How to Build & Sell AI Automations: Ultimate Beginner’s Guide

Concise summary (technology, product features, tutorials, and monetization)

Core concepts

AI automation = systems that use AI to automatically perform complex tasks that previously required humans. Unlike classic IF/THEN automations (Zapier‑style), LLMs and generative models add reasoning, summarization, extraction, natural language generation, voice, vision, and decision‑making.

• Three mental buckets of AI automation:

  1. Conversational AI (chatbots, voice agents)
  2. AI tools (on‑demand helpers a person calls to perform a job)
  3. AI workflow automations (end‑to‑end workflows that trigger, decide, and execute)

• Common automation architecture (assembly‑line model):

  • Trigger → Filter → Actions → Intelligence layer (AI/prompting) → Formatter (data cleanup) → Output

---

Key platforms and products mentioned

• Workflow builders (command center): make.com (used in tutorials), Zapier, n8n
• Databases / spreadsheets: Airtable (used heavily), Google Sheets
• Forms / intake: Tally (used), Typeform
• Communication & scheduling: Gmail (OAuth setup), Slack, Calendly
• AI models / providers: OpenAI (ChatGPT, GPT search models), Google Gemini (mentioned), ElevenLabs (voice)
• Voice agents: Vappy (used in tutorial) — for outbound/inbound AI phone calls, transcriptions, call summary and analysis; alternatives: ElevenLabs, Twilio/Vonage for phone numbers
• Proposals / e‑signatures: PandaDoc (referred to as “Panda Do”) for templated proposals with tokens and e‑sign workflow
• Misc: HTTP/API calls, JSON payloads, OAuth client ID/secret, API tokens

---

Three tutorial builds (course structure)

Chapter structure: Foundations → Hands‑on builds (3 progressive builds) → Monetization blueprint

Build 1 — Basic AI lead qualification (beginner)

Tools and flow:

• Tally form → Airtable base (fields mapped to form)
• Airtable AI fields for automatic qualification and an AI‑generated sales message
• make.com scenario watches new Airtable records (triggered by created_on)

Workflow actions:

• Filter for “qualified” leads
• Router to run parallel actions
  • Gmail module sends scheduling email (Calendly link)
  • Slack module posts the AI sales message
  • Update Airtable contacted_on timestamp

Practical features covered:

• Airtable AI fields, mapping form → DB, make.com scenario setup, routers (parallel branches), formatters/regex to clean Slack text, scheduling automation runs, connecting Gmail (full OAuth client creation in Google Cloud explained)

Build 2 — Add outbound AI voice agent and personalization (intermediate)

Adds:

• Vappy voice assistant (assistant prompt, voice choice, ambient noise, voicemail detection) integrated into make.com

Flow:

• When lead qualifies → Vappy places outbound call → pause (sleep) while call runs → HTTP request to Vappy API to fetch call record → route based on call outcome (answered vs not answered)
  • If answered: use Vappy’s call summary + evaluation (pass/fail whether they want a proposal) → update Airtable (summary, interested checkbox)
  • If not answered: fallback email & notify sales (from Build 1)

Personalization:

• Pre‑call web research using OpenAI (search model) to generate company research passed into the assistant via variables (assistant override) so the call is personalized

Technical notes:

• Switched from built‑in Vappy module to manual HTTP POST for advanced variable passing
• Use API keys, bearer authorization header
• JSON body building and sanitization (remove line breaks via formatter) to avoid invalid JSON

Build 3 — Automated proposal generation and delivery (advanced)

Adds:

• OpenAI generates a structured proposal JSON (prompted as a sales expert and instructed to output JSON)
• Use “Pass JSON” module to break out fields
• PandaDoc create/send document using a tokenized template (proposal.goals, proposal.services, pricing, etc.)
• PandaDoc sends e‑sign link → update Airtable proposal_sent_on

Features covered:

• Creating PandaDoc templates with placeholders
• Filling tokens dynamically
• Generating proposal text programmatically
• Sending documents and leveraging PandaDoc notifications for views/signatures

Outcome:

• Full lead lifecycle: form → qualification → personalized call → call analysis → auto proposal → e‑sign

---

Technical implementation details & tips

• Authentication:

  • Airtable tokens (read/write)
  • Gmail requires OAuth client ID/secret & consent screen configuration
  • Vappy API keys
  • Store tokens securely (some are view‑once)

• API handling:

  • Use make.com HTTP modules for custom API calls (GET/POST)
  • Set headers (Authorization: Bearer , Content‑Type: application/json)
  • Pass/parse JSON responses

• Data formatting:

  • Use formatter modules (text replace, regex) to clean AI outputs before embedding into JSON or messages

• Scheduling:

  • Configure scenario schedule (e.g., every 15 minutes) once tested
  • Use created_on triggers to avoid reprocessing old records

• Reliability improvements:

  • Replace fixed sleep delays with webhooks to detect call end events
  • Handle call timeouts/long calls, voicemail detection, and fallback paths

---

Troubleshooting and learning advice

• Expect breakage: platforms change; debugging and iteration are normal

• Tools for debugging:

  • LLMs (ChatGPT) for stepwise troubleshooting
  • Google searches, platform docs, YouTube tutorials, community forums/Discord
  • Screen‑assist tools (e.g., Google AI Studio screen viewers)

• Problem solving:

  • Include detailed context in LLM prompts (screenshots, full error messages)
  • Practice reading docs and build the “debugging muscle”

• Common errors:

  • JSON formatting / invalid characters
  • Wrong field references (e.g., deleted modules)
  • Permission/scope issues in OAuth setup

---

Monetization strategy and market analysis

• Market opportunity:

  • McKinsey and World Economic Forum stats cited on automation/job change — strong demand for AI skills
  • Large TAM among SMBs (example: 1.7M US businesses in $0.5–10M revenue range)

• Three sellable services:

  1. Education (workshops, training, courses)
  2. Consulting (identify automation opportunities and strategy)
  3. Implementation (build and deploy automations — higher technical bar)

• Go‑to‑market tactics:

  • Warm outreach to existing network (build an outreach list, offer value first)
  • Community + content flywheel: create tutorials and post into communities to build audience and credibility (case example: community member Rory)

• Practical advice:

  • You don’t need to build a global AI product — selling automation services to SMBs is a lucrative, immediate path
  • Build a knowledge gap and monetize it; decide whether you prefer building (implementation) or advising/teaching

---

Resources, templates, and where to follow along

• All step‑by‑step resources, templates, prompts, and scenario files available in the speaker’s free community on “School” (link in the video description)
• Recommended practice path:
  • Complete the free course builds
  • Build extra projects to expand experience before selling implementation services
  • Or start monetizing sooner with education/consulting

---

Main speakers and sources cited

• Primary speaker/instructor: Liam Mley — founder of Morningside AI and creator of Agentive SaaS
• Referenced organizations / sources: Morningside AI, Agentive, School (community), Naval Ravikant (clip cited), McKinsey, World Economic Forum
• Tools/vendors demonstrated: make.com, Airtable, Tally, Gmail/Google Cloud OAuth, Calendly, Slack, Vappy, OpenAI, PandaDoc, Twilio/Vonage, ElevenLabs

---

If you want, I can produce:

• A compact checklist of required accounts, API keys and permissions to run the three builds
• Ready‑to‑paste prompts used for Airtable qualification, Vappy assistant, OpenAI research, and proposal JSON generation