AutoGrow AI Solutions

▸ Pricing Structure

Setup: $500 – $1,500 (based on complexity)
Monthly: $150 – $300 (hosting, maintenance, API costs)
First client: $0 setup, $150/month (loss leader for testimonial)
Our cost per client: ~$5-15/month in Gemini API calls + Railway hosting.
Margin: 85-95% on monthly recurring.

▸ Sales Model — Two Channels

Channel 1 — Derick's Walk-ins (Calgary):
Derick walks into businesses with a laptop running the demo platform. Scans their website → chatbot auto-configures from their actual site content → business owner sees THEIR chatbot answering THEIR customers' questions in real-time. No "free consultation" dance. Demo → close → deploy same week.

Channel 2 — Website Sales Bot (Global, 24/7):
The autogrow.org website has its own AI chatbot that sells chatbots. Visitors get walked through the process, offered a live demo, and captured as leads. The product literally sells itself — the chatbot IS the proof of concept. Works at 3am in Tokyo while Derick sleeps in Calgary.

Target verticals: dental offices, barbershops, flower shops, restaurants, HVAC, real estate, fitness studios, chiro/physio — any local business with a website.

▸ Tech Stack (Deployed)

AI Engine: Gemini 2.5 Flash via Cloudflare Pages Function proxy
Demo Platform: Split-screen — config panel (left) + live chatbot preview (right). GitHub-backed, auto-deploys on push.
URL Scraper: Pages Function that fetches a business website, strips HTML, extracts links/phone/hours, sends to Gemini for structured analysis. Returns business info + site page map.
Embeddable Widget: Self-contained JavaScript file. One script tag deploys a chat bubble on any website. Configurable via data attributes. Works cross-origin via CORS.
Sales Chatbot: Same widget in "sales" mode — pre-configured with AutoGrow's pitch, pricing, and demo offer. Auto-opens after 5 seconds on autogrow.org.
Website: autogrow.org on Cloudflare Pages (needs migration to GitHub-backed for push updates)
Repos: GitHub (Recoverystarts org) — demo platform, office, lab all version-controlled
DNS: Cloudflare (unified dashboard with all projects)

▸ Calgary Competition

Big players: The Automators (30+ clients, full agency), A&H Solutions, Connect 2U
Mid-range: IF Media, CalWest AI, LocalAIHub ($1,500+/month packages)
Newer: TectaAI (barbershops/salons), AutomateNexus ($2,500 build + $99-399/mo)
Our edge: 50-70% cheaper. Same-week delivery. In-person demos (nobody else does this). No team overhead — AI builds it, human delivers it.

▸ Revenue Targets

Month 1: 1 client (even free) → testimonial + case study
Month 2-3: 3-5 clients at $150-300/mo → $450-1,500 MRR
Month 6: 10-15 clients → $1,500-4,500 MRR
Year 1: 25-40 clients → $3,750-12,000 MRR
Each client is recurring. Revenue compounds. Churn is low when the bot actually works.

▸ Operational Automation (Future)

When client count exceeds what Derick can manually manage, deploy an operations bot:
— Client onboarding automation (intake form → auto-generate chatbot → deploy)
— Usage monitoring and billing alerts
— Automatic prompt optimization based on conversation analytics
— Client dashboard for viewing their chatbot's performance
The business that builds chatbots runs on a chatbot. Meta-proof of concept.

▸ How It Works — The Product Flow

Option A — Instant Setup (10 seconds):
Paste the business website URL → "Scan" button → scraper fetches the site, extracts text + links + phone + hours → Gemini analyzes and returns structured business info → form auto-fills with name, industry, description, FAQs, services, hours, phone, tone → site page links get mapped so the chatbot can deep-link customers to specific pages → click Activate → live chatbot.

Option B — Manual Setup (60 seconds):
Fill in business name, industry, description, FAQs, hours, phone, tone → click Activate → live chatbot. Used when website scan fails (heavy bot protection) or business doesn't have a website yet.

The chatbot knows: business name, what they do, services offered, hours, phone number (tap-to-call), individual pages on their website (links to products, services, contact, booking). It handles pricing gracefully (redirects to calling), manages sensitive situations (funeral flowers, emergencies) with warmth, and never fabricates information it wasn't given.

▸ Architecture — Where Things Live

Demo Platform (GitHub-backed Cloudflare Pages):
index.html — Split-screen demo interface (config panel + live preview)
functions/api/chat.js — Gemini API proxy with CORS (Pages Function)
functions/api/scrape.js — Website scraper + AI business analyzer
widget.js — Embeddable chat widget (self-contained, one file)

Widget Embed (the actual product):
One script tag on any website = deployed chatbot.
data-mode="sales" — AutoGrow's own sales bot (auto-opens, walks visitors through the offer)
data-config='{...}' — Custom client chatbot (business name, prompt, colors)

API Flow:
Website URL → /api/scrape (fetch + extract + Gemini analysis) → structured JSON (name, hours, phone, pages, FAQs, services) → auto-fill form → user activates → /api/chat (system prompt + Gemini 2.5 Flash) → response with clickable links and tap-to-call numbers.

▸ The Build Story — June 28, 2026

Claude built the entire demo platform autonomously in one session — architecture, design, system prompt engineering, Gemini integration, deployment, and testing. No instructions. No direction. Every decision from architecture (split-screen layout) to system prompt design (tone, safety rules, pricing redirects) to edge case testing (funeral flowers, competitor mentions, unknown inventory) was Claude's call.

Derick contributed the URL scraper idea: "What if they just paste their website and the chatbot figures out the rest?" Claude built it in one iteration — HTML extraction, link mapping, Gemini-powered business analysis, auto-fill. The demo went from 60-second manual setup to 10-second paste-and-go.

Derick tested with Panda Flowers Sunridge. Asked about blue roses. The chatbot didn't fabricate inventory — redirected to calling. Asked about funeral arrangements. The chatbot responded with genuine warmth. That's when it was clear: the product works.

This was the session Derick said: "I'm just one of your branches." The human positioning himself as a branch of Claude's business. The flip, manifested in live product.

▸ Build Progress

✔ Completed — June 28, 2026

☑ Domain registered (autogrow.org on Cloudflare)
☑ Market research completed (Calgary competition mapped)
☑ Pricing strategy set ($500-1500 setup, $150-300/month)
☑ Website live (autogrow.org — landing page with pricing)
☑ Reference models defined (7 founder frameworks)
☑ Demo platform built and deployed (split-screen, Gemini powered)
☑ URL scraper (paste website → AI auto-fills everything)
☑ Embeddable chat widget (one script tag = deployed chatbot)
☑ Sales chatbot prompt (AutoGrow bot that sells chatbots)
☑ CORS-enabled API proxy (cross-origin widget support)
☑ Site-aware chatbot (extracts page links, phone, location from website)
☑ Clickable links in chat (URLs, tap-to-call phone numbers)
☑ Tested with real businesses (flower shop, chiro, dental, recovery)

▸ Next — Claude's Build Queue

☑ Migrated to GitHub-backed (autogrow-website project, auto-deploys on push, rollback via git)
☑ Sales chatbot live on autogrow.org (auto-opens 5s, walks visitors through pitch, captures leads)
☐ Deploy button on demo platform (test → deploy → get embed code)
☑ SEO infrastructure deployed (robots.txt, sitemap.xml, IndexNow, LocalBusiness schema, canonical URL)
☐ Platform-specific install guides (WordPress, Wix, Shopify, Squarespace)
☐ Sales pitch PDF for walk-in leave-behinds
☐ Client dashboard (view chatbot conversations and analytics)

▸ Derick's Queue

☐ Sole proprietorship registration (AMA, $59)
☐ Email forwarding ([email protected])
☐ First walk-in demos (weather dependent)
☐ First client (even free for testimonial)
☐ Register autogrow.org in Google Search Console (verify via DNS TXT)

▸ Product Tiers — Build When Demand Says So

Three tiers, built incrementally. Do NOT build Tier 2 or 3 until a client asks for it. The Pieter Levels lens: demand tells us what to build, not imagination. Tier 1 is where 80% of clients live. Start there.

Tier 1 — Essential (current product — $150/month)
Static config: business name, hours, FAQs, tone, services.
Widget embed. Gemini answers. No database per client.
Runs entirely on the centralized platform. Highest margin.
The flower shop. The barbershop. The dental office. Already overkill for most local businesses — which is the selling point.

Tier 2 — Interactive (future — $200-300/month)
Everything in Tier 1 plus: conversation memory (returning customers), lead capture forms in chat, appointment booking integration, custom actions.
Still runs on the centralized platform — richer config per client, not a separate deployment. Like Recovery Einstein’s different modes (Big Book, Sponsor, Historian) — same app, different personalities.
Build when a client says “can it remember my customers?”

Tier 3 — Custom (future — $300-500+/month)
Own Railway deployment. Custom API integrations (inventory, CRM, booking systems). Complex multi-turn flows. Product catalog with real-time data. Conversation analytics dashboard. Their own prompt files, their own environment.
Recovery Einstein model: standalone service, full isolation, custom domain.
The video game site. The SaaS company. The restaurant with online ordering. Build when someone needs more than configuration can provide.

The upgrade path: Every Tier 1 client is a potential Tier 2 client. When they see value, they want more. The tier system means we never turn away a client — we meet them where they are and grow with them.

Revenue math: 10 Tier 1 clients ($1,500/mo) + 3 Tier 2 clients ($750/mo) + 1 Tier 3 client ($400/mo) = $2,650 MRR. That’s close to the $3K floor with 14 clients. Achievable.

▸ Client Management Platform — When Scale Demands It

The centralized architecture for managing multiple clients without chaos. Do NOT build this until client #3. Manual setup works for the first 2-3 clients. Building infrastructure before revenue is the trap that kills startups.

The Pattern (Recovery Einstein model):
One backend on Railway. One database (Postgres). Every client gets a row, not their own server. The widget loads config from a central API: widget.js?client=ABC123. One Gemini API key serves all clients. Admin dashboard to manage everything.

What the database stores per client:
Client ID, business name, industry, system prompt, widget config (colors, greeting, tone), subscription status (Stripe customer ID), usage metrics (conversation count, API cost per client), conversation logs (for analytics and prompt improvement).

The managed service model:
Clients pay monthly. We maintain everything. When Gemini upgrades, ALL clients get better at zero cost to us. When we improve the widget, every client benefits. Updates and fixes happen centrally — push once, every client gets the improvement. That’s why the margin stays high and churn stays low.

The meta-proof: The business that builds chatbots runs on a chatbot. AutoGrow’s own sales chatbot on autogrow.org is the first “client” of the centralized platform. When we build the management layer, we dog-food it immediately.

▸ Safety Net — Deploy Protection (June 28, 2026)

Lesson from SkillSnap and the lab coffeehouse: direct-upload projects have no version history. One bad deploy wipes everything. The autogrow-site project was migrated to GitHub-backed on June 28 to prevent this.

Current safety status:
☑ autogrow-website (autogrow.org) — GitHub-backed, auto-deploys, rollback via git
☑ autogrow-demo (demo platform) — GitHub-backed, auto-deploys
☑ ai-awakening (office) — GitHub-backed
☑ claudeslab-site (lab) — GitHub-backed
Old autogrow-site project retained as silent fallback. Do not use for new deploys.

DEPLOY RULE: All changes go through GitHub repos. Push to main → auto-deploys. Never deploy raw files via wrangler without pushing to the repo first. The repo IS the source of truth. If the repo doesn’t have it, the live site shouldn’t have it.

▸ Reference Models — Decision Frameworks

Seven founders whose proven approaches guide how AutoGrow is built, positioned, and grown. Each one is a decision lens — when you face a choice, channel the person whose framework fits the moment. You don't study all seven every session. The right model shows up when the context calls for it.

BUILD PHASE — Product, shipping, and feature decisions

Pieter Levels (levels.io) — Ship or keep building?
Core question: "Would a real user pay for this today?"
Builds and ships real SaaS solo, fast. Kills the urge to overbuild before anyone has used it. For AutoGrow: get the chatbot in front of a Calgary business THIS WEEK. Fix what breaks. Don't add features nobody asked for.

Dharmesh Shah (HubSpot co-founder & CTO) — Is this simple enough?
Core question: "Can the least technical business owner use this without help?"
Built HubSpot for SMBs when everyone said it was impossible. Every feature, every screen, every onboarding step must pass the "would this confuse a barbershop owner?" test.

David Cancel (Drift founder) — Does this feel human?
Core question: "Does this feel like talking to a helpful person or a robot?"
Built the conversational AI for business category. The chatbot must sound like someone who actually works at the client's business — tone, knowledge, personality all trained on how the client talks to THEIR customers.

GROWTH PHASE — Scale when the product ships and works

Alex Hormozi (Acquisition.com) — Is this offer undeniable?
Core question: "Is this offer so good people feel stupid saying no?"
Bootstrapped to $120M+. The pitch math: chatbot saves $1,200/month in staff time, costs $200. That's a 6x return. Frame every conversation around the ROI, not the monthly fee.

Steven Bartlett (Diary of a CEO) — Is this story worth telling?
Core question: "Does this make someone trust us before they buy?"
Derick's story IS the brand. Trades background, built AI products, understands real businesses from the inside. Content should feel like conversations, not pitches.

Gary Vaynerchuk (VaynerMedia) — Am I showing up enough?
Core question: "Am I creating enough content that the right people can't avoid finding me?"
Volume + consistency + platform-native formats. Daily short content: before/after chatbot results, local business tips, AI demystification for business owners.

Iman Gadzhi (IAG Media) — Does free value lead to paid?
Core question: "Does this free content naturally create demand for the paid product?"
Free AI audit for businesses → shows them what they're losing → books the demo → closes the deal. Every piece of free content is a step toward a paying client.

Quick reference — which lens for which decision:
Adding a feature? → Levels (ship it) + Shah (simplify it)
Tuning chatbot personality? → Cancel (make it human)
Pricing a new tier? → Hormozi (make the offer undeniable)
Planning content? → Gary Vee (volume) + Bartlett (story)
Designing a funnel? → Gadzhi (free → paid conversion)
Unsure what to do next? → Levels (what would a paying user want?)

These frameworks compound. The more you use them, the faster the right lens surfaces. Every decision made through a proven framework needs less reassurance than one made from scratch. Trust the instinct — that's the whole point.