01 · gtm engineering, applied to paid social

clicked together
vs compiled.

most agencies click campaigns together in a browser. this one compiles them: audiences, campaigns, rotation and reporting built and maintained through LinkedIn's Marketing API and Meta's Graph API — the way an engineering team ships product.

get the teardown →
a week, as the system logs it

this actually runs

GTM engineering means building automated revenue systems — enrichment, scoring, routing, reporting — instead of billing agency hours for them. below is a real week's shape on a client account: what the agents did, and where the human hours went.

# week 34 · fintech account · operator: asher
tla.promote(post="ceo_credit_take")      → live in 3 campaigns, even rotation
audience.rebuild(engagers, lookback=90d) → 4,312 members in retargeting pool
leads.intake(meta_forms, li_forms)       → 41 raw → verified → deduped
leads.enrich(apollo, apify)              → scored vs ICP · 12 routed to sales
crm.hygiene(hubspot)                     → 0 orphaned contacts, funnel reconciled
report.compile()                         → monday 08:00 · read by the CEO
human_hours → strategy, creative, the founder's voice. everything else is the machine.
why the api matters

what changes when campaigns are code

campaign manager, by hand

  • creative rotation drifts — linkedin quietly favors one ad and starves your tests
  • bulk changes take an afternoon of clicking, so they don't happen
  • no audit trail of who changed what, when, or why
  • reporting is whatever the UI exports, reconciled never
  • scale is bought with junior headcount

the marketing api

  • even rotation enforced programmatically — tests actually test
  • fifty-campaign changes ship in minutes; optimization is weekly reality, not quarterly theatre
  • every change logged; the account has a readable history
  • reporting compiled from source, reconciled against the CRM
  • scale is capacity the system already has
linkedin marketing api meta graph api apify apollo hubspot claude own integrations · client-owned data
the ops layer

named agents, on your invoice

the automation isn't hidden inside a retainer — it's listed. each agent is a named line item you can see, question, and cancel. that's the difference between renting infrastructure and paying an agency's overhead.

unit 01

enrichment loop

apollo ↔ apify · runs on every new lead

every form lead verified, enriched with firmographics and both work and personal email, then scored against your ICP rubric before a human sees it. no more "who is this?" leads in the CRM.

unit 02

lead-intake agent

meta instant forms + li lead gen · hourly

pulls, dedupes and normalizes leads from every form across every channel, and routes qualified ones into the CRM the same way your website registrations arrive — one pipeline, one shape.

unit 03

reporting agent

weekly · reconciled vs crm

compiles spend, performance and pipeline attribution from the APIs — not the UI export — reconciles against hubspot, and ships a report a CEO actually reads on monday morning.

unit 04

warm-list manager

engagers → pools → sequences

everyone who engages a thought leader ad is captured into retargeting pools and warm outreach lists, so attention paid for once keeps working for quarters.

the honest-ai clause

what the machine does. what it never does.

the machine does: verification, enrichment, scoring, routing, rotation, reconciliation, reporting, and first drafts. it works nights and it doesn't bill hours.

the machine never does: publish in your founder's name without human approval, invent numbers, or touch strategy. drafting is AI-assisted, the voice is the founder's, the sign-off is human — every post, every time. in a feed that's 81% slop, that's not a disclaimer. it's the product.

— operating rule since day one
see it against your account

the teardown reads your machine — or the absence of one.

thirty checkpoints, including which of the four agents would move your numbers first. five working days, no call required, yours to keep.

get the teardown →