I Built a Permanent AI Sales Assistant in 30 Minutes. It Changed How I Prospect.

I am an enterprise AE. New logo hunting role. No existing customers to farm. Every meeting has to be created from scratch.

For the first year I used AI the way everyone else does. Open ChatGPT. Paste in some research. Ask it to write an email. Get something generic. Edit it for 20 minutes. Send it. Close the tab. Start over the next day.

Last quarter I changed how I work with it. I spent 30 minutes setting up a Project inside Claude that knows my voice, my product, my ICP, and my top 3 case studies. Every conversation I have with it now picks up where the last one left off. It writes emails that sound like me. It builds POVs that reference my actual customers. It tells me when I am being delusional about a deal in my pipeline.

I do not paste my company background into a prompt anymore. I just open the project and start working.

This is the part of AI that most sales reps have not figured out yet. It is not about better prompts. It is about a better system.

Claude has a feature called Projects. ChatGPT has the same thing. Both work the same way:

You create a Project. You give it permanent instructions (who you are, what you sell, how you write). You upload files (your product one-pager, case studies, voice samples). Now every conversation inside that Project starts with all of that context already loaded.

Compare the two workflows:

Old way: Open ChatGPT. Paste 3 paragraphs of context. Ask for an email. Realize it does not sound like you. Paste 3 more paragraphs to fix the voice. Get a draft. Edit heavily. Repeat for every email, every account, every day.

New way: Open your Project. Type: “Cold email to VP of Digital at [Company]. Here is the research.” Get a draft in your voice that references your actual product and uses real case studies. Edit lightly. Send.

The difference is not the prompt. The difference is that the AI is no longer a stranger you are explaining yourself to every time.

Five things. Took about 30 minutes total to assemble.

1. System instructions. A one-page brief on who I am, what I sell, who my buyers are, my voice rules, and how I want the AI to work with me. This is the “always on” context.

2. Product one-pager. Plain language description of what I sell. Not marketing copy. The version I would give to a smart friend who asked.

3. Three case studies. Real customers, real problems, real results. One paragraph each. This is the magic ingredient. When the AI writes outreach, it uses these instead of making up generic stats.

4. ICP definition. One page on who I sell to and who I do not. Helps the AI qualify accounts and stop me from chasing bad fits.

5. Voice samples. Five of my best emails, pasted into a document. The AI reads these and learns to write like me. This is the single highest-leverage thing in the entire setup.

This is the prompt that turns “AI that writes okay emails” into “AI that writes emails I would actually send.”

Run this once, after you have uploaded your voice samples to the Project.

I uploaded 3–5 of my real emails to this project. Read them and extract

my voice rules.

Specifically:

1. What sentence structures do I use most often?

2. What phrases do I avoid?

3. How long are my emails typically?

4. What tone do I default to (formal, casual, peer, executive)?

5. How do I open and close messages?

6. What patterns do I use when I am asking for something vs. sharing

something?

Give me the rules back as a checklist I can save to project instructions.

This will help you draft in my voice going forward.

The AI will give you back a list of your voice rules. Things like “uses short sentences, avoids ‘I wanted to reach out,’ opens with a specific observation about the prospect’s business, closes with a clear ask.” Copy that list into your Project’s permanent instructions.

From that moment forward, every email it drafts sounds like you. Not like a vendor. Not like ChatGPT. Like you.

Outreach is the obvious use case. But once the assistant knows your context, it is useful for the entire AE workflow.

Monday morning pipeline review. I paste my top 10 deals into the Project and ask the assistant to tell me which ones are real, which are bullshit, and what I should do this week to move them. It is brutally honest because I told it to be. My manager’s forecast call is easier when I have already done one with my AI first.

Account research. Paste a Perplexity output, get a clean account brief in 90 seconds. The AI synthesizes instead of summarizing.

POV building. Before I reach out to a new account, I have the assistant draft a one-page POV: why change, why now, what is at stake. Forces me to have a real opinion before I write the email.

Pre-call prep. The morning of every discovery call, I paste the prospect’s name and the assistant briefs me. Top 5 discovery questions, top 3 likely objections, what a good outcome looks like. 5 minutes instead of 30.

Post-call debrief. Right after the call, I dump my notes and the assistant tells me what the deal status really is, what the risks are, and drafts the follow-up email.

Reply triage. When a prospect responds with something tricky, I paste their reply and ask for 3 response options ranked by aggression. The assistant tells me which one it would send and why.

None of this is exotic. It is the same workflow I was doing manually. The Project just compresses the time from hours to minutes.

Here is one I will give away because it is the most powerful single prompt in my system. It is the one I run on every new target account before I write a single email.

The POV Builder.

Build a one-page POV for [COMPANY NAME].

Here is what I know about them:

[PASTE PERPLEXITY OUTPUT OR YOUR RESEARCH]

Structure:

1. What is likely true about their current situation

2. Why the status quo is costing them

3. What a different future looks like if they solve this

4. Why now

Use the case studies in this project for real examples. Plain language.

No vendor pitch. Should feel like a peer wrote it, not a sales deck.

If you have a Project set up with case studies, this prompt produces a POV you can use as the spine of every email, every LinkedIn message, every cold call you make on that account. It takes 60 seconds.

If you do not have a Project set up yet, this prompt will still give you something better than 90% of the outreach in your prospect’s inbox.

The voice calibration prompt and the POV builder are 2 of the 10 prompts I run in this Project every week. The full system covers:

Plus the full setup guide: exactly how to build your Project, what to put in the system instructions, what files to upload, how to calibrate voice, and the weekly rhythm that turns the assistant into a habit.

It is $19 on Gumroad. Setup takes 30 minutes the first time and zero minutes every day after that.

If you are an AE in a quota-carrying role and your bottleneck is throughput, this is the fastest $19 you will spend this quarter.

Get the AE Assistant Playbook on Gumroad

The AEs winning right now are not smarter than you. They are working with systems that compound.

A Project takes 30 minutes to set up and gives you leverage every day for the rest of your career at this company. Maybe the rest of your career, period. The knowledge you build into it transfers. The voice rules transfer. The win patterns transfer.

Most reps will read this article, agree with it, and not build the Project. They will keep pasting their context into a fresh tab every Monday morning.

The 10% who actually do it will quietly be running circles around the rest.

That is the whole game.

-Cole