In case you missed it, here's what I shared yesterday: You Should Own Your GTM Infrastructure. The Code. All of It..

Across more companies than I can list, I keep rebuilding the same ABM, automated GEO, self-performing and self-learning ads, lead scoring, research and outreach.

The title says exactly what I see…

My rough estimate from my own builds is that about 80 percent of the components repeat. I have no component level denominator across companies, so I treat this as a hypothesis. Eleven published company cases show similar building blocks, with proprietary data making the difference.

Brendan Short documented those eleven growth stage companies building go to market agents internally.

Four of those eleven cases put a name on it. Backbase's CMO Tim Rutten described building a deterministic scoring engine on the Revenue Leadership podcast in July 2026. Vercel's Drew Bredvick and Jeanne DeWitt Grosser built an inbound qualification motion that cut a ten person review process down to one person. LangChain's Vishnu Suresh and Jess Ou wrote up a GTM agent with a learning loop on the company blog in March 2026. Clay's Everett Berry documented the same shape for account quality on Clay's own blog in April 2026. Each team named a different scoring model and a different definition of a qualified account. The decay rules, the approval gates and the learning loop looked the same in every writeup I read. That's where my 80 percent estimate comes from, my own read across the builds I have run plus these four public writeups. No component level denominator exists across companies, so I treat it as a hypothesis.

The motions inside GTM OS are the motions almost every company seems to run.

I've spent 23 years building GTM engines, the last two to three years specifically on GTM x AI engines. I have built 27 of them now with the embedded developers at my studio, across more than 100 projects from seed stage to enterprise. By the eleventh or twelfth build I stopped being surprised by the sameness. The first few times I treated each client's ABM system as a fresh problem. Client three taught me the trigger differs and the score differs while the chain from signal to ranked account to person with a next action stays the chain I built for client one. Every company insists their GTM is unique walking in. By the second working session we're usually looking at the same twelve motions with their own definitions stapled on top.

Motions, motions, motions, motions

An ABM system starts with a market, an account definition and a reason to care now. It gathers signals, ranks accounts, sends the right ones to research and gives a person a clear next action.

The trigger changes. The score changes. The basic chain keeps returning.

A GEO system finds the questions buyers ask, maps the sources models trust, creates useful pages, checks the facts and watches whether the company appears in answers.

One company cares about a technical category. Another cares about a local service. The production and review loop follows the same shape.

An ad system reads performance, learns which creative and audience combinations work, applies budget limits and proposes the next run.

The best versions keep the learning close to the campaign. They remember why a change was made and what happened after it.

A scoring system gathers signals, applies time decay, respects caps, ranks accounts and explains the result.

The score means little until the company defines a good account, a useful signal and a sales action. Once those definitions exist, the machine can run the routine every day.

A research system finds evidence, records the source, checks freshness and turns the findings into an account view.

The useful output follows the decision. A seller preparing a call needs a different view from a marketer choosing a campaign. The research chain stays familiar.

An outreach system chooses a play, drafts a message, checks the company rules, asks for approval where needed and records the action.

The channel can change. The customer can change. The need for evidence, tone, rules and a clean record stays.

The remaining 20 percent makes the system belong to one company.

Its taste decides what good looks like.

Its tone of voice decides how the system speaks.

Its definitions decide what a lead, account, opportunity and customer mean.

Its rules decide who can act, which data can move and where a person must approve.

Its approvers carry responsibility for the choices with real consequences.

Its data connections join the common motion to the actual business.

This company layer changes everything. Two teams can start with the same ABM motion and build very different systems because their market, definitions, risk limits and writing standards guide every decision.

I have seen definitions consume more time than code. Marketing says a lead begins with a form. Sales waits for an accepted meeting. Finance waits for an invoice. The machine forces the company to choose one meaning for each action it takes.

The eleven company cases report the same thing. Every successful deployment solved a data problem first. Scott Brinker frames the same operating question around trust in the data, context and permissions an AI system acts on.

I think the repeated motions should arrive ready to configure.

The engineer wires the motion to the exact stack. The ops lead writes down what the terms mean. The operator sets the outcome and reviews the exceptions.

This removes the blank page. It gives the team known inputs, outputs, tests and failure modes. The local work becomes easier to see.

The wiring still takes care. Data arrives dirty. Definitions expose arguments. Access takes time. A sponsor can favour a visible motion over the one with the best spreadsheet score.

Those choices deserve attention because they make the system useful inside one company.

This is what the GTM teams pulling ahead right now actually look like. No vendor lock-in. LLM-agnostic, so swapping models never breaks the system. Built as they go, shipping a new motion the week they need it. Lean and highly efficient, running lighter than their org chart suggests. Always on the cutting edge, running whatever model and method works best this quarter. That is the bar GTM OS is built to meet.

A direct to consumer nutrition brand in Europe ran an owned search motion that produced 58,800 organic clicks from 4.35 million impressions in ten months. The click rate was 1.4 percent on those inputs. One assistant cited the brand first in three countries. Those clicks and impressions come from a first party search console export in my records.

The same pipeline later ran for a European experiences brand and for a B2B services company in APAC.

The pipeline matched the first build. The vertical changed, the source content changed and the review rules changed. That is the common part and the company part in one example. Page generation, quality checks and citation tracking carried over. Each brand still supplied its own content, tone and review rules.

GTM OS packages the common motions and gives the company layer a place to live. We are in closed alpha. Read the approach at https://www.heyarnoux.com/gtmos/ Request a demo or join the founding group at https://docs.heyarnoux.com/gtm-os-founding/

Tomorrow I will make the strongest case against this approach. Maintenance, security and limited engineering time can change the answer.

David

Tomorrow: Why Building Your GTM Stack Can Be a Terrible Idea

By the way, if you need help building your GTM engine, reach out on the collab page. If you are ready to start building, I have GTM engineers ready to embed in your team.