This is the page you record from. Talk through it in order, show each surface as it comes up, and by
the end the whole picture lands: the agency runs on numbers, the builder pulls work off a board,
and Akash sells. Part 1 is the spoken script. Part 2 is the scoreboard for "are we actually done"
— objective, checkable, no vibes.
Part 1
The Loom Script
Ten scenes, about ten minutes. Each scene names the surface to show and the words to say —
plain and casual, the way you actually talk. Read the italics as stage direction, not out loud.
Scene 1 · Cold open0:00 – 0:45
What this business actually is
On screen: the homepage stpierre.ai, then the offer page start.stpierre.ai.
Hey. Quick walkthrough of how the whole thing runs, so you can operate it without me in the loop.
Here's the business in one line. We sell foundation and waterproofing contractors an AI-run lead
system — we bring them their own local homeowner jobs so they stop buying leads and stop waiting on referrals.
That's it. One offer, one kind of customer. Everything you're about to see exists to deliver that.
Scene 2 · The insight0:45 – 1:45
Ninety-five percent of this is text and pictures
On screen: the ads live dashboard stpierre-ads-live-dashboard.pages.dev — scroll the live ads and strategy cards.
Here's the thing that makes this handoff-able. Almost all the work in this agency is generating text and images.
Ad headlines, captions, creative, landing-page copy, chatbot replies, follow-up messages. That's the job, ninety-five percent of it.
And the tools to make that stuff are already dialed in. The templates exist, the written rules exist, the automatic checks exist.
So you're not inventing machines. You're operating machines that already work. That's the whole mindset — operate, don't reinvent.
Scene 3 · The team1:45 – 2:45
Four functions, two humans, one AI
On screen: the four-functions SLA SOP team.stpierre.ai/sops/four-functions-sla/.
The company does four things: sell, market, deliver, keep clients happy. Sales, marketing, delivery, client experience.
Two humans run it and one AI helps. I'm full-time on sales now — I own the relationships and the number.
You're the builder — you operate the four engines and keep them fed. And Viktor is the AI voice that posts the machine's briefings.
Big rule on how work comes to you: I hand you outcomes as numbers, never a to-do list. Anything smaller than a number, the agents handle straight off. You own moving numbers.
Scene 4 · The spine2:45 – 4:00
The machine wakes up on its own every morning
On screen: the morning status panel (the cockpit briefing), then the dashboard field guide stpierre-dashboard-guide.pages.dev.
Before you touch anything, a whole chain has already run. Watch the clock: Meta ad numbers sync at 6:10, bank at 6:15,
website behavior at 6:30, the scorecard and daily huddle bake at 6:31, the finance and sales pages render at 6:45, the call sheet at 7:30.
Then at 9 a.m. a checker looks at every page, makes sure it's fresh, and drops any red light into the morning status panel. That's your one-glance panel — green means go, red means open the fix-it guide for that page.
You just read this. You never rebuild it. There's one freshness table that tracks all twelve surfaces — what should be up to date and when it last refreshed. Trust that table, not any old notes file.
Scene 5 · Six systems, part one4:00 – 5:30
Sales, chatbot, B2B marketing
On screen: the CRM sales.stpierre.ai, then one of the homeowner chatbots, then the creative rulebook (our reference of ad rules).
There are six systems you run. First three:
One — the B2B sales engine. That's how we get our own contractor customers. Per prospect, it reads the whole relationship and drafts the next message in my voice.
The CRM timeline, the draft-suggestion tool, the sales command center — all built. You keep those drafts fresh; I hit send by hand. This one moves booked calls.
Two — the B2C chatbot engine. That's the product we run for clients. The bot turns a homeowner into a qualified lead for the client and hands it off. It's live — a fleet of homeowner chatbots plus the Dallas Messenger bot we're piloting. This one moves leads for clients.
Three — the B2B marketing copy engine. Every headline, caption, image, landing-page line — informed by the full context and our own test history, tied to a specific hook from our hook list. You test, you deploy, but it goes through my approval gate first. Always. This one moves new ads / day and our cost per lead.
Scene 6 · Six systems, part two5:30 – 7:00
B2C marketing, onboarding, setters
On screen: the client delivery board, then a fresh-client kickoff checklist, then the call-review trainer.
Other three:
Four — B2C marketing copy. Same craft, pointed at client campaigns: video transcripts, captions, ad headlines for the ads we run for them. This moves leads for clients and keeps up client retention.
Five — onboarding automation. This is the one that matters most and it's the least built. Goal: turn a signed sales call into a live launch fast.
Two numbers rule it — time to launch and time to first lead. The bar is a lead delivered on day one, every client. Right now this is a blank slate — build it, and put the numbers on a live page so we can see them.
Six — the setter performance system. Setters are our commission-only cold callers. Track each one's numbers, listen to their calls, auto-prep a daily coaching brief per rep. Also a blank slate — and step zero is turning call recording on in the CRM, because right now it's off. This moves setter shows and booked calls.
Scene 7 · The pull system7:00 – 8:00
Nobody assigns you work — you pull it
On screen: the board team.stpierre.ai/builderpipeline — walk the columns left to right.
Here's how work flows without me handing out tasks. It's a board with five columns.
Backlog is everything, in priority order, and every card spells out what "done" looks like.
Ready is the cards that are fully written up and unblocked — you pull the top card from Ready, you don't pick your favorite.
In progress is where you work — max 2 going at once. Proof is where you post evidence: screenshots, before-and-after readouts, the change log. Done means the number actually moved.
The rule is simple: a card is done when its proof is posted. Anything flagged for my decision can't be pulled — that's a call only I make. And the daily recurring stuff lives in its own lane off to the side, so it doesn't clog the flow.
So the board feeds you. Pull the top card, do it, post the proof, move it. No standup needed.
Scene 8 · The rules8:00 – 8:45
Seven things that never break — and the gate
On screen: the seven-rules section of the builder handoff pack.
Seven rules. Break these and you can cost real money or a client. Memorize them.
1 · Approval gate before activationNothing goes live on Meta until I've seen a contact sheet plus ad and landing-page screenshots.
2 · Never edit a spent ad objectNew creative ships as a new ad set in the same campaign. Never touch one that's spent. Respect the 50-ads cap.
3 · EyeFly is off-limitsDon't touch its systems, don't resurface its invoice, don't mix its history into our numbers.
4 · The live client chatbot is untouchedThe Dallas bot serves a real client. No change to its live line without a tested undo. The final app-review submit to Facebook is mine.
5 · Do-not-contact is sacredAnyone who opted out, said stop, or is in quiet hours is a hard stop. Live tests only ever hit my own test phone.
6 · One daily save of the workspaceThe 6:00 a.m. snapshot is the only thing that saves the shared workspace. Your own build projects use normal code branches.
7 · Log it, with a way to undo itEvery change to an outside system gets a logged record and a clear way to undo it. Money, sends, and signatures still stop for me.
The one that matters most: the approval gate. Everything auto-sends off, drafts only,
until I approve a lane. If it's going to spend, send, sign, or delete — it stops and waits for me.
Scene 9 · What stays mine8:45 – 9:15
What stays Akash's, permanently
On screen: the "what stays yours" panel — money levers, face and voice, relationships, the bar.
A few things never get handed off. Not this year, not ever.
The money leversAd budgets, activation, purchases, payroll. The systems stage it — I flip it.
My face and voiceThe reels, the VSLs, the sales calls. That's the moat. Agents script and cut; I record.
The relationships and the closeBooked qualified appointments, the actual sale, client escalations, hiring.
The bar and final approvalsCashflow, the weekly stretch target, and every go/no-go gate.
Scene 10 · What "done" means9:15 – 10:00
Done is when it runs without me touching it
On screen: the 12-metric scoreboard sp-scoreboard.pages.dev, then back to the board.
So what does "done" look like? Three things, and they're all checkable.
The agency runs on numbers — every metric on this scoreboard has a builder who owns it.
The builder pulls work off the board without me assigning it. And I spend my days selling — calls, scripts, cashflow — not building.
When my calendar is calls and scripts and cash, when you're logging the weekly predictions instead of me,
and when I can be gone a week and every system still runs — that's done. Part two of this page is the exact checklist for it. Let's build.
Part 2
Done when — system by system
One block per system. "Done" is not "it feels handed off" — it's every box below checked, each one
something you could show me on a screen. Numbers, streaks, and proof only. Each block is tagged with the
scoreboard number(s) it moves.
System 1Booked calls
B2B sales engine
This system is done when the sales engine drafts and stages every prospect message
itself, and Akash only ever approves, edits, and sends.
For 5 business days in a row, every active prospect has an owner, a dated next step, and a status by end of day — no blanks in the CRM. (check: CRM shows 0 un-owned or un-dated active contacts)
The draft-suggestion tool writes 2 message options plus a short summary for every prospect waiting on a reply, refreshed daily. (check: a draft ready for 100% of open threads)
A fresh qualified lead has a draft waiting within 5 minutes of qualifying, measured across 3 days in a row. (check: time from qualify to draft is 5 min or less)
Akash writes zero cold messages from scratch for a full week — his only job here is approve, edit, send. (check: the send log shows edits, not new writing)
System 2Leads for clients
B2C chatbot engine
This system is done when the bot turns homeowners into qualified, handed-off leads
with no human writing the qualifying messages, and new prospect bots spin up from a config file alone.
The bot takes a homeowner conversation to a qualified lead and clean handoff for 3 real leads in a row, with no human writing the qualifying messages. (check: 3 clean transcripts, human touches = 0)
The Dallas bot stays up and checking in every 30 minutes with zero misses through the Facebook review period, and nobody edits its live line. (check: check-in log, no gaps)
Bot wording issues — the "real person" line, the unwanted auto-opener — are cleared on every live client bot, and no bot contradicts its own setup. (check: bot-by-bot review passes)
Every new prospect bot ships from one config file plus a deploy — no custom code — checked live on the site. (check: last new bot = one config change)
System 3New ads / dayCost per lead
B2B marketing copy engine
This system is done when a daily batch of ads ships on its own, passes every rule check,
and Akash touches nothing but the approval.
The daily batch of ads ships 7 straight days as a fresh ad set with a preview sheet and screenshots — Akash touching only the approve step. (check: 7-day streak, 0 owner build actions)
Every batch passes the automated creative rule-check before anything goes to Facebook, and the paused preview matches the plan. (check: 0 rule violations shipped)
Every headline, caption, and image ties to a hook on our hook list, matched to its angle and past test results — no stray copy. (check: 0 off-list hooks in the batch)
No edits to any ad set that already spent money, and no ads turned on without approval. (check: Facebook change log is clean)
System 4Leads for clientsClient retention
B2C marketing copy
This system is done when every active client's campaign copy is produced on schedule,
passes the same rule check, and is never mislabeled.
Client campaign copy — transcripts, captions, headlines — is produced on schedule for every active client, with no client waiting on creative. (check: delivery board shows 0 clients waiting on copy)
Every piece of client creative passes the automated rule-check with the right per-trade job value, the "Learn More" button, and the @alexstpierre handle. (check: 0 rule failures)
Every batch is labeled our-own-ads vs client-ads with zero mislabels for a full week. (check: label review = 100% correct)
System 5Time to launchTime to first lead
Onboarding automation
This system is done when a signed client goes live in a day, gets a lead on day one,
and Akash does nothing but the kickoff call — with both numbers showing on a live page.
Time to launch is 1 business day or less (from signed to ads live) for 3 clients in a row. (check: onboarding timestamps)
Time to first lead is 1 day or less — first lead delivered on day one — for 3 clients in a row. (check: first-lead timestamp per client)
A new client's whole setup — their CRM account, template, settings, and kickoff checklist — is created automatically, with Akash doing the kickoff call only. (check: 0 manual setup steps)
Both numbers are built and showing live on a page. (check: they read from real data — today they aren't tracked at all; this system is about 24% built)
System 6Setter showsBooked calls
Setter performance system
This system is done when setter numbers and calls are captured, and every rep gets an
auto-prepared daily coaching brief without Akash putting it together.
Call recording is turned on in the CRM and catching every setter call (the first step — it's off today). (check: recording covers 100% of calls)
Each active setter's numbers — dials, conversations, appointments set, show rate — are tracked daily on a scoreboard. (check: each rep's row updates daily)
A daily coaching brief per rep is auto-built from real pieces of their calls and delivered as a daily touchpoint, 5 straight days. (check: 5-day streak of briefs, per rep)
A new setter passes the full check-out — offer quiz, pitch practice, objection practice, system test, and a supervised trial call — before touching a live list. (check: sign-off record complete)
OverallKeeps the machine running
The whole handoff is done when…
The agency runs on numbers, the builder pulls work, and Akash sells. All five, at once,
held for real — not just a good day.
Akash's calendar is calls, script practice, and cashflow only — no build tasks on it. (check: one week of calendar, 0 build tasks)
Every scoreboard number — booked calls, cost per lead, new ads / day, leads for clients, time to launch, time to first lead, setter shows, client retention, keeps the machine running — has a named builder who owns it. (check: owner filled in, no blanks)
Weekly predictions are logged Monday and actuals plus change notes written Friday by the builder, not Akash, for 2 weeks in a row. (check: who wrote the weekly prediction and change logs)
The builder pulls the top card from Ready, keeps max 2 going at once, and every done card has its proof posted — screenshots, before-and-after readouts, the change log. (check: quick board review)
Akash can be gone for 7 days and every system still sets targets, does the work, and reviews itself — only the exceptions reach him. (check: a real 7-day test)
Internal builder-pipeline document · St. Pierre · prepared 2026-07-23. Operating detail only —
no compensation figures, no credential values, no private financials. This page is for recording and
reference; it is not an outbound message.