GTM Engineering

Most sales problems get diagnosed as headcount problems. Leads are slow, so hire an SDR. Outreach is not landing, so hire two more. Qualification is inconsistent, so write a longer playbook and hope.

Usually the bottleneck is not people. It is that nobody built the system, so a person is standing in for one.

That substitution is expensive and easy to miss, because a person filling a system-shaped hole looks like work getting done. They read the inbox, they classify the lead, they look up the company, they decide who it goes to. It all looks productive. It is a human doing the job of a switch statement.

GTM engineering is the practice of noticing those holes and building the thing that belongs there. Not a workflow diagram, not a tool purchase. Something that runs.

What I keep finding

Speed beats quality, up to a point nobody reaches. Almost every inbound improvement I have shipped came from answering faster, not from answering better. Interest decays fast, and most teams reply hours after it is gone.

Timing is most of outreach. The same message sent at a moment when the buyer actually has the problem converts several times better. At NuMarket, watching public license filings took cold conversion from 3% to over 25%, and the copy barely changed.

Use a model for the fuzzy part and rules for the rest. Classification is genuinely hard and benefits from an LLM. Routing is a business rule, and making it a model decision buys unpredictability for nothing.

Clean the data before you score it. Scoring on top of a bad CRM produces confident nonsense, which is worse than no score.

The approval step is the adoption step. Teams do not object to the agent doing the work. They object to it acting without asking. Adding a confirmation is usually what gets a system switched on.

Where I have done this

The projects below are the real ones, with the numbers attached. Most of it was built at inBeat across their sister brands, plus a contract with NuMarket.