RIP is Marketing Gurus with 1000X ROAS
Cody Plofker, CEO of Jones Road Beauty, humorously critiques Manus AI—a recently integrated tool in Meta’s Ads Manager for campaign optimization—after it recommended increasing ad spend, prioritizing traffic campaigns, avoiding bid caps, and shifting budget from competitors like AppLovin, likening it to sales tactics from Meta representatives. - Grok.
That popup looked like a funeral notice. Felt the “time to die” in the OP — and every media buyer who saw the same screenshot this week had the same flicker:
Is this the Monday my job got shorter, or just the Monday my slide deck did?
The slide deck. The slide deck died. You’re fine.
Meta paid $2 billion for Manus.
Hammered it into Ads Manager seven weeks after closing the deal.
Here’s what the version sitting in your Tools menu right now literally cannot do:
Adjust a single bid
Change targeting on one ad
Move a dollar of budget
Make or edit any creative
Run on anything except Sales-objective campaigns
Touch 8 entire industries — Housing • Employment • Credit • Finance • Pharma • Politics • Social Issues • Elections
Operate in the EU
Last 60 days without silently breaking when your token expires
What it can do:
Pull weekly performance decks
Audience research questions
Competitor analysis from Meta’s Ad Library
Anomaly detection — “why did CPM spike Tuesday?”
Recurring scheduled reports
Plain-language Q&A on your data
Restaurant prints a 200-item menu.
Kitchen knows six dishes.
Manus in 2026 is the six.
The roadmap is real — Meta’s CAIO publicly called Manus “the execution layer.” But the 2026 version is read-only, one-objective, eight-industries-blocked, EU-locked.
The job that actually compresses today
The junior analyst who builds the weekly slide deck.
That role is going.
Structurally untouched: senior buyer • strategist • creative lead • attribution analyst.
Manus literally cannot reach those.
Wall Street ran this movie 20 years ago
Floor brokers — humans who manually chunked big trades to avoid moving the price — got vaporized by execution algos through the 2000s.
The traders who survived weren’t the ones who fought it. They were the quant researchers, risk managers, and strategy heads.
They got paid more after the transition than before.
 The role that died |
 The role that ate the boats |
|---|---|
| Floor broker — manual chunked execution | Quant researcher / strategy head |
| Media buyer — manual dayparting, pacing, pause/resume | Strategist / attribution lead / creative judgment |
The pivot is up the stack, not out of it.
I’ve been running it on a Sales-objective DTC account since the day after launch
The weekly deck I used to grind 40 minutes on every Monday morning takes 4 now.
It hallucinates competitor stuff every other deck — last week it invented a campaign name that didn’t exist.
So I spot-check 2-3 numbers against native Ads Manager before anything goes near a client.
That’s the actual workflow.
Heads up — if your Manus integration goes silent after 60 days and starts feeding stale numbers? That’s just the default Meta access token expiring with no error message.
Switch to a System User token in
Business Settings → Users → System Users.4 minutes • zero breaks • lifetime.
📅 Where Manus is actually going — and how long you've got
The read-only ceiling is a beta-stage trust gate, not a permanent feature.
Stated roadmap (per ALM Corp’s Feb 2026 deep guide):
analysis (today) → optimization recommendations → supervised execution (single-confirm approve) → fully autonomous execution
Each phase is a separate ceiling-lift.
Best estimates:
12-24 months to supervised execution in Ads Manager
36+ months before senior buyer roles feel real squeeze
Cross-platform expansion — WhatsApp Business + Instagram Creator tools next per Meta’s Q4 2025 investor commentary
The Manus team is now housed in Meta Superintelligence Labs (MSL) — a new division.
The window to pivot is real and big — but it is a window, not a hallway.
Quiet advantage: the buyers who learn Manus’s prompt grammar now, while it’s read-only and harmless, will be 12 months ahead of agencies still arguing whether to “trust the output” when supervised execution lands.
🎤 Who actually said what (the receipts)
“the execution layer Meta needed to compete with Google, Microsoft, and OpenAI”
— Alexandr Wang, Meta Chief AI Officer (ex-Scale AI founder), on the Manus acquisition
“AI that can act on behalf of individuals and businesses, not just answer questions.”
— Mark Zuckerberg, “personal superintelligence” letter, July 2025
“The graphical user interface of the advertising manager is living on borrowed time…”
— Barbaros Ozturk, on the trajectory of agentic advertising tools (cited in Marketing Agent Blog)
“Right now I’m not taking any of the outputs and sending them to clients because they’re just not reliable enough.”
— Chris Rigas, VP Media at Markacy, Digiday Mar 3, 2026
“Autonomous media buying is no longer theoretical.”
— Tim Vanderhook, Viant CEO, same Digiday piece
“I’ve seen a lot of the early adopters that automated tasks get really hurt in the long term when chatbots were hallucinating data.”
— Ryan Schuster, Director of Paid Search and Social, Brainlabs
🪤 The credit-drainage trap (if you go anywhere near standalone Manus.im)
Different product from the Ads Manager connector.
Standalone Manus.im is the bigger, more capable agent — it has a CodeAct feature that runs Python/Bash in a sandboxed VM, plus a Browser Operator that logs into accounts and clicks UI elements.
Treat it as the connector’s roadmap minus the safety rails.
It’s also where the horror stories live:
“Lost $39 in under 10 minutes” — pattern reported by multiple operators on Reddit. Agent Mode on the Plus plan ($39/month) drained the entire monthly credit pool when the agent got stuck in a loop.
No real-time credit alerts. No way to stop a runaway.
Tasks that should cost a few credits spiral into hundreds when the agent loops, makes extra API calls, or processes more data than expected.
Every script runs in a disposable VM — no persistent state, no installed libraries between sessions, no version control.
▸ Run scheduled tasks only — deterministic credit cost
▸ Avoid open-ended Agent Mode prompts
▸ Set a hard monthly cap in billing
▸ Watch usage like a hawk for the first two weeks
If you’re testing standalone Manus to scout the connector’s future — cheapest plan, eyes open.
🚪 Who can ignore all of this entirely
If you advertise in any of these — go back to your normal Monday, Manus literally cannot touch your account this year:
Housing
Employment
Credit
Financial Services
Pharma
Politics
Social Issues
Elections
Plus the entire EU (GDPR + EU AI Act).
Plus any campaign objective that isn’t Sales — so Awareness, Traffic, Engagement, Lead Gen, App Install, Video Views, Messages, and Catalog Sales are all walled off.
Plus View-only account permissions = blocked.
If you’re outside the intersection of:
Sales-objective
Non-restricted industry
Non-EU
Admin or Editor
Above Meta’s spending threshold
…the 2026 panic doesn’t apply to you.
Use the 12-24 month buffer to build the human-trust-and-judgment moat AI structurally can’t reach.
The funeral was for the slide deck.
Floor brokers died in the 2000s; the traders who became quant researchers bought the floor brokers’ boats.
Same movie, new theater — just don’t be the slide deck.
!