Builder Vision Handoff — stpierre.ai
For Collins, our first Builder · From Akash · 2026-07-13 · Accepted on the Jul 13 call · Start Jul 14
What this page is: the picture of what we're building, and the 7 systems you'll own.
Why it matters: it's your map. The vision is fixed. The "how" is yours to plan and build.
What to do with it: skim the cards. Open a card's detail when you start to build that system.
Time to read: about 6 minutes.
What stpierre.ai is
We're a media company for foundation and waterproofing contractors. We run their ads and get them booked jobs. AI agents do most of the work. We stay small — 3 people, max.
Akash sells and records videos. You build and run the agents that do everything else. You're judged on how well the machine runs, not on tasks finished. Your goal: cut 3+ hours off Akash's week by week 4, and automate 1–2 jobs each week.
The two pipelines
Every job in the company fits one of two pipelines. If a task fits neither, it probably shouldn't exist.
Sales pipeline AKASH OWNS
People and money. Move every lead toward a happy, paying client.
Ads spend, outreach, cold email, calls, closes, keeping clients, and hiring. Delivery lives here too. This is all Akash.
Builder pipeline YOU OWN
Outcomes and numbers. Ship work that moves a metric.
Content, ad production, sales automation, reporting, and the systems around them. You plus your AI agents run this.
How Akash hands you work — read once, then move on
- Akash delegates outcomes, ideally as numbers — big ones. He does not manage the middle.
- The AI floor: anything smaller than a number goes straight to AI, not to you. If you'd just be a middleman between Akash and an agent, Akash tells the agent directly. Your job is the layer AI can't do alone: running the machine, wiring the loops, QA, and judgment.
- Every lane owns one lead/lag pair. The lead metric is volume you control (posts published, ads staged). The lag metric is quality the market decides (conversion rate, cost per result). We build agents that lift the lag metric using volume and testing.
- Lock the format first. Before an agent runs on a format, version one gets locked as a spec (like "what does tweet-style #1 look like"). Then agents iterate inside it. Locking those specs is real builder work.
- Email split: newsletter email auto-sends. Anyone in the sales pipeline never gets auto-sent — you pre-draft it and stage it as one manual action for Akash.
The 7 systems you'll own
1 · Media production BUILDING furthest along
What it is: the content engine. AI writes hooks, scripts, and posts, and publishes with no founder needed.
Your job: ship the full daily slate every day, then automate the running so it needs no hands.
End state, today, and gaps
- End state: AI drafts from performance data and fans every recording into statics, carousels, tweets, IG, newsletter, and YouTube cuts with locked links. Akash only records flagged face/voice items and approves waves.
- Today: organic FB + IG posting is live at one post per day, on its own. Statics system is hardened (last run passed 12/12). 11 carousel templates, 3 B-roll libraries (~150 clips), locked copy contract, 41 approved headlines, 44 VSL videos in flight.
- The daily slate you run: newsletter · organic posts · one AI-UGC (customer-style) video · a carousel · tweet-style and note-style statics. Hooks, offer, and formats are already templatized. Every hook must name the niche (foundation & waterproofing) — hook contract v2.
- Gaps: the daily cadence isn't running yet. Hooks are still hand-picked into the queue (the self-improving loop isn't built). Queue auto-filler is waiting on an approval gate.
2 · Sales system BUILDING blocked on traffic
What it is: AI runs the ads and follow-up that turn strangers into booked calls.
Your job: automate speed-to-lead, follow-up, and no-show recovery. Akash keeps the close.
End state, today, and gaps
- End state: a media-buying agent creates, kills, and scales ads inside caps. A setter agent runs speed-to-lead, follow-up, confirmations, and no-show detection. A sales agent builds pre-call briefs and custom demos. The VSL runs as feed creative, not a standalone page.
- Today: zero paid traffic is flowing — the single biggest constraint in the business. Cold email is warming (3 accounts, ready around Aug 1). The GHL sales system (our customer database) is live. CAPI to Meta is staged, one approval from firing.
- Gaps: outbound-batch sending isn't built yet (10 drafts stalled on small gates). Cold SMS goes only to people who opted in on a form — TCPA law, there was a live incident. The custom-demo engine isn't built yet.
3 · Client onboarding + delivery YOUR FIRST BUILD
What it is: when a client signs, agents build their account and run delivery.
Your job: this is your first build project — the automated client-onboarding system (GHL account setup + an AI sub-agent). Build the machinery now so it's ready when the first client signs. A separate plan spells it out (sp-client-onboarding-automation SOW, 2026-07-14). Delivery-at-scale stays later.
End state, today, and gaps
- End state: client signs → an agent builds the GHL account from a snapshot (settings, bot persona, wiring, checklist). Akash does the kickoff call only. Delivery = 5 qualified inspections per week per client, with a credit ladder auto-applied on a miss. A nightly watchdog scores every bot chat and fixes data errors on its own.
- Today: all the database schema exists (clients, bot checks, health). Onboarding, GHL, and QA skills exist as the "quarry" — parts to reuse. An overnight runner exists in shadow mode.
- Gaps: the onboarding automation is your first build. Delivery-at-scale still waits on a client-count trigger — that part gets built when its trigger fires. Akash writes each client's "qualified" definition once at onboarding, then agents enforce it forever.
4 · Reporting + client experience BUILDING
What it is: daily health scores and client updates, drafted by AI.
Your job: wire the drafts. Sending stays with Akash until it's trusted.
End state, today, and gaps
- End state: a weekly health tier per client, computed daily. Client updates drafted every day and auto-sent once trusted. Complaints auto-routed. One roll-up Review Panel where every department's target, actual, and plan is browsable. The test: Akash goes off-grid 7 days and everything still runs.
- Today: the cockpit, briefing, and capture are live. 74 dashboards registered. Client-health, inbox, complaint, and scorecard skills all exist (built in the EyeFly era, before the pivot).
- Gaps: too many dashboards (99 pages vs. a target of 13). Client-facing dashboards are deferred on purpose. Health should become a daily computed score instead of hand-built.
5 · Data tracking + attribution LIVE
What it is: tracks every lead from ad click to payment. This is our source of truth.
Your job: keep it clean and closed. Turn on the last pieces, like CAPI.
End state, today, and gaps
- End state: person-level tracking as good as Hyros — trace any lead from ad click → VSL watch → application → booking → payment across devices. Every test result feeds a learnings table that agents read before they draft.
- Today: the core is live — 7,352 touches stitched into 3,993 people, with a live timeline dashboard. The brain (our Postgres database) is the single source of truth: 254 tables, 982 change-log rows in 30 days. CAPI is staged, one approval from firing.
- Gaps: email-open tracking is missing. Per-lead organic engagement is missing (we only see totals). CAPI v2 is written but not built. n8n (our automation tool) hit its usage limit — a runaway webhook fired 1,933 times and needs a guard.
6 · Project management BUILDING the core missing loop
What it is: the company runs by scoreboard, not task lists. Targets set Sunday, results checked Friday.
Your job: finish the scoreboard people can see. The engine under it already runs.
End state, today, and gaps
- End state: every department agent sets weekly targets with confidence and reasons (Sunday). Actuals auto-compare (Friday). Learnings get logged. The hit-rate tightens from 100% toward 75% to force stretch. Akash is pulled in only on misses. His ops time drops to 10 minutes a week.
- Today: capture and task tracking are live (527 handoffs, 417 tasks done in 30 days). The Loop Engine is built and running: 8 department lanes with metrics, weekly targets set, the Sunday rite logged, 12 metric keys flowing fresh. What's missing is the scoreboard screen and its history (about 1 week of results so far).
- Gaps: capture is still mostly manual. About 1,050 not-started tasks sit in the backlog. The loop passes when we get two weeks in a row of locked targets, logged results, and an exceptions-only brief.
7 · Builder pipeline BUILDING you are here
What it is: how builders turn ideas into live agents: idea → scoped → built → QA → deployed → documented.
Your job: follow the build order. Automate a job before you hand it to a person.
End state, today, and gaps
- End state: builders own the agent layer — prompts, workflows, dashboards, QA, rollback, and data quality — with a clear pipeline and an agent registry. Rule: a job moves to a human only after its backend runs reliably.
- Today: the board and 6-week plan are live at the builder board. Each builder gets their own login, one shared brain, never a shared identity. Three autonomy tiers are defined (below). 37 EyeFly-era skills are mapped into the 7 St. Pierre agent loops — the "quarry" you graduate one by one. Delete nothing.
- Gaps: approved build order for new tools: fleet-health → recruiting → outbound-batch → n8n registry → session-launcher — all human-gated on any send.
Rules of the road
The 3 autonomy tiers
- Auto — reversible and logged. Just do it.
- Auto-with-gate — do all the work, leave Akash exactly one click. Never a stack of approvals.
- Report-only — read and suggest. No changes.
Every outside change goes through the ledger with a way to undo it.
Locked — do not touch
- The SMS worker (live TCPA compliance case).
- The Meta ad account — budgets are Akash-only.
- The command-center launcher and scorecard tables.
- Anything a client sees. Test only to the phone ending 5477.
Which doc wins
| Doc | What it's for |
|---|
| Operating brief (2026-07-10) | The rulebook. Wins over everything older. |
| This page (2026-07-13) | The vision, and where each system stands today. |
| Daily Operating Rhythm | How each day runs: the 1 PM daily meeting, Slack, scoreboard, the 4 PM report, logoff. |
| Custom Audit + Onboarding SOWs | Active build scope (in the reference shelf below). |
| Builder board | Live. But the FOUND-1 scorecard card is stale — agents build that now, not you. |
| Old Gantt, content-operator packet, phase-1 (Jul 4) | Old. Ignore them where they conflict with this page. |
Where the gated work stands right now — 2026-07-13
Akash reviews each of these with you directly. Nothing below is approved yet. Do not build, render, send, or activate any of them until his feedback lands.
| Item | Where it is today | Status |
|---|
| B-roll look | 3 sample looks are built. The 48-video batch and the AI B-roll pilot both wait on Akash picking a look. | Waiting on Akash |
| Carousel pilots | 3 pilots are staged on the Receipt → Reframe → Rule → Proof → Belief pattern, sitting on the approve/revise page. No scale until all 3 are signed off. | Waiting on Akash |
| Newsletter ("The Backlog Brief") | Wired to its send gate. 4 items only Akash can clear: verify the small email list, fix the old name/cadence in the canon doc, check the domain setup, confirm name + cadence. The first send also needs his explicit go. | Waiting on Akash |
| Week-1 framing | Two versions exist in older docs: daily production slate vs talking-head recording. Akash settles which one is your week-1 spine. | Waiting on Akash |
| Cold email (Instantly) | 3 inboxes are warming. No campaign has been sent. When this lane hands to you is not yet set. | Waiting on Akash |
| Voice-note loop | The plan is: Akash's voice notes become tasks and board cards you pick up each morning. Where the audio lands is not yet confirmed. | Waiting on Akash |
Your focus — set by Akash 2026-07-13
- Focus 1 — Run the daily production machine. Ship the full daily slate every day using the ready hooks, offer, and formats. Then automate the running so it needs no hands.
- Focus 2 — Automate sales. Speed-to-lead, follow-up, no-show recovery, and pre-call assets. Akash keeps the relationship and the close. Live message tests go only to the phone ending 5477.
- Your first build project — automated client onboarding. The GHL account setup plus an AI sub-agent (system 3 above). A separate plan spells it out. Akash walks you through it on the kickoff call.
- Earlier week-1 ideas (custom demo engine, funnel debug) wait unless Akash re-prioritizes them on the kickoff call.
Reference shelf — nothing here is needed in week 1
- Onboarding Automation SOW — at
/builderpipeline/sow/. This is your first build; open it when you start scoping it. The Custom Audit Engine SOW is a later build.
- 6-week Gantt, content-operator packet, phase-1 assignments (Jul 4), projects-pool — legacy, history only. The $1-gate phasing and content-operator framing are retired. Open only to see how we got here.
Where this page came from
Compiled 2026-07-13 from: the AI-run-business vision (07-09), the Autonomous Agency OS summary + SOW (07-11), the delivery-agents blueprint (07-04), the AI-OS audit (07-12), the operating brief (07-10), the master SOW orchestration (07-10), and the Postgres brain (handoffs, signals, tasks).