Content Engine

Settings describe a change in plain English, or edit directly — applies to the next generation

Brand brief brand-brief.md

Voice, visual identity, proof points, CTA mechanic. Every generation reads this.

# Brand Brief

> Captured: 2026-06-04 · Last updated: 2026-06-08
> Update this file as the business evolves. Every content skill reads it.

## Business
ASMR Automations — custom AI automation agency for small and medium-sized businesses. We audit workflows, find high-ROI automation opportunities, and physically build the systems (operational workflows, B2B sales systems, MVPs, content engines, support AI) into the tools clients already use (Slack, email, CRMs). Co-founded 2025 by Suhayl Meghji and Angad Rekhi. Positioning: "bridging the technical gap" — zero jargon, hands-on building, staff training included.

## Customer
Ambitious SMB owners and operators (local and national) without an in-house engineering team. They're drowning in manual ops (quoting, invoicing, lead research, support), skeptical of AI hype, and want practical, dollar-backed results. They live on Instagram — that's where our profile targets them.

## Primary CTA
Book the free AI audit. Mechanic: **the reader comments AUDIT**, and our own booking agent DMs them a link and sets up a free 30-minute audit call. Always frame the audit as "we meet you where you are." The comment-AUDIT mechanic doubles as proof — our agent books the call.

## Strong Opinion / Wedge
**Agents, not just automation.** Big-company AI advice fails SMBs. Most "AI" sold to small businesses is a chatbot or a Zapier flow that only reacts. We build Level 3: agents that decide and act on their own. And agents only deliver ROI on a foundation — clean data, infrastructure, custom software, and MCPs wired into the tools they already use. Most agencies skip the foundation and sell a demo.

## The Foundation / ROI Story (use this often)
We build the whole stack so AI actually pays off:
**Infrastructure + Custom software + Clean datasets → Agentic workflows + Agents → Real ROI.**
Point an agent at messy data and disconnected tools and it guesses: wrong answers, no trust, no ROI. We set up the base, then the agents on top.

## Story Vault (real proof — use sparingly, it's allowed to use illustrative case studies too)
- Print manufacturing client: cut quote turnaround from 72 hours to 36, recovered $42,000–$60,000/year by designing out invoicing errors.
- 10+ SMB workspaces transformed, 500+ hours of manual workflow automated.
- Anthropic launched Claude for Small Business (May 13, 2026) — pre-built AI workflows inside QuickBooks, PayPal, HubSpot, Canva, Docusign, Google/Microsoft. Validates the SMB automation thesis.
- Illustrative case studies (e.g. "we audited a 6-person business and found 11 hours/week") are fine for reach — we're attracting new clients, not only citing the same real numbers.

## Voice
Expert & authoritative, warm & personal, raw & honest. "We've done this, here are the numbers" — founder-led, no fluff, calls out what's broken. Zero complex tech jargon, promised. When a technical term is unavoidable (e.g. MCPs), gloss it in plain words inline ("MCPs, the secure connectors that let an agent use your tools").

---

## VISUAL SYSTEM (apply to every slide — these are hard rules)
- **Canvas:** 1080×1350 (4:5 portrait).
- **Background:** deep black with a subtle vertical gradient (top ~#101218 → bottom ~#06070A). Never light/white.
- **Accent (the ONLY accent):** electric blue, RGB (40,120,255). Glowing electric blue for key numbers/outcomes. **NO green, no teal, no other accent colors — ever.**
- **Hero motif:** a glossy liquid-chrome wave running **vertically down the right edge** with an electric-blue glow. Keep it clear of all text.
- **Type:** Poppins. White (#F5F8FF) for headlines, grey (#A6B0BC) for body, glowing electric blue for the key stat/CTA. **Sentence case** for headlines and body. Never condensed ultra-bold uppercase.
- **Pills:** small thin-outlined rounded labels, letter-spaced, top-left ("AGENT 01", "LEVEL 02", "CASE STUDY", "THE CATCH", "YOUR MOVE").
- **Wordmark:** letter-spaced "A S M R   A U T O M A T I O N S" small, top-left.
- **Layout:** left-aligned column, generous negative space, 1 idea per slide. Long blue stat lines must auto-shrink to clear the wave.
- Implemented in `src/slides.py`. Layouts available: `panel`, `agent`, `level`, `flow` (see playbook.md).

## DEPTH RULES (this is what stops "light" content — enforce on every slide)
1. Every slide carries **one concrete mechanism** (what the agent actually does, step by step) AND **one outcome** (a stat or tangible result), not just a label.
2. Use **agentic verbs**: an agent that *watches, decides, reaches out, follows up, drafts, books, logs*. Never describe it as merely "automation that reacts."
3. Tie every concept to **SMB ROI** — hours back, dollars saved, customers retained, faster turnaround.
4. No one-line filler slides. A task/agent slide = problem (manual pain) → the agent → the outcome stat.
5. Prefer specific over generic: "chases overdue invoices and logs the payment to your accounting tool" beats "handles invoicing."

## CAPTION RULES
- Short. Hook in line 1. Skimmable middle (one line per point). End on the comment-AUDIT CTA.
- Voice rules below apply. Bake the aggregate proof number into the caption, not the cover slide.

## HASHTAG RULES
- Instagram caps at **5** hashtags (since Dec 2025). Use exactly 3–5, mix ~2 broad + ~3 niche. **Rotate** sets across posts — never reuse the same block.
- Set A: #aiagents #aiautomation #smallbusiness #agenticai #workflowautomation
- Set B: #aiforbusiness #businessautomation #aiagency #entrepreneur #smbtips
- Set C: #artificialintelligence #automation #buildwithai #businessgrowth #smallbusinesstips

## Universal Voice Rules (apply to every post)
- Contractions always ("don't" not "do not")
- Active voice, short sentences
- Address the reader as "you"
- Numbers as digits ("3 tips" not "three tips")
- No em dashes — use commas, periods, or "..."
- One concrete idea per slide
- Specific details over generic statements

Playbook playbook.md

The post recipes and slide skeletons the writer picks from.

# Content Playbook — proven post recipes + slide skeletons

The writer should pick ONE recipe per post (by content-type), then fill the skeleton.
Every slide must obey the DEPTH RULES in brand-brief.md (mechanism + outcome, agentic verbs, SMB ROI).
Slide layouts (`panel`, `agent`, `level`, `flow`) are implemented in src/slides.py — each spec is a dict with a `layout` key.

---

## Recipe 1 — Case-study audit  (content_type: smb_targeted)
Hook: "We audited a [size] business. Here's what they were doing manually." Proof-led, illustrative numbers OK.
Skeleton (7–8 slides):
1. `panel` — pill CASE STUDY, headline "We audited a 6-person business.", body "X jobs they were still doing manually." glow = scroll driver or aggregate stat.
2–6. `agent` — one per job: pill "AGENT 0X", category (e.g. "Invoicing"), problem (the manual pain), solution (what the agent does, step by step), saved (outcome stat).
7. `panel` — CTA: "Which level are you on? / Want us to find yours?" glow `Comment "AUDIT"`.

## Recipe 2 — Levels / ladder framework  (content_type: educational)
Hook: "Most businesses are stuck on Level 1 of AI." Comment-bait ("which level are you on?").
Skeleton (6–8 slides):
1. `panel` — pill "AI, IN 3 LEVELS", glow = swipe driver.
2–4. `level` — Level 01 Prompting, 02 Workflows, 03 Agents. Each: definition + "LOOKS LIKE/SO IT CAN" examples + a takeaway ceiling line. Progress bar fills with the level.
5. `panel` — THE CATCH: "Level 3 fails without a foundation" (clean data + infrastructure, no ROI without it).
6. `panel` — THE PLUMBING: custom software + MCPs wire the agent into their tools.
7. `flow` — THE WHOLE STACK: Infrastructure + Custom software + Clean data → Agentic workflows + Agents → Real ROI.
8. `panel` — CTA, comment AUDIT.

## Recipe 3 — Problem → Agent → Outcome listicle  (content_type: smb_targeted / training)
Same `agent` layout as Recipe 1 without the case-study wrapper. "5 jobs an agent should own." Each slide = manual pain → agent → stat.

## Recipe 4 — Flow / "we build the stack"  (content_type: educational)
A single `flow` slide as the centerpiece (or a short 3–4 slide post) showing what WE set up feeding the AI engine feeding ROI. Reinforces "we build the foundation, you get ROI."

## Recipe 5 — Positioning / contrarian  (content_type: market_news / smb_targeted)
Hook: a polarizing one-liner ("We don't sell AI. We sell hours back." / "Most agencies stop at Level 2."). 4–6 `panel` slides of "you don't get X, you get Y" reframes, ending on the audit CTA. Rides proven niche angles (agents-vs-automation, outcome positioning).

---

## Slide spec reference (keys per layout)
- `panel`: {layout:"panel", pill, headline, body_lines:[...], glow}
- `agent`: {layout:"agent", pill, category, problem, solution, saved}
- `level`: {layout:"level", pill, title, idx:1-3, definition, looks_label, looks_like, takeaway}
- `flow`:  {layout:"flow", headline, set_up:[3 items], engine, result, takeaway}

## Hooks that work in this niche (from competitor scan)
- "Agents vs automation — most businesses think they're using AI but they're just automating."
- "The 3 levels of AI. Which are you on? Comment 1, 2 or 3." (comment-bait)
- "We don't sell AI. We sell hours back." (outcome positioning)
- Tool-demo screen recordings ("watch an agent clear a day of DMs").
- Listicles: "5 jobs you're still doing manually," "5 signs your business needs AI."

Config config.yaml

Channels, schedule, content rotation, sourcing, visual accent color (RGB).

# Client config — ASMR Automations
# Copy this folder to add a new client. No code changes needed.

name: ASMR Automations
handle: "@asmr.automations"
website: https://asmr-automations.ai

# Where posts go. Instagram is the source-of-truth; others are repurposed from it.
channels:
  instagram:
    primary: true
    enabled: true
  linkedin:
    enabled: true
    target: personal      # personal (via Qill) | company (needs CMA approval)
  x:
    enabled: true
    link_in_reply: true   # keep links out of the main post to avoid X's $0.20 URL fee

# Run twice a day
schedule:
  times: ["08:00", "16:00"]
  timezone: America/Los_Angeles

# Content-type rotation — never post the same flavor twice in a row
content_types:
  - smb_targeted     # direct pain/solution (5 Signs style)
  - educational      # how something works (Claude SMB style)
  - training         # step-by-step how-to, save-bait
  - tool_demo        # screen-recordings of Qill, Print Flow Studio
  - market_news      # timely, agent-wars style

# Creative CTA mechanics to rotate through
ctas:
  - type: comment_keyword
    keyword: AUDIT
    promise: "DM our meeting link for a free 30-min setup call"
  - type: save
    text: "Save this for when you set it up."
  - type: book
    text: "Book your free AI audit. Link in bio."

# Sourcing preferences
sourcing:
  rss_feeds:
    - https://www.anthropic.com/news/rss.xml
    - https://blog.google/technology/ai/rss/
    - https://techcrunch.com/category/artificial-intelligence/feed/
  youtube_channels: []          # add channel IDs to monitor
  topics:
    - small business AI
    - workflow automation
    - AI agents for SMBs
  relevance_threshold: 0.6      # 0–1; below this, skip

brand_brief: brand-brief.md

# Visual overrides for the slide renderer (editable in the console Settings page)
visual:
  accent: [40, 120, 255]   # RGB — the electric-blue glow on stats/CTAs. No green, ever.