The ROI of Agentic GTM

Everyone knows AI makes developers 10x more effective. Is the same true for GTM or is it just creating AI-powered SDR slop cannons?

Let's dive into the ROI of building agents to run your go-to-market and why you should invest. I first wrote about this three years ago and the tech has improved dramatically since then.

AI agents in GTM focus on two different types of things. There's a foundation, your business intelligence layer. Then on top of that foundation, you build line-of-business agents that augment tasks your humans already do.

I think about these agents in two categories: excellence agents and efficiency agents. The mental models for each are different, and it's worth getting them straight before you try to put a number on any of this.

Efficiency agents produce artifacts people would have already created, but at a faster clip.

Excellence agents do incremental work that your best reps are doing but your average reps aren't. It raises the floor. Now everyone is executing best practices, all the time.

What if we had perfect enablement?

That's the question excellence agents answer.

What if every rep had perfect enablement skills, 100% of the time, that was 100% effective? In-the-moment guidance on every deal, every call, every follow-up. The kind of guidance your best reps have internalized but your average reps forgot two days after onboarding.

Organizations with strong enablement functions achieve 49% higher win rates on forecasted deals. But "strong enablement" has always been "inconsistently applied enablement." Your best reps internalize it and use it but your average reps can't find the Notion doc. Excellence agents make everyone perform like your best reps.

The practical example: post-call follow-up generated immediately after the call with 100% accuracy of what was stated, evaluated against your sales framework, with coached next steps. Normally, deal reviews happen once a week if you're lucky. Maybe you catch up in a one-on-one with your manager. So deal advice runs at a once-a-week cadence. Imagine that advice being constant, right after every call. That helps you book the next meeting faster and compresses deal cycles.

Gartner predicts AI-driven enablement will deliver 40% faster sales stage velocity than traditional methods by 2029. I'd argue that number is conservative if you're building this yourself rather than buying off-the-shelf tooling.

What if this person was superhuman?

That's the efficiency agent question.

What if you had a superhuman version of this person who could handle 10x or 100x the throughput? I wrote about this when we built Vercel's first GTM agent. We went from 10 inbound SDRs to 1 and saved $2M+. The agent made the remaining human's output 3x by automating the tedious parts (research, qualification, email drafting) and letting the rep focus on the high-value work.

These two categories stack. Excellence agents improve win rates through better execution. Efficiency agents improve throughput. Combined, you're moving more deals through the pipe, faster, with better outcomes.

The intelligence layer is different

The foundation underneath both agent types is the intelligence layer, and it solves a different problem. It gets product insights and sales gaps from the field to the right place in the business, faster.

This is about iteration velocity. On sales plays. On product feedback. Your team actually builds the right thing because customer signals aren't trapped in call recordings nobody watches or CRM fields nobody reads. The intelligence layer connects your engineers with the right stakeholders internally and the right customers externally.

I've written about the knowledge base problem before. It's one of the hardest challenges in this whole stack. But when you solve it, you create a context layer that makes everything else work.

The I in ROI keeps shrinking

I've been talking about the return side. The investment side is changing just as fast.

There's been a 100x price decrease between GPT-4.5 and GPT-5-nano at the same performance level. LLM API costs have dropped at a median rate of 50x per year, accelerating to 200x per year since January 2024. The model you need for post-call summaries and lead qualification is effectively free.

The engineering costs are real but manageable. Our first lead agent at Vercel cost about $60K/year in engineering hours and API costs. It saved $2M+. That's 33x ROI on something we built in a weekend. And the tooling has only gotten better since. GTM use cases are greenfield and AI handles them well.

We've got a narrow window before everyone figures this out. The organizations adopting AI-powered GTM now are the ones your sales team will be competing against next quarter.

Best to get some of these early returns before they're competed away.

The stuff that doesn't show up in a spreadsheet

There's some ancillary benefits worth mentioning.

Employees want to work at companies that are AI-forward. BCG's research shows that when leaders demonstrate strong support for AI, the share of employees who feel positive about their careers rises from 15% to 55%. People get into sales because they love people. They don't love updating the CRM and digging through case studies to find the right one and sitting through enablement training. They do these things because they get paid. AI can take away the hardest, least-fulfilling parts of the role while leaving the parts people love.

There's another benefit to living on the edge. Your product probably has some AI functionality in it. By leaning into AI in your own domain, you understand it better, sell it better. And honestly? It's just plain fun. Sales has always been hard work, but massive technology changes can bring a lot of wonder and joy to jobs.

How to measure it

It's tough to model out all of these gains. You're raising the bar of what the default is, improving efficiency, and humans remain fundamentally required for all of it.

AI is speeding up your deal cycles. It is improving win rates via perfect enablement. It is speeding up your business. These gains show up in so many different ways that it's impossible to attribute all of it as it all stacks together.

That's why you want to look at actual business metrics.

My approach:

• study function specific metrics (stage progression, win rates, reply rates, no shows, etc.)
• study GTM-wide metrics

For GTM-wide, I like something like GEM (GTM Efficiency Metric, GEM ≈ Net New ARR ÷ Total GTM Spend). Pick a metric where there's no where to hide.

Yes, it will be hard to decouple AI's impact from other changes you make in your business, the product, and the market. But AI is such a powerful technology that regardless of those things, this will show up in the numbers. If the numbers aren't improving, you're not doing your job.

Let's go ship.