Sessão Mentoria Hoe Seat dia 02/09/26

Business-focused summary (mentorship session on using AI “digital teams” in a dance-school business)

1) Core idea: shift from “employee execution” to “owner orchestration”

• The owner/CEO must operate with an ownership mindset: set priorities, give direction, review results, and respond to key operational signals (e.g., WhatsApp messages).
• AI agents are treated as digital employees; the owner should function like a director/manager, not a bottleneck doing every task manually.
• “Companies need meetings” → run a consistent rhythm of reviews with agents (weekly guidance; adjust after seeing outputs/results).

Operating principle (process):

1. Decide (owner) → set weekly priorities + approvals
2. Align (human team) → provide context and expectations
3. Execute (agents) → analyze data → propose recommendations → produce plans/campaigns/creatives
4. Test/iterate (agents + owner review) → start/stop/change based on performance data
5. Owner decides → accept agent recommendations (question if needed, but ultimately decide)

---

2) Agent-driven operational improvements (real example + KPI)

A presenter reports using a Financial Agent with weekly meetings and business inputs (e.g., statements and billing data). The agent identified unnecessary expense drivers, including:

• Service packages
• Credit card

Outcome:

• Revenue increased by 52% in a month (as shown in the subtitles)

Key lesson: agents must be actively used and managed—not left idle.

---

3) Organizational structure: departments → multiple agents → escalation by problem type

They reference a “dance school of the future” model:

• In the future version, there are many agents.
• A general director/CEO agent coordinates department agents.
• In the session’s current implementation: 8 trained agents, one per department.

Escalation rule (which agent to ask):

• If you don’t know which agent to contact, ask the system/chat:
  • “I have a problem… which agent should I contact?”

Example: refund request after signing but not attending

1. Administrative agent first (contract terms, signing date, cancellation policy)
2. Financial agent next (refund impact / calculation)
3. Commercial agent after (offer renegotiation such as reschedule/pause instead of cancellation)

---

4) The training playbook: “audit → command → execute”

A disciplined method is taught to keep agent outputs accurate.

A) Personalization + “school project” (build the brain)

Two preparatory steps are emphasized:

• GPT customization / personalization

  • Define tone (direct/didactic/analytical), style, and communication preferences
  • Specify what to consider before research
  • Define what must be remembered (what’s convenience vs official source)

• School project (central knowledge dossier)

  • Upload business data: files, spreadsheets, decisions, curriculum, prices, schedule, capacity, etc.
  • This becomes the official source of truth, preventing guessing

B) Specialized agents: run an “Audit” command first

Each agent should execute an Audit to confirm:

• identity/department
• what data it has access to
• responsibilities and decision limits
• gaps (e.g., missing updated spreadsheets)

Purpose: prevent hallucinations/misrepresentation when data is incomplete.

C) “Task command” pattern (objective-first instructions)

When requesting execution, use a command that includes:

• objective
• context
• audience
• offer/promise
• format/deliverables
• limits (e.g., no exaggerated claims)
• validation/review process

They stress doing this once, then keeping it updated (e.g., yearly).

---

5) Marketing + creative system (GTM-style execution for enrollment)

The session focuses heavily on ad campaigns to fill classes using AI-generated marketing assets.

Framework: Ferrari analogy + “stop using tools lightly”

• “Ferrari for bread” metaphor: don’t underuse the system—use it to generate a full marketing pipeline.

Campaign performance approach

• Use a paid traffic model:
  • creatives → ad campaigns → clicks → landing page (not just organic posts)
• Testing loop:
  • generate many creatives → run tests → keep winners → iterate

Concrete creative workflow (static + video scripts)

They demonstrate:

• Static ad creatives (square formats like 1080×1080 mentioned for Meta ads)
• Scripts for recording (then use teleprompter + CapCut workflows)
• Generate multiple variants (e.g., “send Felipe 20 creative ideas”) → test → after results, rely on only a few winners

Why volume matters: if one creative fails, you still have enough alternatives to find what works.

Specific creative quality rules (examples)

• Avoid overloading creatives with too much text/info:
  • keep artwork lean; move details to captions and landing pages
• Don’t include “value” in the creative if it hurts engagement/relationship building:
  • build relationship first, manage objections later
• Distinguish:
  • Headline = text on the image
  • CTA = call-to-action (e.g., “Learn more. Click here.”)
  • CTA should prompt clicks, not long explanations

Example target audience + offer structure (GTM prompt style)

An “owner command” can include:

• targeting (e.g., women 35–55, shy, believe they lack coordination)
• offer (trial dance class, ballroom dancing)
• messaging angles
• number of deliverables (e.g., 3 angles + headlines/scripts with CTA to WhatsApp)
• approval gate before production

---

6) Lead funnel + sales operations change (process redesign)

They propose a funnel redesign to avoid unnecessary commissions:

• Route paid traffic to landing pages (they reference an “Arena” domain with multiple pages).
• Customer flow:
  1. Click ad → landing page
  2. Pay for trial lesson there
  3. Money goes directly to the business account
  4. WhatsApp message triggers scheduling
  5. Deal closes in class with owner/business team

Sales commission adjustment:

• Stop paying 30% commission on leads that convert directly with the owner.
• Pay commission only for leads that exit class without closing and are then handed to sales recovery.

Positioned as automation of the school sales process.

---

7) Experimentation, data, and KPIs (what they track implicitly)

Few numeric KPIs appear beyond the 52% revenue increase, but the KPI mindset is clear:

• Weekly review of:
  • sales totals
  • monthly fees
  • cash needs / revenue targets to pay expenses by deadline
• Marketing testing:
  • A/B testing creatives via paid distribution
  • data-backed decisions over months

Explicit target/timeline signals:

• Install autonomous agents: within 1–2 weeks
• System tuning/leveling by December
• Testing cadence: within the next 7 days, adopt the habit of asking which agent should handle a task

---

Actionable recommendations (from the session)

• Implement the owner operating system:
  • weekly priorities + approvals + review meetings
• Maintain agent audit + command discipline:
  • audit first (confirm data access)
  • then execute with objective/context/audience/limits
• Build your GPT “brain”:
  • personalization for your tone
  • school project as the official source of truth (prices, schedule, curriculum, spreadsheets)
• Marketing execution:
  • generate many creative variants (start with ~20) and test
  • keep artwork lean; move details to captions/landing pages
  • use WhatsApp CTA at the end of scripts/copies
• Sales process:
  • send traffic directly to landing pages for trial purchase
  • reserve sales-team commissions for recovery leads only

---

Presenters / sources mentioned

• Rosana Cabral (host/mentor)
• Marcelo Filho (mentee; referenced)
• Eugene (mentee testimony credited in subtitles)
• Regiane (mentee; referenced)
• Keila (participant referenced)
• Natalia (participant referenced)
• Silvia (participant referenced)
• Daisy / “Daisy has it” (participant referenced; provided revenue spreadsheets)
• Felipe (mentioned as the one generating/testing creatives)
• Aia (mentioned in instruction correction context)
• Elsa / Elzinha / Elsinha (participants referenced during GPT personalization tutorial)
• “Matthew” (Bible passage) referenced for a “two people believe” principle