Menu

Cognition

A marketing ops agent that runs on tested code

Anyone on the marketing team can ask in Slack for a campaign, an audience or an event review, and an AI agent runs it through tested code

Will Hunter GTM Engineer, Cognition LinkedIn

  1. The problem

    Marketing ops runs on playbooks: load the event list, create the campaign, build the audience, manage the ads, pull the numbers. Every run needs judgment calls, and the usual first move is to hire someone to own all of it. Cognition wanted to see how far it could get by engineering the work first.

  2. What they built

    Marketing Workbench: every marketing ops workflow built in code so Devin, Cognition's AI agent, can run and debug it. Anyone on the team asks in Slack, in plain English. Devin can stage a LinkedIn campaign from a brief as a draft that needs a separate yes to go live. It can lay out pending event nominations, with each nominee's title, account, pipeline and past attendance, then log the reviewer's decisions in Salesforce. And it can find accounts that came to three or more events this year with no open opportunity, then build and size an audience of their engineering leaders in Primer.

  3. So what

    Campaign and audience requests go from a Slack message to a staged draft without a new hire. Nothing goes live without a person's approval, and event decisions land in Salesforce with an audit trail.

Under the hood

For GTM engineers

Each job has a skill that tells Devin what to do, but the work runs through tested code, so Devin can't invent a field, guess at a record or write somewhere it shouldn't. LinkedIn Ads has no MCP, so the team wrapped its API for Devin to use safely. The playbook is managed like production code, with PRs, code-owner review and tests in CI, and merging to main updates everyone's Devin at once. Every write is read back from the target system, and when something breaks Devin debugs it and ships the fix through the same reviewed PR flow.

Source

More examples

Other systems GTM engineers have built, or see all of them.

  1. Modal

    Forecasts each customer's usage and alerts the team when a customer drifts from it, where closed-won and churn don't have clean definitions

    Chris Prinz · GTM Engineer

  2. Air

    Rebuilt outbound so no account quietly falls through the cracks; meetings booked went from 1% to over 6% of assigned accounts

    Tom Bonagura · GTM Engineer

  3. Vercel

    Turned what customers say on calls, in support cases and in Slack into a weekly report that tells each product team what the field is asking for

    Brandon Galang · GTM Engineer

Hiring someone who can build like Will?

We search for GTM engineers who have shipped systems like this one, and Joe speaks to every candidate before you do.