AI Will Reshape GTM in Four Phases. Most Teams Haven’t Even Entered Phase 1.

Over the last decade, I’ve watched hundreds of go-to-market teams wrestle with the same challenge: when to scale, how fast, and how to do so effectively. That work ultimately became The Science of Scaling, a set of frameworks that guide those decisions with more rigor.

Now, AI is disrupting go-to-market in all kinds of ways that we couldn’t have predicted ten years ago. Some founders are operating with seed-level resources, raising a single round, stretching dollars, and focusing on efficiency; others are backed by $100M+ rounds and are experimenting at scale. For most founders, the reality is somewhere in between the penny pinching and the hype. 

But, what holds true regardless — even amidst the hype and the massive funding rounds — are the fundamentals that guide smart decision-making, which is what The Science of Scaling is all about.     

The fundamentals still stand because they’re grounded in principles like product-market fit and customer retention. Even if it’s AI sellers selling to AI buyers, those definitions don’t change. What does change is how we operationalize those fundamentals in practice, and that shift is happening fast.

The way forward is to move past dabbling and into the first real phase of value creation. I think it’s helpful to think of this as a framework: four phases for how AI will reshape GTM.

It’s tempting (and fun) to speculate about Phases 3 and 4. But the reality is that most teams haven’t even entered Phase 1, which is what we are going to focus on in this article. 

For as long as I’ve been in sales, we’ve obsessed over one metric: selling time.

What percentage of a rep’s week is actually spent in front of the customer? Most teams are at 20–25%. A great team might break 40%. Almost no one gets to 50. Phase 1 is about changing that equation. With the right agent ecosystem, seller time can approach 80–90%.

And getting there isn’t theoretical. It’s something any company can implement today, with agents built on top of GPTs. Here’s how it can play out across the customer journey.

Ideal Customer Profile Definition. An ICP was always meant to be dynamic, shifting with market changes, competition, and product evolution. But the reality is that retooling an ICP has been expensive, time-consuming, and painful. Now, an AI ICP agent can constantly analyze everything from micro-segments where retention and unit economics are strong to identifying new opportunity areas and guiding GTM process enhancements. 

This is huge, because it means that product-market fit and go-to-market fit can be pursued in parallel. Another implication is that ICP expansion now becomes granular. Instead of saying, “We now sell to enterprise,” the ICP agent might add “273 accounts” to the ICP with little human understanding of the rationale that drove it. 

This constant analysis also means that the first sales hire won’t have to masquerade as a part-time product manager. The agent surfaces insight; the salesperson focuses entirely on buyer interactions.

Account Selection. Once ICP is defined, the next question is: which accounts do we go after? Today, that decision often sits with a very green SDR.

Think about that: you’re about to drop $5 million into GTM next quarter, and the first call on where to place that bet comes from someone with a few months of experience, scrolling through a prospecting tool.

That’s a really expensive decision to get wrong. With an account-selection agent tied to ICP and capacity data, the list updates dynamically: not “50 accounts for the quarter,” but the 17 accounts that should be in play this week.

Persona Mapping & Outreach. Now we drill into the buying committee. Who’s the champion? Who owns the budget? Who’s the security gatekeeper? An agent can surface that roster, pull signals on each individual, and help shape outreach. At Stage 2, we’ve documented real-world builds of these kinds of agents. For example, one agent dynamically selects accounts, while another powers a modern outbound workflow. These aren’t theoretical sketches; they’re working systems GTM teams are running today.

Meeting Prep & Live Support. Say you’ve got a first meeting tomorrow with Mary, the Director of Marketing at Aventra. An AI agent trained on your sales methodology, battlecards, and product data builds you a playbook — tailored to Mary and Aventra. Then it role-plays the meeting with you, acting as “Mary,” pushing back in the ways she’s likely to. That’s way better than waiting for your manager to squeeze in a coaching session. And when you walk into the real meeting, the agent is there with you, listening, surfacing the right questions, suggesting the right case study.

Post-Call & Coaching. After the meeting, the agent doesn’t just draft your follow-up email. It updates the CRM, the forecast, the deal room, the rep scorecard. And it coaches you. “Great job, Asad. But you’re still a 3.9 on urgency development. Next time you’ve got a 7-minute Uber ride, here’s the video to watch.” That’s what enablement should have been all along: continuous, contextual, specific. We even published an example of this approach in practice, building a scalable GTM playbook using off-the-shelf GPTs (MedScout case study here).

Hiring Feedback Loops. Finally, think about recruiting. A good CRO already looks back every six months: which hires ramped fastest, which ones flamed out, and did we see those signals in the interview process? AI can help run that loop, connecting performance back to hiring criteria, and over time, get better than us at predicting who will succeed in our context.

This is the leverage in Phase 1. It’s not about replacing reps. It’s about reclaiming their time for the conversations that matter.

Leaders jump in with demos or tack AI onto an existing workflow, and then wonder why nothing changes.

If you want to avoid that outcome, a few principles matter:

AI is empowering dramatic changes in workflows, which a minority of companies are transforming into a huge competitive advantage: reps that spend 75% of the week in front of customers are going to be three times more productive than reps spending the traditional 25%.  

It’s important to acknowledge that, as with most new tools, productivity leaps won’t be felt immediately the first time a company experiments with AI. The first time a rep automates an email, they’ll probably feel like it slows them down. The first time ops uses AI to tune the forecast, they’ll probably feel like they could have done it better and faster themselves. In the 2000s, Google was famous for its 20% time policy, which allowed employees to dedicate 20% of their time to personal projects in order to drive company innovation and personal happiness. Companies could adopt a similar approach in order to incentivize employees to experiment with AI, without the burden of achieving positive ROI immediately. 

The key is for companies to commit and encourage employees to learn, iterate, and work toward the massive step function that AI can enable today.

Become a Topline insider by joining our Slack channel.

This Made Us Think: Curated by Kyler Verney.

We want to hear your feedback! Let us know your thoughts.