Directing human attention in AI-native GTM

I think we’re still underestimating how much of GTM is spent working out what deserves our attention.

I recently spoke to a VP Sales with 80 deals in their pipeline.

To understand where they need to get involved, they open the CRM, scan the fields, look at activity, read notes, check calls and emails, speak to the AE.

They’re reconstructing what’s actually happening.

→ What’s changed?
→ What matters?
→ Where do I need to get involved?

We’ve built an entire GTM stack around storing evidence, then asking humans to make sense of it.

Now we’re adding AI.

ChatGPT, Claude and copilots can make that reconstruction faster. Summarise the calls, analyse the deal, research the account, work out what to do next.

But that’s still the old operating model.

The bigger shift is building GTM around AI from the beginning.

That requires an Operating Layer across the GTM stack:

→ State for what’s happening now across every prospect, deal and customer
→ Memory for the context that needs to persist
→ Skills for how your organisation thinks and works

Now the system can continuously understand the operation.

New evidence arrives → State changes → agents reason over it with your Memory and Skills → the system decides what matters, what it can act on and where it needs human judgement.

So that VP Sales doesn’t need to inspect 80 deals to find the three where their experience matters.

They start with the three.

What changed. Why it matters. What the system recommends.

→ And what judgement it needs from them.

That’s HITL.

Not another alert.

The system is directing scarce human attention towards the decisions where it can change the outcome.

Then it acts.

And what happens next matters too.

The decision, action and outcome create a trace the system can evaluate and learn from.

→ Observe
→ Understand
→ Decide
→ Judge
→ Act
→ Learn

Continuously.

This changes the economics of GTM.

AI makes execution increasingly cheap. Human judgement becomes increasingly valuable.

So you stop spending expensive human capacity searching, inspecting and reconstructing the operation.

A VP Sales can apply their judgement across 80 deals without reviewing 80 deals.

An AE can operate across more accounts without understanding less about each one.

RevOps can encode how the organisation works rather than continuously enforcing it through process and meetings.

And the system keeps operating between the moments humans traditionally catch up.

That’s the shift from AI-assisted to AI-native GTM.

Not AI helping humans reconstruct the operation faster.

A system continuously understanding, deciding, acting and learning, with humans brought in where their judgement has the most value.

The more work it does, the better it should become at doing the work.

That’s the Operating Layer for AI-native GTM we’ve built at Revenue Labs.

🔔 Interested in AI-native GTM? Follow along.
🧱Building with AI and Agents in GTM? DM me.

Originally posted on LinkedIn

‍