>_ Rasul Shaikh

// Operator case study ยท greenfield build no. 3

Zero to first revenue in nine months.

Company: Omnibound AI · AI search marketing platform · B2B scale-up
Role: AI GTM Engineer, in-house · Oct 2025 to Aug 2026
Operator experience. A full-time engineering role, not a client engagement.

< 3 days
Engine deploy time
5
Logos closed
28
Deals in follow up
1.5%+
Positive reply rate
4
Positioning shifts
300+
Mailboxes orchestrated

The situation

Omnibound sells AI search marketing, helping brands win visibility inside AI answers. New category, no inherited GTM infrastructure, no playbook to copy, and scale-up expectations on pipeline.

The usual answer is to hire a marketer, two SDRs, and a RevOps contractor, then wait two quarters for the machine to assemble itself. That is roughly 400 thousand dollars a year in loaded headcount before the first meeting is booked. The mandate here was the opposite: one GTM engineer, one system, and revenue accountability from day one.

The build

Phase 01 · Strategy

Strategy compiles first

Before any tooling: TAM sized, ICP defined, personas mapped, and first, second and third party signal sources selected. Which job posting patterns, stack signatures, website visitors and LinkedIn behaviours actually indicate buying intent. The engine was designed on paper before a single API key existed.

Phase 02 · Engine

Full stack live in under three days

An end-to-end chain that had previously been a two week deploy was engineered, tested and sending in under three days. 150+ domains and 300+ mailboxes warmed and orchestrated, with deliverability engineered before volume rather than after it broke.

Phase 03 · Workflows

Unattended workflows layered on top

A seven stage LLM copy engine running with no human review, agentic research that took six hour batches down to under thirty minutes, 500+ personalised ABM landing pages tracking Share of Answer, RB2B visitor de-anonymisation, and the full event and webinar engine. Every one wired into CRM instrumentation.

Phase 04 · Revenue

The loop closes on revenue, not activity

Attribution ran through a self-built instrumentation layer, so every closed deal traced back to the signal that started it. Five customers closed, eight opportunities qualified, twenty eight deals in active follow up.

Why this is worth reading

The interesting part is not the tooling. It is that the system kept running without anyone babysitting it, and that every decision it made was traceable afterwards. That is the difference between a workflow and a system, and it is the only thing I am really selling.