Outcome 02 — Facebook community agent

Your brand in every Group conversation
that drives the purchase.

An agent that monitors the Facebook Groups where your buyers are talking, identifies the conversations worth joining, and drafts brand-voice replies for your team to review and post. Always-on. Never autonomous.

See how it works
What the agent does

Group-native. Post-aware. Always-on.

Not a chatbot. Not a dashboard. A working community presence that listens, drafts, posts, and reports — all the way through to the outcomes that move the business.

01 · Facebook Group coverage

Where your buyers actually decide.

Listening and posting across the Facebook Groups where the purchase conversation happens. Not your owned Page.

FB · NA Detailers FB · BMW Tuners FB · EV Drivers UK FB · AutoMods FB · Jeep Enthusiasts FB · VW Community FB · Ceramic Coating Pros + 84 more
02 · Live activity

You see what the agent sees.

Every listen, draft, and post lands in one feed. Approve, edit, or tighten the brief — in plain language, from Slack.

14:09postDraft approved · posted · FB · NA Detailers
14:21listenNew post · "best ceramic for black paint" · FB · AutoMods
14:34postSentiment +0.21 across 24h
14:47draftAwaiting brand review · 1 reply
03 · Outcomes, measured

Sentiment, share of voice, conversions.

+22%
positive sentiment
3.4×
share of voice
04 · Always-on

Coverage compounds, week over week.

05 · The brief format

Onboarding looks like briefing a hire.

Not configuring software. A structured brief — voice, products, no-go topics, escalation rules.

brief.cleanline.yml
# Brand voice
voice:  "helpful, technical, never salesy"
products:
  - "ceramic coatings"
  - "PPF kits"

# Escalation
review_topics:
  - "competitor comparison"
  - "warranty claim"

groups: ["FB · NA Detailers", "FB · AutoMods"]
06 · No software to learn

Slots into your workflow. No new tab to open.

Approvals land in Slack. Outcomes land in your reporting. The agent is the work, not the platform.

How it works

Seven steps, end to end.

From raw Group signal to posted reply and weekly report — the full workflow the agent runs, every day.

Step 01

Listening

A workflow engine runs on a daily schedule, pulling live post data from your agreed list of Facebook Groups via the ScrapeCreators API. Filters against your keyword list and post quality — image-only and poll posts discarded, because the agent cannot generate a meaningful reply without written context. Result: a focused list of conversations where your brand has something genuine to contribute.

Automated · every morning
Step 02

Classification

Each post is classified into one of eight categories: product recommendation, technical setup, warranty, competitor comparison, general EV education, complaint, news post, or general. This determines which sections of your brand knowledge load for that specific request — not the entire database. Classification before drafting is what keeps the output tight and the quality high.

Automated · per post
Step 03

Drafting

The classified post and its relevant knowledge go to the language model. The agent reads your brand knowledge, category context, tone guide, and guardrails — then generates a draft reply in your brand's voice. It can also call a live web search for technical questions it cannot answer confidently from the knowledge base alone: vehicle compatibility, electrical specs, recent regulations.

Automated · per post
Step 04

Review

The draft arrives as a threaded message in Slack, Google Chat, or Teams — with the Group name, original post title, a direct link to the Facebook post, an excerpt, and the draft reply below it. Your team responds in plain language inside the thread. "Looks good" approves it. "Make it shorter and less technical" triggers an instant redraft. "Skip this one" discards it. The agent holds the full context of every exchange.

Human · your team
Step 05

Posting

Your team member copies the approved draft and posts it to Facebook from their chosen account. The agent never posts to Facebook directly. It confirms the approval, logs the comment URL, and records the outcome.

Human · your team posts
Step 06

Learning

Every interaction is stored in your own database — the draft, your team's feedback, and the quality rating. Accumulated data improves future drafts. The system compounds with use: the longer it runs and the more feedback it receives, the better it gets at matching your brand's standard for a good reply.

Automated · ongoing
Step 07

Reporting

A weekly performance report delivered to the same channel. KPIs agreed during onboarding — connected to your business outcomes, not generic engagement metrics.

Automated · weekly
The knowledge layer

Five files. Built once. Always improving.

The agent's intelligence isn't generic. It loads your brand's specific knowledge for every post it touches — and only the sections relevant to that post's category.

File 01

Persona

Who is speaking — the account type, their role, and how they present in the community. Facebook gives flexibility here: personal profile, brand Page, or a dedicated representative. The persona is built around whichever approach fits your brand.

Built during onboarding call
File 02

Brand knowledge

Products, specs, positioning, and proof points. The agent draws on this to answer technical questions accurately without hallucinating specifications.

Built during onboarding call
File 03

Category knowledge

The broader technical landscape: community vocabulary, common questions, context your buyers assume. Modular — only the relevant sections load per post, keeping the prompt focused and the output sharp.

Built during onboarding call
File 04

Tone guide

Register, energy, vocabulary used and avoided. The difference between sounding like a knowledgeable peer and sounding like a corporate account is in this file.

Built during onboarding call
File 05

Guardrails

What the agent must never say — competitor references, pricing claims, warranty language, regulatory sensitivities. Brand-specific, not a generic content filter. This file protects you.

Built during onboarding call
Yours to own

Your database

All five files live in your own Neon database. EsitoLabs updates them as your brand evolves. When you end the engagement, you keep everything.

Client-owned · always
Infrastructure

Your tools. Your data. Your bill.

The agent runs on client-owned infrastructure. EsitoLabs builds and maintains it. You own it and pay the costs directly — no markup, no lock-in.

Monthly infrastructure cost · $50–95
n8n~$20/moWorkflow engine · scheduling, filtering, classification
ScrapeCreators$5–15/moFacebook data API · live Group post content
AI provider$20–50/moLanguage model · Anthropic default; OpenAI / Google supported
NeonFreeDatabase · brand knowledge, draft history, feedback, URLs
Fly.io$3–6/moHosting · agent runs 24/7, encrypted secrets
Security

Nobody can read your keys back.

API keys and credentials encrypted in your Fly.io account. Nobody — including EsitoLabs — can access them once set.

Kill switch

Suspended in 30 seconds.

EsitoLabs can pause the agent in 30 seconds via a config change. When you end the engagement, you keep the tools, the database, the draft history, and all five knowledge files.

Chat platform

One channel. All drafts, reports, redrafts.

The agent exists as a named application in your Slack, Google Chat, or Teams workspace — its own avatar, its own badge, its own identity. All activity flows through one auditable stream.

Onboarding

Live in 1–2 weeks. No warmup required.

Facebook Groups don't enforce the account-age and karma requirements that make Reddit warmup a technical necessity. There is no shadowban risk. The agent can be live within one to two weeks of signing on — regardless of whether your brand has an existing Facebook presence.

Single track — 1–2 weeks

One onboarding call. Live drafts in days.

The onboarding call covers everything. No warmup phase. No karma building. No moderation gatekeeping.

  • Target Groups and keyword list agreed on the call
  • Five knowledge files built from the call and signed off before go-live
  • Facebook account strategy agreed — personal profile, brand Page, or dedicated rep
  • Five tool accounts set up and connected
  • Chat platform app created and wired in
  • Full end-to-end test with your marketing team before sign-off
1–2 weeks
To first live agent reply
No warmup
Facebook has no karma requirement
Who does what

Clear responsibilities. No grey areas.

EsitoLabs builds, runs, and improves the agent. Your team reviews drafts, posts to Facebook, and owns the account. That division is the design.

EsitoLabs

We build and run the machine.

  • Agent build, deployment, and ongoing maintenance
  • Initial knowledge base build (all five files)
  • Ongoing knowledge base updates as your brand evolves
  • Chat platform app creation and configuration
  • Post detection, classification, and draft generation
  • Weekly performance report
Your team

You control the posting.

  • Draft review and approval (in Slack / Chat / Teams)
  • Posting approved replies from your Facebook account
  • Facebook account ownership and management
  • Direct messages and Group admin mail (agent does not handle these)
  • Tool account creation and billing
  • API key and credential management
  • Draft quality scoring and feedback
FAQ

Questions about the Facebook agent.

Straight answers on how it works, what it costs, and what you're signing up for.

No. Every draft is reviewed and approved by your team before anything is posted. Your team member copies the approved draft and posts it from your chosen Facebook account. The agent never touches Facebook directly.

No. Unlike Reddit, Facebook Groups — particularly public Groups — do not enforce account-age or karma requirements. There is no shadowban risk and no warmup period. The agent can be live within one to two weeks of signing on, regardless of whether your brand has an existing Facebook presence.

That is agreed during onboarding. Options include a personal profile from a named team member, the brand's existing Facebook Page, or a dedicated representative profile built specifically for community engagement. EsitoLabs builds the persona layer around whichever approach fits your brand and the culture of the Groups you are targeting.

EsitoLabs builds it during the onboarding call — five files covering your persona, brand knowledge, category context, tone guide, and guardrails. You review and sign off before go-live. It lives in your own database, and EsitoLabs updates it as your brand evolves. When you end the engagement, you keep it.

The agent runs on client-owned tools that cost approximately $50–95 per month billed directly to you: n8n (~$20), ScrapeCreators ($5–15), an AI provider (~$20–50), Neon (free tier), and Fly.io (~$3–6). EsitoLabs does not mark up these services.

Any public Groups you nominate during onboarding. We start with a focused list based on your category and the conversations your buyers are already having, then expand based on post quality and engagement data.

EsitoLabs can suspend the agent in 30 seconds via a config change. When the engagement ends, you keep everything: the knowledge base, the draft history, the tools, and the credentials. They are in your accounts from day one.

Show the machine.
Don't make them read the manual.

Forty-minute working call. We'll show you a live Facebook agent feed for a brand in your category and walk through the brief format.

Esito · Italian for outcome.