AI Sales Academy
en

Build your AI Sales Machine

Automation 25 min
Business problem

Knowledge from 14 modules without assembly into a system changes nothing: prompts in a notebook, a workflow in your head, AI in Chrome. The final project forces you to build ONE working machine — connected, monitored, with human-in-the-loop — that really does the work for the salesperson.

Proven method

Build the machine in 13 links (each from the modules):

  1. LEAD SOURCES: where leads come from (a form, an import, a tool),
  2. ENRICHMENT (module
  3. : company and contact data,
  4. ICP QUALIFICATION (module
  5. : AI JSON + rules (score/qualified/priority),
  6. RESEARCH (module
  7. : an automatic Prospect Brief,
  8. LEAD SCORING (module 5/Prospecting): the full score,
  9. PERSONALIZATION (module
  10. : L4 openers from signals,
  11. OUTREACH (module
  12. : a sequence with stop conditions,
  13. REPLY CLASSIFICATION (module
  14. : 8 categories,
  15. AI DRAFT (module
  16. : response drafts,
  17. HUMAN APPROVAL: approval before sending,
  18. CRM (module
  19. : automatic updates,
  20. FOLLOW-UP (module 6 Cold Email + classification): reactions to replies,
  21. ANALYTICS: metrics (reply, meetings, cost/lead, quality score). How to build: start from the MINIMUM (4 links: source → qualification → outreach → CRM) and expand (add research, classification, drafts in next iterations); test every link separately before connecting; human-in-the-loop on sending and replies; monitoring from day one (module
  22. . Completion criterion: the system processes a lead from source to CRM with minimal intervention, and you only do approvals and conversations.
Market example
BAD
Manual, inconsistent execution or chaotic automation lacking data validation.
GOOD
A Minimum Viable Machine (day 1): Google Sheets (lead source) → Make: enrichment (Hunter) + AI scoring (ChatGPT API → JSON) → rules: score ≥ 70 → email campaign (Instantly) → a reply → classification (AI) → a draft + task in the CRM. You do: send and reply approvals, conversations. Iteration 2: research (a Prospect Brief) before personalization. Iteration 3: full 8-category classification + auto-CRM. After 30 days: the machine handles 200 leads/month, you spend 1h/day on approvals and conversations instead of 4h on work.
Ready-to-use prompt 
You are an AI Sales Machine reviewer.

Context: I built my machine: [DESCRIPTION — links, tools, what works]. Planned links: [WHAT'S NEXT].

Task: Assess and develop my machine:
1. Map my links onto the 13 steps (Lead Sources → Analytics) — which I have, which are missing
2. Minimum: what's absolutely needed for the machine to work end-to-end
3. Weak links: 3 places that can fail (data, AI, deliverability) + how to strengthen them
4. Human-in-the-loop: where a human approves (check that nothing sends without consent)
5. Monitoring: what 5 metrics I track from day one (cost/lead, quality, reply, bounces, fails)
6. A 30-day plan: iterations (what I add in weeks 1/2/3/4)
7. An end-to-end test: describe the path of 1 lead through my machine

Format: 7 sections.
Model rekomendowany: Claude 3.5 Sonnet / GPT-4o Copy and paste into ChatGPT, Claude or Perplexity
ALTERNATYWNE SCENARIUSZE

Variations & alternative scenarios

Minimum version: 4 links (source → qualification → outreach → CRM).

Full version: 13 links + monitoring + an agent (module 12).

Practical exercise

Build the machine minimum (4 links: source → qualification → outreach → CRM) — configure 1 test workflow. Define 5 monitoring metrics. Write a 30-day plan (4 iterations).

Zapisano w przeglądarce

Steps to take

0/4