due.

You’ll get some value using just this even if you decide not to create your own GPT.

Prompt:

Please write a 5-step outbound email sequence for [company name] sales reps to send to potential prospects. Include variables for personalization points.

Make the first email brief, around 50 words, and ask up front if the prospect is facing challenges [company name] can solve.

Include a 1 line case study with a sample company that [company name] helps similar to the prospect company.

Make the second email include a 60-second video demo, and emails 3-4 should address specific outcomes the prospect can achieve with [company name].

Make email 5 a "breakup" email. Here is more information on [company name] to help: [insert company description from website].

Once you get an output, prompt to modify. Please, edit YOURSELF. Don’t just copy and paste.

The first output I’ll admit wasn’t great, so I revised adding my own voice and then re uploaded into the configuration.

Add this line to the “instructions” within the “Configuration” section:

Make sure to reference the email campaign example in order to use my tone when reaching out to the personas and departments that are requested.

Then repeat the email outbound sequence prompt above. This wont give you the perfect output - that’s okay. I still revised the second go around but total time was about 30 - 45 minutes.

Future instances won’t take nearly this long for this account, and I can still use email language and templates that have worked prior for new custom GPT’s. Over time I can refine and get faster at more targeted outreach for each GPT which will give more relevant messaging since I’ll be able to train it faster with better examples and refining prompts. It’ll also be cool to start chaining these together to see the output I get when applied to longer or more complex workflows.

Other things that I didn’t cover which I’m sure will have useful application:

Web browsing - tried this to get more relevant news, will experiment more. Wasn’t too satisfied plus my new found perspective is keeping things small and simple.
AI agents - I can probably add some more workflow to this ie relevant articles, maybe even source leads or relevant personas based of the info I provide. I’d use a no code platform like Relevance AI to build out the code. Still undecided if I want to create a new GPT or put that functionality into one custom GPT. Leaning towards the former.
Adding case studies or success stories (public facing) which I can upload into the custom GPT configuration.

I’ll be exploring these over future posts as I track what successes, or challenges I run into over the course of this first run.

If you have questions, want me to help your own custom GPT, hit me up at andrew@hackingsales.xyz or schedule some time to chat with me.

As always, thanks for reading and see you all next week.

-Andrew K

Hacking SaaS Sales

Tailored GPT Outreach: Building Precision…