A Qualification First Outreach System
A daily system that finds 50 possible buyers, runs 60 qualification questions in milliseconds, and only lets 10 to 20 emails leave the building.
Most sales and marketing teams do not have a writing problem.
They have a sending problem.
They take a list, let an AI draft 50 “personalized” emails, hit send, and call it pipeline. Reply rates keep falling. Inboxes get burned. The team spends the next week chasing people who were never going to buy.
I got tired of that inside my own business. So I stopped treating outreach as a writing task and started treating it as a qualification system.
This is what I am building, why it converts better than spray and pray, and why I am looking at it as a product for marketing and sales teams, not just a tool for myself.
The old loop was already better than most teams have. It was still wasteful.
When Grok Bot launched, I moved all of my cold outreach onto it.
I defined one thing first: the perfect client. Not a vague ICP slide. A working definition the system could actually use.
Every day the loop looked like this:
- Grok Bot used Cursor through a CLI to research businesses that matched that profile.
- It produced a daily list of the top 50 fits.
- It wrote an individualized email for each one.
- It sent the emails.
- Everything lived in Craft: sent, needs a reply, needs a follow up, or remove.
That already beat a VA plus a spreadsheet. The emails were specific. The list was researched. Tracking was clean.
The leak was obvious once I watched it run.
I was still paying the cost of 50 first touch emails when only a slice of those 50 should have been emailed at all.
Some had no path to the owner, CEO, or person in charge. Some would never buy enough to matter. Some were one time buyers when I needed recurring work. The writing was good. The targeting was not finished.
Volume was not the bottleneck. Judgment was.
The new loop: 50 names in. 60 questions. 10 to 20 emails out.
Jev is a decision model, not a chatbot. It does not write the email. It answers structured questions fast, in parallel, with a confidence score on each answer.
That is the missing layer.
Instead of taking the 50 and sending 50, the system now runs those 50 through more than 60 questions before a single message is allowed to go out.
The list compresses to the 10 to 20 people who actually fit.
Three of the questions matter more than the rest for sales:
- Do we have a real path to the owner, CEO, or person in charge, or are we about to email someone who cannot buy?
- How much product or service could they actually purchase?
- Is this a one time order or a recurring account?
Those are the questions a good closer asks in the first five minutes of a call. Most outreach tools never ask them. They just write.
Because Jev evaluates the full question set in milliseconds, the filter does not add a research day to the calendar. The daily motion stays intact. The send list gets smaller and sharper.
That is the product.
Not “AI writes your cold email.”
A system that refuses to send the email until the account clears the bar.
What the backend actually looks like
I did not want an agent that auto sends whatever it drafts. That is how you torch a domain.
The architecture I am building for Southern Coastal Printing, and the same pattern I will reuse elsewhere, has a hard rule: only an approved version can leave.
The production path is simple:
- My Cursor CLI (FORGE) stores dated drafts.
- Elon (GrokBot) and Grok generate the payload.
- Jev runs quality control on the email.
- I review the exact copy.
- I approve, edit, or reject.
Only the approved version gets sent, which did I mention will have over 60 quality checks before even arriving to me in less than a few seconds.
Craft is not the writer. Craft is the tracking and duplicate suppression layer, think of it like Notion. It answers the operational questions every sales team actually lives with: what went out, what is waiting, what needs a follow up, and what should be killed so we never email that account again. It’s currently all organized as a Kanban board.
The email itself has a version lifecycle. That sounds like engineering. It is really sales control.
- Version 1: Jev quality control fails. The writer revises.
- Version 2: Jev passes. I edit.
- Version 3: Jev passes. I approve.
The system has to know exactly which version exists. It cannot overwrite the previous one. It has to leave an audit trail. Later, the same version model has to support a day 3 follow up without mixing that follow up into the first touch record.
That last point is where most “AI SDR” tools fall apart. They can draft. They cannot govern.
If you run a team, you already know why this matters. Somebody will ask what was sent, who approved it, and whether the follow up contradicted the first email. If you cannot answer that in 10 seconds, you do not have a system. You have a pile of drafts.
Why marketing and sales leaders should care
Your cost is not the email tool.
Your cost is:
- SDR hours spent on accounts that were never going to close
- Domain reputation burned by volume
- Managers reviewing copy instead of reviewing fit
- CRM notes that do not tell you if the buyer can actually buy
- Follow ups that drift off the original promise
The market is full of tools that make step 3 cheaper. Write faster. Personalize more. Send more.
Almost none of them make step 0 stricter. Should this account get a first touch at all?
That is the gap I am productizing.
A team does not need 50 average emails a day. A team needs 10 to 20 emails a day that were allowed to exist because the account cleared a written standard:
- Right company
- Right person
- Right order size
- Right purchase pattern
- Clean send history
- Human approved copy
Give a sales manager that standard and the conversation changes. You stop arguing about subject lines. You start arguing about the 60 questions. That is a better argument. The questions become the playbook. The playbook becomes the product.
This is not a printing tool. Printing is just the first install.
I am building it first against a real book of business so it has to survive contact with a real inbox.
The design is not locked to one industry. The perfect client definition changes. The 60 questions change. The data sources change. The loop does not.
Same motion for a media company, an agency, a software team, a local service business, or a national sales org:
- Define the buyer.
- Pull the daily 50.
- Score them.
- Kill the ones that fail.
- Write only for the ones that pass.
- Require approval.
- Track the send.
- Follow up from the same record.
I am also pointing the same system at my real estate work. Instead of living inside Zillow, Realtor, and Redfin, the stack pulls the data, scores the property against the deal I actually want, and drops the result into a document. Different market. Same idea. Do not waste a human day on a target that should have been removed in milliseconds.
If it works in two unrelated businesses, it is not a prompt. It is a product pattern.
What it took to get here
I underestimated the backend.
Between Cursor and Grok I have already spent about 12 hours in build conversations, and that is before calling this finished. The writing layer was the easy part. The hard part is the part buyers will pay for:
- A durable definition of the perfect client
- A question set that actually predicts fit
- Versioned copy that cannot silently change
- A quality gate that can fail an email
- A human approval step that cannot be skipped
- Tracking that prevents duplicate sends
- A follow up model that does not corrupt the first touch
That is not a weekend ChatGPT workflow. That is an operating system for outbound.
Who this is for
This is for the person who owns pipeline and is done buying more send volume.
- Agency owners.
- Sales directors.
- Founders who still sit on first touch email.
- Marketing leads who are blamed when the list is bad.
- Anyone paying people to research, write, and follow up by hand, then watching those hours disappear into accounts that were never qualified.
If your current stack can write 1,000 emails a day, you do not need my system for speed.
You need it if you want fewer emails and more conversations with people who can say yes.
If you want this built around your buyer, say so
I am taking this to market.
Not as another generic AI writer. As a qualification first outreach system that can be adapted to a company’s perfect client, question set, offer, and send rules.
If you run marketing or sales and you want to see whether your book of business would compress from “50 sends” to “the 15 that deserve a send,” comment or send me a message.
Tell me your industry and what a perfect client looks like in one sentence. That is enough to start.
Hope you are having a strong week. Now go look at yesterday’s send list and ask how many of those emails should have been blocked before they were written.