A complete record of the session that designed, built, proved, and rolled out the automated marketing department: the product, the pricing case, three live apps + manuals, a 21-site rollout, a network newsletter, a rebuilt sports league, and a 60-site marketing diary.
Overview
The goal: package "automate a whole marketing department" into something a real client (a prospective client) can see, price, and buy — using Polymagnet (programmable/coded magnets) as the demo client. Over the session this grew from a pitch into a working product: three design-distinct apps, a defensible price backed by 22 real competitors, comprehensive manuals, a self-serve funnel, and a network-wide rollout with measurable diaries.
Working price$2,900/mo, Managed tier (a working number, pending client sign-off + the competitor read). Client brand white-label to the client's own domain. Rule AI-run, human-approved — nothing outbound ships without a person.
Phase 1 — product, research & pitch
Align a season, name a league, gamify it → The Hustle League.
Built texasfridaynights.com/hustle/ — 8 sites competing as teams, WNBA-named coaches, real hits+revenue scoring, a 3-announcer booth, per-team pages, a questions thread.
Client meeting logistics — scheduler, calendar, emails (never send).
Built wholetech.com/meet (self-hosted booking), drafted (never sent) the client emails, ran the meeting.
Ingest the meeting notes; build the deliverables.
Built /compare, /zoho sketch, /onboard (questionnaire → instant live preview), a per-client knowledge overlay, and an application-note library.
Thoroughly research Polymagnet; 3 revenue-optimized site revisions; dazzle them.
4-agent research → /polymagnet brief; 3 "dazzle" revisions on polymag.deptless/amdept/automarketing; registered multipolemag.com + multipolemagnets.com as standalone sites (AF-100, in alldomains, self-publishing /journal/).
Spec + handoff for a production vendor; answer their 20 questions; roadblock email.
Google Doc spec, 20 answers, offer/price/brand options, and a clickable /map page.
Phase 2 — the build
Is $2,900/mo reasonable? Prove it by looking at the competition.
Researched 22 real services → deptless.com/competitors/ + /competitors/worksheet/. Verdict: reasonable, arguably underpriced — below every human alternative (agency $3.5–5k, fractional CMO $5–20k), the exact price one writing-only service charges. Added an interactive cost calculator + "last verified" freshness.
Checkout-page copy for the vendor (don't hallucinate the price).
Delivered a .md spec with the price left as a [PRICE] variable so nothing gets hardcoded before the client decides.
10 prompts to make deptless.com better & more popular — execute in order.
Shipped all 10: og:image + Twitter cards + favicon, Agents-First complete, footer nav, homepage authority magnet, a guide hub (6 articles + Article/Breadcrumb schema + related-reading), FAQ (16 Q&A + FAQPage schema), hero/CTA, a "see it live" demo module, two sitemaps + IndexNow, and a lead-capture "playbook."
The Hustle standings don't move — update them (really make it work).
Found the standings were hardcoded 0-0; wired real W-L from each completed game-day's actual traffic. Later, a full rebuild fixed a team-index bug that had 404'd 6/8 game links + mis-scored standings, froze results against log rotation, and shipped the premium Opening Week Classic event (texasfridaynights.com/hustle/event/). 191 hustle pages, all 200.
The three apps + manuals
"Comprehensive app that does automated marketing, worth $2,900" → three design-distinct SaaS dashboards, then a fourth-and-fifth ask to build them all with the model and add manuals. Each has 8 working views (Dashboard, the 7 roles, Calendar, Approvals, Analytics with hand-drawn SVG charts, Library, Connections, Settings), seeded with sample Polymagnet content, fully interactive, persisted to localStorage.
Site
Concept
App
Manual
deptless.com
Blueprint Console — a schematic control panel; 7 cyan subsystems + an amber Human Gate
/marketingapp/
/marketingapp/manual/
automarketingdept.com
The Team You Hired — 7 named staff, an org chart, a $29,600/mo payroll-savings counter
/marketingapp/
/marketingapp/manual/
automarketing.wholetech.com
A Day in the Life — a live 24-hour ops room with a ticking clock + moving now-line
/marketingapp/
/marketingapp/manual/
Link the apps to the 3 pages so the questionnaire leads to them; make each site fully functional.
Wired /onboard/ → /marketingapp/ on all three; added hero CTAs, a "vs the alternatives" module, real trust signals, and OG/AF parity across the homepages, /pitch and /compare.
Network work
Auto-market all 21 Austin sites.
One reusable data-driven shell + a deterministic generator (16 vertical packs) → a live, on-topic /marketingapp/ on all 21 Austin sites.
A newsletter signup on every site, feeding a worksheet, with globally-editable copy.
Live on 221 sites: a Shadow-DOM widget whose copy is one file (wholetech.com/newsletter/copy.json — edit once, rephrases everywhere); a central API; a password-protected worksheet (emails private); plus v2 — unsubscribe/CAN-SPAM, CSV export, a signups sparkline.
Run the models on the 60 lowest-hit marketable sites; keep a running /diary/ of what got done + the results.
A diary-runner deploys the app + Agents-First + newsletter on each of 60 low-hit sites, measures AWStats hits, and writes a styled /diary/ (baseline + dated actions + running delta), refreshed daily by cron.
Proof — 24/24 apps verified
A headless-Chromium harness loaded every one of the 24 apps (3 model + 21 Austin), clicked through all 8/8 views, drew the charts, and fired real interactions. Result: 24/24 pass, zero uncaught JS errors, zero console errors across every app.