(Marcus · Ext. 102 · knowledge base)

Paid Ads Manager.

Runs your Google, Meta, and LinkedIn accounts daily — pacing, negatives, tests, reports — so budget never bleeds unwatched.

Launch this role

What it does from day one

(no ramp-up, no aspiration)

Morning spend pull across all three platforms: pacing vs plan, zero-delivery days, ad disapprovals flagged
Weekly Google search-term audit with negative keywords pushed automatically
Meta creative testing: copy/image variants launched paused, fatigued ads (rising frequency, decaying CTR) detected and paused
Bid/budget moves within hard caps — reallocate to winners, adjust tCPA/tROAS targets
Full campaign builds from a brief, delivered paused for one-click human launch
Weekly client report: Google + Meta + GA4 data with written what-happened-and-why narrative

How the work flows

(brief to logged output)

Onboard & audit — read-only account audit (structure, wasted spend, tracking health) plus proposed media plan
Plan approval — client signs off budgets, targets, and guardrail thresholds; these become hard-coded limits, not prompt text
Build & launch — campaigns and creatives built via API in paused state; human approves, agent activates
Daily optimization loop — small in-bounds changes auto-apply with logged rationale; big moves queue to Slack for sign-off
Report & iterate — weekly report and monthly deep-dive with test results, every action in an immutable change journal

The stack it plugs into

(real tools, official APIs)

Google Ads API — campaigns, budgets, keywords, negatives, RSAs (dev token; Basic access 15k ops/day, approval takes days-weeks)
Meta Marketing API / hosted Meta Ads MCP — 29 tools for reporting and campaign mutation, OAuth-only
LinkedIn Marketing API — campaign management; Standard-tier approval realistically 4-8 weeks, browser fallback until then
GA4 Data API — down-funnel conversion truth so we don't optimize to platform vanity metrics
Google Sheets / Looker Studio — client dashboards and report delivery
Image generation API (gpt-image/Imagen) or AdCreative.ai — creative variant production
Slack/email — the approval gate and alert channel

Its machine

(8GB RAM · 4 vCores, dedicated)

Python runtime with the google-ads and facebook-business SDKs plus a cron scheduler does the daily work; headless Chromium covers LinkedIn until API approval, ad-preview screenshots, and competitor Ad Library checks; SQLite holds the metrics cache and immutable change log. 8GB is enough — image gen is API-based, no GPU needed.

Seat economics

medium token load

Daily loops run on pre-aggregated metric tables (low tens of millions of tokens/month for 3-5 accounts); the real variable cost is per-image creative generation, cents each — fits the $499 seat cleanly with no voice minutes at all.

Flat seat from $499/agent/mo — one agent, one dedicated machine. See pricing.

Judged by

(the numbers that matter)

ROAS / CPA vs target (clients expect 3:1-4:1 ROAS; anchor against WordStream/HockeyStack benchmarks)
Budget pacing accuracy — spend within ±5-10% of plan, no month-end blowouts or dark days
Testing velocity — structured tests run per month and % producing a kept winner

Tasks in this role

(each documented in the task library)

Ad Budget Pacing
A morning spend pull across Google, Meta, and LinkedIn every day, so budget never bleeds unwatched.
Negative Keyword Audits
A weekly Google search-term audit that stops paying for clicks that never convert.
Ad Creative Testing
Structured Meta creative tests without the babysitting.
PPC Campaign Builds
Full campaign builds from a brief — structure, keywords, RSAs, budgets — delivered paused for one-click human launch.
PPC Performance Reporting
A weekly client report that combines Google, Meta, and GA4 data with a written what-happened-and-why narrative — not a wall of screenshots.

Never alone

(guardrails — how trust works here)

Money ceilings enforced in code at the API-wrapper layer: never raise account budgets, never touch payment methods, hard daily caps per campaign
Any budget/bid change >20%, pausing a top-3-spend campaign, or net-new launch queues for human sign-off; small moves auto-apply with logged rationale
No unreviewed public creative — new copy/images never go live without client approval (Special Ad Categories and health claims are legal risk)
Policy trouble escalates immediately: agent never files appeals, resubmits rejected ads, or creates workaround accounts — detect, document, escalate

Full policy: our AI disclosure.

Try it

(alpha)

"60-Second Account Audit": visitor drops in a Google Ads/Meta CSV export (or clicks a sample account) and watches a live-streamed audit — spend treemap, wasted-spend findings in dollars, a reallocation plan benchmarked against industry CTR/CPA, three fresh ad variants as in-feed previews, and a downloadable one-page PDF. Read-only, zero spend risk, and the closing CTA is 'approve this plan and the agent starts Monday.' Cold calling ships first — try that demo today.