Why does AI in GTM need executive leadership?

Speed on the old job is worth funding, and your builders should keep shipping it. Some changes are different in kind: a task moves from one team to another, a target changes, or the buyer sees a different process.

Your builders often know which of those changes would matter. What they can’t do is rewrite another team’s targets, budget or comp.

GTM product thinking, running GTM as a product, puts three seats around the shared harness, the setup your team uses to work with AI. Only one of them, the executive, can decide the changes neither functional leader can authorize.

Say your BDRs drive event registration and their targets count it. With a shared harness, anyone who knows a guest can invite them, tracked the way a BDR program tracks it. Your sales and marketing leaders can agree it’s a good idea in one meeting. It stalls after that, because it rewrites a target one of them owns and a number the other answers for. If that target feeds a comp plan neither leader can change alone, the call belongs to the executive who can.

We’d pick meetings held over pipeline from attendees. Meetings sit closer to the BDRs’ own work and show up within weeks of the event. Pipeline is the better number and the worse target for this team: it often lands a quarter later and depends on the AE. Either way someone gives something up. The BDR team trades a target it knows how to hit for one that depends on whether guests show up, and marketing’s event number now depends on people outside marketing.

We run our own invites this way. For Sculpt, Clay’s conference in San Francisco on October 8 (our meetings run October 7 to 9), anyone on our team invites from our harness, and every request is scored and logged to HubSpot. As of October 3, 2026, the calendar send still waited on one production setting. We’re three people, so the decision took one conversation and changed no target. If your registration target feeds comp, yours also needs a comp decision.

Good functional leaders trade all the time, and most of a change should stay with them. A call goes up when it needs authority neither leader holds, or when they’ve tried and can’t agree. Apply that test to five parts of every change:

• Which work moves between teams.
• Whose targets change, and what replaces them.
• Whose budget and whose time pay for the change.
• Whether and when your teams get tools to build with. Which tool, inside your security and spending rules, can stay with them.
• When it lands. Changing a target mid-period is costly and burns trust, so most teams change it at a plan boundary.

Our founder, Sam Gong, learned the tools decision late. As SVP Marketing at a software company, he gave his product team AI building tools a month before an offsite. They should have had them three months earlier: three months lost to a decision only he could make.

His marketing team had bought four AI point tools without deciding what work each was for, and pulled the purchases back after a board meeting. The tools weren’t the problem. Nobody had defined the work.

AI made drafting cheap. It didn’t make deciding cheap. When your teams can build a change in a week, a decision that waits a quarter is the bottleneck.

We failed at this first. In an earlier attempt, 52 of the 55 decisions waiting on our founder stayed open, because they sat as a list in a file and he decides in conversation. Your stalled calls may have the same cause: decisions waiting where nobody has to answer them.

Now every decision carries a recommendation, a default and a date. If the date passes, the default lands, the work moves on, and the log says “landed by default”, not “decided”. Each decision says what it would cost to reverse.

It worked better, not fully. Of the 24 decisions logged from September 21 to 27, 2026, 19 got an answer, most of them in one conversation. Five passed their default date, and the log still shows them open, not landed by default. A default date that nothing enforces is a reminder, so the clock has to belong to someone.

Before a default lands on your team, agree which calls may land that way: reversible, internal, with a named owner and a review date. A target, spending, a price, anything a customer sees and anything legal gets an explicit yes from the person who owns it.

At full strength, which we haven’t reached ourselves, the executive staff meeting opens with that log: what landed by default, what needs an explicit yes, and what’s past its date.