The GTM Launchpad is where we open the hood on B2B revenue - fixing structure, sharpening outbound and building pipeline that actually holds up.

This week I gave two completely opposite pieces of advice.

In my own pipeline, I told an AI worker to stop creating follow-up work because the buyer had already told me what was happening next.

Legal was moving, finance had what it needed and another touch from me would not have created more leverage. It would have created more noise.

A few hours later, I was coaching an SDR and told him almost the exact opposite:
stop waiting.

He had been waiting on someone else to deliver a list he needed and the question I kept coming back to was simple: what can you own next?

Same week, same belief in systems, two completely different decisions.

That contradiction is what stuck with me.

The best AI sales system is not the one that creates the most activity. It is the one that can tell the difference between something that needs action and something that needs restraint.

If we are going to keep handing more of the revenue workflow to machines I think we need a much better brake pedal.

Earlier this week, I had one active opportunity where the buyer had already told me exactly what was happening. There was a defined next event and the work had moved to the other side of the table. In another situation, the agreement and invoice were already in the right hands. Nothing was stalled just because nothing new had happened in 48 hours.

The normal sales-software instinct is easy to imagine: the deal has not moved, so create a task.

No response, create another task.

Three days elapsed, send a follow-up.

Seven days elapsed, send another one.

But elapsed time was not the problem here I already knew the next step and another email would have made the system busier without making the deal better.

At the same time, I had older opportunities sitting in the CRM where there was no agreed next event, no useful reason for the silence and no evidence that anything was moving. Those did deserve attention.

The system pulled together the history, surfaced the relationships, prepared the reactivation drafts and put them in front of me. I reviewed them, made sure they felt right and sent them.

OPEN OPPORTUNITY ≠ OPEN TASK

A deal can be open because something is actively happening on the other side: legal, finance, procurement, internal approval or a promised date. If the buyer already gave you the next step, sometimes sales discipline means respecting it.

But there is an important distinction here because WAIT can turn into an excuse really fast and I ran into the opposite problem while coaching that SDR.

We had talked about a webinar list the week before and by the time we met again he still had not received it. He was waiting but in that situation waiting was not restraint. It was a lack of ownership. So we worked through a better question: What can you do next without waiting for somebody to hand you the answer?

No phone number? Find another route. Dialer down? Shift channels. Missing one list? Work the accounts already in front of you while you escalate what is blocked.

That helped me define WAIT more tightly. A legitimate WAIT state needs three things: a known dependency, an owner and a wake condition.

Legal has the agreement and said Friday? Great, wake the opportunity up Friday afternoon if nothing changes.

Finance has the invoice and gave you a processing window? Respect the window.

You just do not know what to do next and you are hoping somebody else handles it? That is not WAIT. That is a problem.

The more I build AI into my own operating system, the more I keep coming back to three states: PREPARE, HUMAN and WAIT.

I do not think every revenue workflow can be reduced to three labels forever and I am not trying to turn this into another cute AI framework. I like the labels because they force a better question than “Can we automate this?” They make me ask what the machine should actually own, where a person still creates value and when the smartest action is to deliberately hold.

This is where I want AI to be aggressive. Research the account, check the CRM history, read the relevant email thread, map the buying committee, summarize what changed, build the account hypothesis, draft the email, prepare the call brief and organize the meeting notes. That is the work around the work and it eats a ridiculous amount of time when a human has to reconstruct the entire relationship from scratch every time.

On a client planning call this week, I said something out loud that probably captures the whole PREPARE state better than anything else: I do not want the system living in my head, your head, Slack, Trellus, all this stuff. That is the machine's job.

Pull the pieces together, keep the memory, preserve what happened, organize the playbook, score the calls and prepare the operator. A new rep walking in three weeks from now should not need six months of tribal knowledge just to understand how the motion works.

That does not mean the machine owns the decision. It means the machine earns the right to make the human faster by doing the preparation cleanly.

Preparation is valuable precisely because it creates more room for judgment when judgment is actually needed.

There are still moments where the human is the value: the call, the warm reply, the coaching conversation, the pricing decision, the scope change, the interview, the awkward objection and the relationship that needs a little pressure or a little patience.

I saw this clearly in coaching this week. We are building more research, summaries, scoring and preparation around the seller while at the exact same time I am pushing him to sound less robotic once another human picks up the phone.

The point is not to make the seller memorize a more complicated script. It is to give him enough context that he can listen to what was actually said, use the knowledge he already has and ask the next question because it makes sense in the conversation.

Could software eventually execute more of those moments? Of course! But “Can AI do this?” is becoming a less useful question than “Does removing the human improve the outcome?”

For now, the operating rule I keep coming back to is simple

system prepares
system surfaces
human decides
human executes

We can loosen that boundary later when a decision is repeatable, low-risk and proven.

Maximum autonomy does not need to be the starting point just because the software is capable of it.

This is the state I think most teams are missing.

Most automation systems are built around elapsed time because elapsed time is easy to measure: three days pass, do X; seven days pass, do Y; fourteen days pass, create a red task and make somebody feel behind.

The problem is that time passing does not automatically make the next action useful.

The better question is

What condition needs to become true before another action is worth taking?

That changes WAIT from “forget about it” into an intentional state.

The opportunity still has context, the dependency is known, someone owns what is supposed to happen and the system knows exactly what should wake it back up.

That is a very different design from a workflow that panics every time a clock moves. It also creates accountability because intentional waiting is visible.

You can tell the difference between “we are waiting on procurement until Friday” and “nobody knows what is happening here.”

One is a state. The other is a leak.

I found myself using the same idea with a founder this week while we were building his first outbound motion.

We could have gone straight into buying more domains, spinning up more mailboxes and pushing thousands of emails.

Technically, the tooling makes that easy now. The harder part was admitting that we had not earned that volume yet.

So we deliberately kept the first motion small.

Get the data source right, get a few mailboxes warming, start with a narrow group of accounts and have five to ten legitimate conversations with the exact people the product was built for. Learn what they care about, see what gets a reply and pay attention to what falls flat before adding more volume.

If the signal is there, amplify it.

If it is not, fix the ICP, message or offer before giving the mistake more distribution.

That is a brake pedal too and I think it matters more now because the cost of creating scale has fallen so dramatically.

The goal is not to slow the business down.

It is to stop the business from scaling the wrong thing faster.

The more powerful the accelerator becomes, the more expensive bad judgment becomes at scale.

This is the bigger reason I care about all of this. AI is making activity cheap

Research is cheaper, personalization is cheaper, drafting is cheaper and coordination is cheaper.

Soon, almost every revenue team will be able to manufacture more touches, more content, more follow-ups and more “personalized” outreach than they know what to do with.

The cost of a bad action is not the API call. It is the buyer's attention, the trust you burn, the relationship you make noisier and the rep spending Tuesday morning chasing the wrong five things because the system told them everything was urgent.

Cheap output does not make attention cheap.

So yes, I want the accelerator. I want AI doing the research, organizing the mess, preparing the work and making the team faster. I also want a brake pedal.

A good revenue system should make the right work easier to do but it should also make the wrong work harder to do.

I think we may eventually measure that too. Not only emails generated or hours saved but bad actions avoided:

duplicate touches prevented, DNCs caught, existing customers excluded from prospecting, warm relationships surfaced before buying another net-new contact, follow-ups intentionally held because the buyer already gave you the next step, campaigns kept small because the message had not earned scale yet.

That tells me a lot more about the quality of an AI system than raw output does.

Take one recurring revenue workflow and run it through these three states.

PREPARE means the machine can research, summarize, prioritize, draft or organize without creating meaningful risk.

HUMAN means judgment, relationship, commitment or context matters enough that a person should still own the decision or action.

WAIT means the next useful action depends on a buyer, a date, an approval or another external condition and the dependency, owner and wake condition are explicit.

Then inspect the workflow instead of the technology.

What can the machine prepare without risk?
Where should human approval remain on purpose?
What exact condition should wake a waiting item back up?

You do not need an autonomous sales org by Friday; build one workflow where the machine does more of the work around the human moment so the human can be better when the moment arrives.

That is what I am testing inside my own business right now and this week made the principle a lot clearer for me.

Automate the preparation
Protect the judgment
Respect the wait

That is the system I want: not automation for the sake of movement but a revenue motion that knows when to prepare, when to hand control to a human and when doing nothing is the most disciplined action available.

What is one revenue action you would never let an AI system take without a human reviewing it first?

Hit reply and tell me where you are drawing that line.

I am genuinely curious because I think those boundaries are going to become part of how strong GTM teams are designed.

Thanks for reading. See you again next week at 8 am MST!

- Tom Slocum ✌️

P.S. If your outbound motion is busy but fragile, that is the work I do inside
The SD Lab

Know a founder or sales leader who would get value from this field note?
Forward it their way