content: TikTok script 2026-07-08 voice agents

content/instagram/2026-07-06-collection-gap/caption.txt

@@ -0,0 +1,9 @@
+Your business isn't unprofitable. It's unpaid.
+
+59% of small businesses are getting paid late right now. The average one is owed $17,500. Meanwhile the follow-up is a polite email and an awkward call. Whoever shouts softest gets paid last.
+
+We call it the collection gap. The work is done. The cash is missing.
+
+comment the word "cash" and I'll send you the Get Paid Checklist (free PDF): day-0 invoicing, automatic 3-7-14 reminders, one live ledger.
+
+#smallbusiness #cashflow #invoicing #accountsreceivable #operations #automation #businessowner

content/instagram/2026-07-06-collection-gap/resource_brief.md

@@ -0,0 +1,13 @@
+# Lead magnet — "Video 7 - Comment CASH - Get Paid Checklist.pdf"
+
+3 pages, Arqentia Instrument document system (per the V5/V6 resource series).
+
+1. **Cover** — COMMENT CASH pill; title "Get Paid Checklist"; flow INVOICE → REMIND → LEDGER →
+   PAID; core question "How much of your revenue is sitting in unpaid invoices right now?";
+   pass standard "Invoices go out day 0; reminders run themselves; one ledger shows who owes what."
+2. **Step 1 · Find the gap** — checklist: invoices sent days after work closes / follow-up lives in
+   someone's head / no standard reminder cadence / awkward calls instead of process / discounts
+   given just to get paid / nobody knows total outstanding today.
+3. **Step 2 · Close it** — checklist: invoice the day work closes / automatic reminders day 3-7-14 /
+   escalation step at day 21 (call script) / late fee policy stated on every invoice / one live AR
+   ledger reviewed weekly / stop work triggers for chronic late payers.

content/instagram/2026-07-06-collection-gap/script.md

@@ -0,0 +1,39 @@
+# Arqentia — "The Collection Gap" (Video 7) — keyword: cash
+
+**Format:** 9:16 IG Reel ~50s. Remotion atelier + stock B-roll (per arqentia-video5-video6-build).
+**Grounding:** Bluevine SMB study — 59% of small businesses experience late payments; average SMB
+owed $17,500 (https://www.bluevine.com/blog/small-business-late-payment-gaps); QuickBooks Jan-2026
+Insights: 42% report cash-flow problems.
+
+## Spoken VO (HeyGen — VO only)
+
+Your business isn't unprofitable. It's unpaid.
+
+Look at your receivables. 59 percent of small businesses are getting paid late right now. The average one is owed seventeen thousand five hundred dollars.
+
+Meanwhile you chase invoices by hand. A polite email here. An awkward call there. Whoever shouts softest gets paid last.
+
+We call it the collection gap. The work is done. The cash is missing.
+
+So we don't start with software. We audit the operation first. Invoice the day the work closes. Automatic reminders on day three, seven, and fourteen. One ledger that shows who owes what, at all times.
+
+You did the work. The system should collect it.
+
+comment the word cash, and I'll send you the get paid checklist.
+
+## Beat map (9 beats)
+
+| # | Beat | Line | Visual |
+|---|------|------|--------|
+| 1 | Hook title | "isn't unprofitable. It's unpaid." | title over avatar: "Not unprofitable. / Just unpaid." em unpaid. |
+| 2 | Exhibit 01 | "Look at your receivables" | stock: ledger/bookkeeping office (graded) |
+| 3 | Stats | "59 percent … seventeen thousand five hundred" | MG stat cards 59% → $17,500 |
+| 4 | Stamps | "polite email / awkward call / paid last" | word stamps over avatar |
+| 5 | Name it | "the collection gap" | MG typographic moment + sub |
+| 6 | Method | "audit first… invoice day-0… reminders 3·7·14… one ledger" | MG convergence: INVOICE/REMIND/LEDGER → COLLECTS ITSELF |
+| 7 | Exhibit 02 | "You did the work. The system should collect it." | stock: cash counting / register (graded) + device seal |
+| 8 | CTA | "comment the word cash" | COMMENT [CASH] pill |
+| 9 | End card | — | Arqentia. wordmark |
+
+Signature device: **the aging bar** — receivables aging (days) climbing during problem beats,
+snapping to "day 0 · collected" at resolution. (Meter family; distinct labels/motion from V4-V6.)

content/instagram/2026-07-07-dead-stock/caption.txt

@@ -0,0 +1,9 @@
+Your most expensive product is the one that never sells.
+
+Every shelf of slow movers is cash you already spent, frozen. You reordered it because a spreadsheet said you were low. The spreadsheet was wrong. Retailers lose $1.7 trillion a year to stock they bought wrong.
+
+We call it dead capital. Money you can see but cannot spend.
+
+comment the word "stock" and I'll send you the Dead Capital Checklist (free PDF): track what moves, flag slow movers the week they stall, one source of truth for every unit on hand.
+
+#smallbusiness #inventory #cashflow #operations #ecommerce #wholesale #businessowner

content/instagram/2026-07-07-dead-stock/resource_brief.md

@@ -0,0 +1,16 @@
+# Lead magnet — "Video 8 - Comment STOCK - Dead Capital Checklist.pdf"
+
+3 pages, Arqentia Instrument document system (V5/V6/V7 resource series).
+
+1. **Cover** — COMMENT STOCK pill; title "Dead Capital Checklist"; flow TRACK → FLAG → LEDGER →
+   MOVES; core question "How much of your cash is frozen on shelves that will never sell?";
+   pass standard "You know what moves, slow movers get flagged the week they stall, one source of
+   truth for every unit on hand."
+2. **Step 1 · Find the frozen cash** — checklist: reorders driven by a spreadsheet, not real
+   sell-through / no flag when an SKU stops moving / stock counts differ from the system / slow
+   movers discovered at year-end, not weekly / cash-value of dead stock unknown today / buying to
+   hit supplier minimums, not demand / no owner for inventory accuracy.
+3. **Step 2 · Thaw it** — checklist: track real sell-through per SKU / auto-flag SKUs with zero
+   movement in 30/60/90 days / one live inventory ledger reconciled weekly / reorder points set
+   from demand, not gut / a clear the-deck rule for aged stock (markdown/bundle/return) / a named
+   owner for inventory data.

content/instagram/2026-07-07-dead-stock/script.md

@@ -0,0 +1,39 @@
+# Arqentia — "Dead Capital" (Video 8) — keyword: stock
+
+**Format:** 9:16 IG Reel ~50s. Remotion atelier + B-roll (avatar ≤20% per 2026-07-06 rule: hook + CTA
+only; exhibits/MG carry the rest). Grounding: retailers lose $1.73T/yr to overstocks + out-of-stocks
+(Shopify/IHL retail inventory-distortion research); dead stock commonly ties up 20-30% of SMB inventory
+value.
+
+## Spoken VO (HeyGen — VO only)
+
+Your most expensive product never sells.
+
+Walk your warehouse. Every shelf of slow movers is cash you already spent, frozen. You reordered it because a spreadsheet said you were low. The spreadsheet was wrong.
+
+Retailers lose one point seven trillion dollars a year to stock they bought wrong. Too much of the wrong thing. Too little of the right thing.
+
+The overstock is the part you can see. We call it dead capital. Money you can see but cannot spend.
+
+So we don't start with software. We audit the operation first. We track what actually moves. We flag slow movers the week they stall. Then one ledger, every unit on hand.
+
+Your shelves should hold product. Not frozen cash.
+
+comment the word stock, and I'll send you the dead capital checklist.
+
+## Beat map (avatar ≤20%: only hook ~3s + CTA ~3s)
+
+| # | Beat | Line | Visual |
+|---|------|------|--------|
+| 1 | Hook | "most expensive product never sells" | avatar + title "Your most expensive product / isn't selling." |
+| 2 | Exhibit 01 | "walk your warehouse… cash frozen" | B-roll: dim warehouse racks / pallets (graded) |
+| 3 | Exhibit 02 | "reordered it because a spreadsheet said" | B-roll: spreadsheet / reorder screen macro (graded) |
+| 4 | Stats | "$1.7 trillion… too much wrong / too little right" | MG stat card $1.7T + two-line contrast |
+| 5 | Name it | "dead capital" | MG typographic moment + sub |
+| 6 | Method | "audit first… track what moves… flag stalls… one source of truth" | MG convergence TRACK/FLAG/ONE LEDGER → MOVES ITSELF |
+| 7 | Exhibit 03 | "shelves should hold product, not frozen cash" | B-roll: goods moving / forklift flow (graded) + device seal |
+| 8 | CTA | "comment the word stock" | avatar + COMMENT [STOCK] pill |
+| 9 | End card | — | Arqentia. wordmark |
+
+Signature device: **the freeze gauge** — % of inventory frozen climbing during problem beats, snapping
+to "moving" at resolution. Distinct labels/motion from the V4-V7 devices.

content/instagram/topics_log.md

@@ -0,0 +1,15 @@
+# Arqentia Instagram topics log
+
+Source of truth for what's been scripted/produced/posted. The live IG grid outranks Zernio history
+(Rafael posts natively too — never draft content already on the grid).
+
+| date | topic | keyword | status |
+|------|-------|---------|--------|
+| 2026-06-16 | AI needs a workflow | WORKFLOW | posted |
+| 2026-06-16 | Mapping workflow leaks | LEAK | posted |
+| 2026-06-17 | AI traceability / black box risk | TRACE | posted |
+| 2026-07-01 | The silent leak (shrinkage/downtime found late) | LEAK | posted |
+| 2026-07-02 | The one-person spreadsheet / single point of failure | EXCEL | posted |
+| 2026-07-02 | The handoff tax | HANDOFF | posted |
+| 2026-07-06 | The collection gap (late invoices / AR) | CASH | produced · awaiting review (Drive 7/6/26; zernio draft deleted per review-first directive) |
+| 2026-07-07 | Dead stock / dead capital (frozen inventory) | STOCK | produced · awaiting review (Drive 7/7/26) |

content/tiktok/2026-07-06-forward-deploy/caption.txt

@@ -0,0 +1,9 @@
+Amazon just spent $1,000,000,000 admitting something: AI does not deploy itself.
+
+AWS is embedding thousands of engineers inside customer companies to build agents with them. Not selling tools. Shipping outcomes. The model was never the bottleneck. Deployment is.
+
+Steal the play: one workflow, one builder inside it for two weeks, shipped against a real metric.
+
+comment "deploy" and I'll send you The Deployment Brief (free PDF).
+
+#ai #aws #aiagents #automation #buildinpublic #operator #deployment

content/tiktok/2026-07-06-forward-deploy/resource_brief.md

@@ -0,0 +1,7 @@
+# rs - Forward Deploy - Comment DEPLOY - The Deployment Brief.pdf
+4 pages, rs_ document system. 1) ink cover: "The Deployment Brief" + COMMENT "DEPLOY" pill +
+"run an AWS-style forward deployment in your own business, in two weeks". 2) paper §01 "pick the
+beachhead": one workflow, expensive + repetitive + measurable; write its current cost; name one
+owner; define the single metric. 3) paper §02 "the two-week embed": week 1 shadow + map + build
+smallest agent; week 2 run supervised, fix daily, measure; ship or kill on day 14; document what
+it touches. 4) ink back: "tools don't transform companies. deployment does." + @rafa_flows.

content/tiktok/2026-07-06-forward-deploy/script.md

@@ -0,0 +1,9 @@
+# rs_ — "Forward Deploy" (TikTok) — keyword: deploy
+
+~33s full-frame rs_ cut. Trend: AWS $1B Forward Deployed Engineering org (aboutamazon.com,
+2026-06-30) — thousands of engineers embedded with customers to co-build agentic AI.
+Hook "$1B. ONE ADMISSION." → cutaway (engineers at work) on "embedding thousands of engineers" →
+chips NOT SELLING TOOLS / SHIPPING OUTCOMES → stamp "// the bottleneck is deployment" →
+numbered chips ONE WORKFLOW / ONE BUILDER · TWO WEEKS / A REAL METRIC →
+thesis "tools don't transform companies. / deployment does."
+Resource: The Deployment Brief (4pp). CTA: comment "deploy".

content/tiktok/2026-07-07-context-engineering/caption.txt

@@ -0,0 +1,9 @@
+Prompt engineering is already dead.
+
+The people shipping real agents stopped tuning clever prompts months ago. The model is not the variable anymore. What you feed it is: the right files, memory, tools, and limits. Give a smaller model the right context and it beats a genius flying blind.
+
+They call it context engineering. It is the actual job now.
+
+comment the word "context" and I'll send you The Context Stack Checklist (free PDF): task, files, memory, tools, and hard limits, every run.
+
+#ai #aiagents #contextengineering #promptengineering #buildinpublic #automation #operator

content/tiktok/2026-07-07-context-engineering/resource_brief.md

@@ -0,0 +1,12 @@
+# rs - Context Engineering - Comment CONTEXT - The Context Stack Checklist.pdf
+4 pages, rs_ document system (ink cover + paper content + ink back).
+
+1. **Ink cover** — title "The Context Stack Checklist" + COMMENT "CONTEXT" pill + "stop polishing
+   prompts. build the context an agent actually needs, every run."
+2. **Paper §01 · the five layers** — TASK (one job, one output format, done-definition) / FILES
+   (the exact docs, not the whole drive) / MEMORY (what happened before that matters now) / TOOLS
+   (only what this task needs, scoped) / LIMITS (what it must never do, hard constraints).
+3. **Paper §02 · assemble it every run** — pull context fresh per task, not once / trim to what's
+   relevant (more is not better) / verify the agent can see each layer before it acts / log what
+   context produced which output / a smaller model with the right stack beats a bigger one blind.
+4. **Ink back** — "The model is not the edge. Context is." + @rafa_flows.

content/tiktok/2026-07-07-context-engineering/script.md

@@ -0,0 +1,34 @@
+# rs_ — "Context Engineering" (TikTok) — keyword: context
+
+~33s full-frame rs_ cut. Avatar ≤20% (hook + CTA only); cutaways + chips carry it.
+Trend: "context engineering is the new prompt engineering" (AI Engineer World's Fair, this week).
+
+## Spoken VO (HeyGen — VO only)
+
+Prompt engineering is already dead.
+
+People shipping real agents stopped tuning prompts months ago. The model is not the variable anymore. What you feed it is. The right files. The right memory. The right tools. The right limits. Give a smaller model the right context. It beats a genius flying blind.
+
+They call it context engineering. It is the job now.
+
+So stop polishing the prompt. Build the context stack. Task. Files. Memory. Tools. Hard limits. Every run.
+
+The model is not the edge anymore. Context is.
+
+Comment the word context. I will send the context stack checklist.
+
+## Beat map (avatar ≤20%: hook + CTA)
+
+| # | Beat | Line | Top panel |
+|---|------|------|-----------|
+| 1 | Hook | "prompt engineering is already dead" | avatar + "PROMPT ENGINEERING / is dead." |
+| 2 | Cutaway cx_screens | "shipping real agents stopped tuning prompts" | dark room, engineers at glowing screens |
+| 3 | Chips (what you feed it) | "right files / memory / tools / limits" | FILES / MEMORY / TOOLS / LIMITS |
+| 4 | Cutaway cx_stack | "give a smaller model the right context" | components assembling into a glowing stack |
+| 5 | Name it | "context engineering" | MG stamp "// context engineering" |
+| 6 | Numbered chips (the stack) | "task / files / memory / tools / hard limits" | 01 TASK 02 FILES+MEMORY 03 TOOLS+LIMITS |
+| 7 | Cutaway cx_fuel | "the model is not the edge" | fuel/energy into a dark engine core |
+| 8 | Thesis | "model is not the edge. context is." | two-line, "context" blue |
+| 9 | CTA | "comment the word context" | avatar + end card |
+
+Resource: The Context Stack Checklist (4pp). CTA: comment "context".

content/tiktok/2026-07-08-voice-agents/caption.txt

@@ -0,0 +1,14 @@
+Anyone can build a voice agent now. Almost none of them will work.
+
+No-code voice agent builders dropped everywhere this week, so every business is about to launch one. But a founder who deployed agents for 100+ companies says ~90% fail. Not because the AI is bad. Because the operation behind it is: an outdated knowledge base, a CRM that doesn't connect, and no plan for the human handoff.
+
+Before you build, do three things:
+1. Write your top 10 calls with the exact answer to each.
+2. Connect it to your real system of record, not a spreadsheet.
+3. Define the handoff. The moment it stops and gets a person.
+
+A voice agent is not an AI project. It's an operations project.
+
+Comment the word voice and I'll send you the Voice Agent Playbook: the readiness checklist we run before deploying one.
+
+#aiagents #voiceai #automation #buildinpublic #aitools #smallbusiness #operations

content/tiktok/2026-07-08-voice-agents/resource_brief.md

@@ -0,0 +1,39 @@
+# Lead-magnet PDF — "The Voice Agent Playbook"
+
+**Comment keyword:** VOICE
+**Promise:** the readiness checklist rs_ runs before deploying a voice agent — so it lands in the ~10% that
+actually work.
+**Format:** 4 pages, rs_ document system (ink cover + paper content + ink back), 1080×1350.
+
+## Page 1 — Cover (ink)
+- Title: **The Voice Agent Playbook**
+- Subtitle: The readiness checklist to run before you build. Not the AI. The operation behind it.
+- `COMMENT "VOICE"` pill + `rs_` wordmark.
+- Anchor line: "No-code builders shipped everywhere. ~90% of voice agents still fail. Here is why, and the fix."
+
+## Page 2 — Why 90% fail (paper)
+The three operational gaps, each = a header + one line + a one-line test:
+1. **Outdated knowledge base** — the agent answers from stale docs. Test: pull your last 20 tickets; can a
+   doc answer each verbatim?
+2. **A CRM that doesn't connect** — the agent can't read or write the record, so it can't act. Test: can it
+   look up a customer and log the call without a human?
+3. **No human handoff** — it traps the caller in a loop. Test: is there a defined trigger + a warm transfer?
+4. **No guardrails** — it answers questions it shouldn't. Test: is there an explicit "do not answer" list?
+> The pattern: none of these are model problems. They are operations problems.
+
+## Page 3 — The pre-build checklist (paper)
+Run this BEFORE you touch a builder. Each = a check box:
+- Written top-10 call reasons, ranked by volume
+- The exact, approved answer to each (source of truth, dated)
+- Connected system of record (CRM/ERP), read AND write
+- Defined handoff trigger + warm-transfer path to a person
+- A "do not answer / escalate" list
+- Test set: 20 real calls the agent must pass before go-live
+- An owner who reviews the transcripts weekly
+> Score: 5+ checked = ready to pilot. Under 5 = you are building the 90%.
+
+## Page 4 — Build it once, right (ink back)
+- The rs_ order of operations: **Map the calls → Connect the record → Define the handoff → Guardrail it →
+  Test on 20 real calls → Pilot → Review weekly.**
+- One-line proof: "The builder takes an afternoon. The operation is the whole job."
+- CTA: DM "voice" / `rs · @rafa_flows · // build & operate`.