## 1. Executive Summary

Sergio Fernandes is the founder of an agency that scaled past R$400k/month and later co-founded BaseWorks, an AI workspace that lets a marketing team train one company-owned model on its own content, docs, and client material. This episode is a two-track operator playbook: how to scale a service business (sales leadership, CRM hygiene, headcount economics) and how to convert your own operational pain into software that other agencies pay for. His core agency thesis: revenue-per-head has a hard ceiling around R$15k/month per person in paid media services, and AI is the only lever that breaks it. He also argues that no inbound channel will ever book you a R$50k/month client big retainers still close through founder-led relationships.

## 2. Alpha / Sauce

- **Founder runs sales until ~R$300 400k/month.** He would not have hired a sales leader before that point; even at R$400k he was unsure what to delegate. The close on contracts around R$50k still ran through him, with support staff in the process.
- **Cadence beats one-shot outreach.** Up to nine touches across channels (email, WhatsApp, LinkedIn, phone) maximize cold-lead response, and a "breakup" message on touch eight or nine where you formally withdraw often triggers the reply.
- **Healthy agency economics: R$8 12k revenue per person per month; R$15k is the practical ceiling.** AI implementations lift that ratio immediately, which is why BaseWorks' pitch to agency owners is margin, not novelty.
- **Cohort-by-segment analysis found the wedge.** Instead of average retention, he split retention and content volume by customer segment; agencies showed dramatically higher retention than infoproductors or e-commerces, so in February 2023 the company pivoted to B2B agency focus.
- **Build the product as your own guinea pig.** His agency (300 350 clients at the time) was the first BaseWorks deployment; internal copywriter feedback loops ("three things bad, one good") drove iteration.
- **Distribution before ad spend:** classes inside friends' mentorships, event stages, and influencer integrations where creators *use the tool as a teaching example* rather than a "subscribe now" plug measured quarterly, with the influencer roster rotated every three months.
- **Google Ads worked from day zero; Meta required a different shape.** Google: high-intent AI keywords straight to a home page with a 2 3 minute demo video plus free trial. Meta: a paid-grade lead magnet (templates, spreadsheets), a thank-you page CTA to book a demo, and a nine-step qualification form that cost him R$1.80 per lead.
- **CRM is the operating system of sales.** Pipeline stages (lead generated qualified demo held proposal sent closed), conversion-rate tracking between stages, a written playbook, and automated qualification gates inside PipeDrive.
- **Generic AI-content agencies are dying to ChatGPT.** The defensible layer is proprietary training data, per-client "second brains," and manager-level control not raw generation.

## 3. Methods & Setup Step-by-Step

1. **Document how work actually gets done.** At R$50 60k/month revenue he wrote Notion SOPs per role (designer, media buyer, content, copy/CRO, account manager) this preceded any scaling ambition.; setup: Notion process docs per role.
2. **Audit your own bottleneck list:** more clients, higher margin, growth without proportional hiring, more service capacity per team member.; setup: one-page founder bottleneck list.
3. **Fix operational friction before automating.** His example: 15 ad creatives spread across Drive folders with permission steps slowed the media buyer more than the creative work itself.; setup: audit file/permission flows per recurring task.
4. **Design your team's AI policy deliberately.** Different employees using different AIs with company data was his trigger to build one controlled workspace.; setup: single governed AI tool per team.
5. **Make each project a "second brain":** load winning ads, brand PDFs, tone of voice, ICP description; every generated asset draws from that training set, segmented per client.; setup: one trained project per client.
6. **Stay head-of-sales yourself:** learn cadence flows (nine-touch multi-channel sequences), CRM stage design, and playbook writing before delegating.; setup: CRM cadence + stage conversion dashboard.
7. **Qualify hard on the front end:** a long form (site, revenue, headcount, segment) feeding CRM automation that auto-rejects non-fit leads and routes fits to an SDR.; setup: multi-step form CRM automation.
8. **Run intent capture on Google:** brand terms, "AI for marketing" head terms, long-tail, and indirect competitor terms.; setup: Google Ads, budget always above the daily cap.
9. **Run volume on Meta via lead magnet:** deliver genuinely paid-quality templates free, then CTA a demo on the thank-you page.; setup: lead magnet page thank-you CTA CRM.
10. **Earn third-party credibility:** speak at events, teach inside partner communities, demo live on your phone at stands; pay influencers fixed + variable but only for organic *usage-as-example* integrations.; setup: quarterly influencer budget + UTM/tracking plan.
11. **Evaluate channel ROI on a quarterly horizon** influencer and community work rarely attributes last-click within 90 days; rotate and reassess every three months.; setup: quarterly channel budget review.
12. **Hire for versatility first:** his first three hires (copywriter, CS, operations) came from inside his own agency; operations was chosen for range, not specialization.; setup: agency product internal transfer plan.
13. **Lifestyle design check:** a "perfect average day" exercise (a la Peter Kell) kept him honest about why he wanted the numbers he wanted.; setup: written ideal-day exercise.

## 4. Tools & Stack

| Layer | Tool / Asset | Role in the system |
|---|---|---|
| AI workspace | BaseWorks | Proprietary-trained AI, per-project second brains, manager control |
| CRM | PipeDrive | Pipeline stages, cadence flows, automated lead gating |
| Docs / SOPs | Notion | Role-by-role process library written early |
| Paid search | Google Ads | Intent capture: brand, AI terms, competitor terms |
| Paid social | Meta Ads | Lead magnet funnel with 9-step form (R$1.80/lead) |
| Training data | Winning ads, brand PDFs, tone guides | Fed into each project's model |
| Distribution | Events, mentorships, influencer integrations | Credibility channel; quarterly measurement |
| Feedback loop | Internal agency staff | First users; "3 bad / 1 good" iteration rule |

## 5. Actionable Checklist

- [ ] Write Notion SOPs for every role before the next hire.
- [ ] List your top 5 operational bottlenecks and mark which one a tool could remove.
- [ ] Map every recurring task's file/permission flow; eliminate friction first.
- [ ] Consolidate team AI usage into one governed workspace.
- [ ] Build one trained "second brain" per client (ads, brand docs, tone, ICP).
- [ ] Run founder-led sales until revenue per head is proven, not until it feels busy.
- [ ] Implement a multi-channel cadence up to 9 touches incl. a breakup message.
- [ ] Track stage-to-stage conversion weekly inside your CRM.
- [ ] Add a long qualification form with auto-routing to an SDR.
- [ ] Launch Google Ads on high-intent terms to a short demo video + trial page.
- [ ] Ship a lead magnet you would normally sell, with a demo CTA on the thank-you page.
- [ ] Book one speaking or community slot per month and demo your work live.