Revenue Optim
← All articles
GTM

GTM engineering 2.0 is here: AI has moved faster than the sales role

GTM engineering 2.0 is here: AI has moved far ahead of the sales role, and the job now is to redesign the role around it, not to bolt it onto the old one.

By , Partner, RevOps + GTM Engineering

GTM engineering 2.0 is here, and not because a new tool came out. It is here because AI has moved much further than the sales role. Today a model can research an account, summarise a call and draft a decent first message in seconds. And the rep still works, more or less, the way they did five years ago.

That gap is the thesis of this article. Version 1.0 of GTM engineering used AI to do the old job faster. Version 2.0 starts from a different question: if the machine can already do all this, what should the person's job look like?

And a second idea keeps coming up in our conversations with sales teams: fixing one leg is not enough. You have to look at the whole system.

The thesis: AI has moved faster than the sales role

Think about a B2B rep's Monday. They open the CRM to see what is pending. They open the call recorder to remember what the customer said on Thursday. They open the sequencer to see who replied. They open LinkedIn to see who changed jobs. And they open a spreadsheet, because what actually matters to them is in none of the other four.

By mid-morning they have gathered the context. They still haven't spoken to anyone.

Now think about what AI can already do: read all of that at once, join it up and say which account is worth starting with, and why. This is not science fiction. The models, the assistants and the connections between tools already exist. What most companies do not have is a sales role designed to use them.

That is what we call GTM engineering 2.0: stop adding AI to the same old tasks, and redesign sales work around what AI already does well.

What version 1.0 solved (2024-2026)

Version 1.0 was useful, and much of what we build still rests on it. Nearly everything a GTM engineer did in these years fits into ten use cases:

  1. Data and CRM enrichment, from several sources in a waterfall.
  2. Automated inbound: enriching, scoring and routing every form in minutes.
  3. AI outbound: personalised messages built from account data.
  4. TAM sourcing from the ICP, not from a bought database.
  5. PLG workflows, to pass product users who fit over to sales.
  6. Rep-assisted workflows that prepare the work before a call.
  7. Account research and signals: hires, funding rounds, job changes.
  8. Territory planning with data, not by gut feel.
  9. Just-in-time enablement: the right argument when it is needed.
  10. ABM, with marketing and sales on the same accounts.

All of it made tasks faster. But the role stayed the same. The rep got better data and was still the one who pieced it together, made the call and logged in the CRM what had happened.

Why 1.0 is no longer enough

Because automating pieces of an old job gives you more activity, not better decisions.

We see it in three ways. First: enriched data no longer sets anyone apart. When everyone uses the same sources and the same templates, the buyer gets the same "personalised" email from five vendors in the same week. Second: every automation lives in its own corner. The inbound workflow does not know what was said on the last call, and outbound does not know the account opened a support ticket yesterday. Third, and most important: the rep still carries the heaviest part, which is working out what is going on.

Version 1.0 answered "how do I do this task faster?". Version 2.0 answers "what should I do now, with this account, and why?".

What teams tell us: fixing one leg is not enough

In our conversations with B2B companies, the same conclusion keeps coming back: projects that fix a single leg of the commercial system improve that leg and leave the result almost where it was.

There are five legs, and they touch each other: data, process, tools, content and people. Work on only one and the problem moves next door. We have seen it in many forms:

  • An enrichment tool is bought and outbound gets better data, but the CRM still loses track of whoever replies.
  • The CRM is cleaned up and a few months later it is back where it was, because nobody changed the process that made it dirty.
  • An AI assistant is rolled out and the team stops using it within a few weeks, because nobody redesigned their day around it.
  • More content is produced and nobody connects it to the accounts reading it, so it does not start a single conversation.

None of those projects was wrong. They were alone. And this gets worse in version 2.0, because a shared context layer cannot be built from parts that don't talk to each other: either it brings together data, process, tools, content and people, or it is not shared.

That is why GTM engineering 2.0 is, above all, a change of approach. You stop asking "which tool are we missing?" and start asking "how does our commercial system work end to end, and where does it break?".

What GTM engineering 2.0 is: nine pieces, three layers

GTM engineering 2.0 is nine pieces. They are easiest to follow in three layers: what the system knows, how the person uses it, and how it reaches the buyer.

What the system knows

  1. A GTM-wide context layer. One place holding everything known about each account: the company, the conversations, product usage, support tickets, the content they have read, the commercial history. This is where the game will be played, because whoever holds this context decides what every other system does.
  2. The next best play for everything. With that context, the system proposes what to do with each account, in sales, customer success, growth or media. Not one more alert: a recommendation with its reason.
  3. Systems that learn. Every play is stored with its outcome and the next ones adjust. Without piece 1, there is nothing to learn from.
  4. Simulate before you launch. Estimating from history which accounts will respond and which message will work, before burning the market on a test.

How the person uses it

  1. An assistant as the rep's new interface. The rep asks the way they would ask a colleague: "which accounts should I start with today?". The assistant answers with the context already joined up, and whatever screens are needed are generated around the way each person works.
  2. Reps who run their agents. The person stops carrying out every step and starts deciding, reviewing and correcting what their agents do: research, preparation, drafting, follow-up. Each rep becomes, in small, the CRO of their own team of agents.

How it reaches the buyer

  1. Selling to people and to agents. More and more buyers hand the first round of research to an AI assistant. Your website, your documentation and your pricing have to be readable by a machine too.
  2. Content and distribution engineering. Content stops being a separate marketing asset and starts feeding the system: it creates signals that outbound acts on. We cover it in the 2-person GTM team.
  3. New channels: MCP, plugins and agents. Being available where the buyer already works with AI is a distribution channel, the way SEO once was.

And one consequence you can already see coming: after GTM engineering, the next thing many companies will build in-house is their own media and creator teams.

Where to start: context before agents

If AI is already ahead, the temptation is to start at the end: an agent that says what to do every morning. It is a mistake we see often, and it is expensive.

An agent is worth what the context it reads is worth. If the CRM has half its opportunities with no close date, three definitions of "qualified lead" and duplicate accounts, the agent does not recommend better than the rep. It recommends what the CRM has wrong, only with more confidence.

Careful: having an order is not going back to fixing the legs one at a time. The system is designed whole from the start, knowing how each piece fits the others, and then built in steps. The order we suggest is boring, and it works:

  1. A written ICP and a clean data model. If your ICP is partially identified and not documented, start by calculating your real ICP.
  2. The context layer: connecting the sources to a single account model.
  3. The next best play, first as a suggestion the rep accepts or rejects.
  4. Logging outcomes, so the system has something to learn from.
  5. The assistant and the agents, once what sits underneath deserves trust.

The CRM's screen dies, not the CRM

There is one idea in all this we like a lot: that the rep never has to log into the CRM again, or the call tool, or the sequencer. Those screens are a punishment, and making someone spend the afternoon filling in fields wastes their time.

But it pays to be precise. What dies is the screen, not the system of record. Something still has to know which account belongs to whom, what stage each opportunity is at and what has been signed. The CRM keeps doing that. The rep simply uses it through an assistant. Luis develops this idea in the future of CRM with AI.

And here is the paradox. If nobody looks at the CRM directly, the quality of its data matters more than ever, because nobody will notice at a glance that something is wrong. That is why RevOps gains weight in version 2.0: it looks after the data model the agents work on. If you have doubts about yours, go through the signs your CRM needs an audit.

What AI doesn't change

AI has moved faster than the sales role, but it has not answered the old questions for us:

  • Who you sell to. No agent fixes a fuzzy ICP.
  • Why they buy from you. That still comes from talking to customers.
  • Who answers for it. A system that decides needs a person who stands behind what it decides.
  • That there are no shortcuts. More agents do not guarantee more meetings; they guarantee more activity. Pipeline moves with activity aimed at the right accounts.

What does change is where the rep spends their time. Less searching, copying and logging. More thinking, talking and deciding. Which is, when you think about it, the part of the job that was always theirs.

What to do this quarter if you are still on version 1.0

Most of the B2B companies we talk to have not finished version 1.0. That is fine: it is exactly the base version 2.0 needs. What matters is building it knowing where it is going.

  1. Take stock of where context lives. Write down every tool that knows something about your accounts, and what it knows.
  2. Pick one source of truth per data point. One data point, one owner, one system.
  3. Start logging plays and outcomes today. What was done, with which account, what happened. It is the raw material for learning, and almost nobody keeps it.
  4. Try one small "next best play". For example: which inbound accounts to call first each morning, with the reason.
  5. Teach the team to run agents, not just to use them. Reviewing, correcting and deciding is a different skill.

If you don't know where to start, that is what the RevOps and GTM Diagnostic is for: it reviews the commercial process and the system behind it, and tells you in what order to act. And if you want someone to build and run that base inside your own stack, that is what we do as an external GTM Engineer.

Frequently asked questions

What is GTM engineering 2.0?

It is the second stage of GTM engineering. Instead of adding AI to the same old tasks, sales work is redesigned around it: a shared context layer that proposes the next play for each account, learns from the outcomes and is used through assistants and agents.

Why do you say it is already here?

Because the pieces exist: models that research and write, assistants that query several tools at once, connections such as MCP. What most companies lack is not technology. It is a redesigned role, and data put in order so the role can use it.

Why does it take a holistic approach?

Because the commercial system has five legs that touch each other: data, process, tools, content and people. Fixing only one moves the problem next door. Version 2.0 makes this even plainer, because its central piece, the shared context layer, needs all five at once.

Is the CRM going to disappear?

Not as a system of record. What loses weight is its screen: reps will query and update the CRM through an assistant. That is why data quality matters more than before.

Do I need a GTM engineer to get to version 2.0?

You need someone who owns the whole system, not one tool. It can be a GTM engineer, a RevOps profile or both; we set out where each role sits in what RevOps is. And to get the team using AI in their day-to-day work, we offer the AI transformation of the sales team.

Want to go deeper?

Download the RevOps Playbook and get the full framework in writing.

Message us on WhatsApp

Cookies. This site doesn't use analytics, advertising or tracking cookies. We count visits anonymously and in aggregate, without storing anything in your browser. Learn more