Go-to-Market in the AI Age: What Changes, What Doesn't

By Michael Gaudet — Benevity employee #63 (through a $1B+ exit), 20 fractional-CMO engagements delivered, Co.Labs Executive in Residence, Haskayne MBA.

Most of the founders I talk to are quietly anxious that AI has rewritten go-to-market and they're already behind. The truth is the opposite of scary: the fundamentals haven't moved an inch. What changed is the how — and the how gets rewritten roughly every six months, which is exactly why it's the wrong thing to anchor to.

The three things venture investors actually want to see — your ICP and buyer journey, your KPIs and measurement, your marketing strategy and execution — haven't changed in years, and an AI revolution hasn't moved them. They won't have moved four years from now either. So this piece is built the way the work is built: on the layer that holds. I'll be honest about the how, because it's useful, but treat it as the part that decays.

Has AI actually changed go-to-market?

The honest answer is two-part: how revenue gets made has changed completely, while the principles behind it haven't moved an inch.

On the how side, it's a different world. AI agents now sit on the front line in outbound and BDR work, and increasingly in client success. Data enrichment and automation run through everything — meeting-recording AI, CRM augmentation, outreach drafted at scale. It's a whole new alchemy of people, skills, time, and tooling, and it's genuinely transformative.

But the principles underneath are untouched. Knowing your customer still decides whether you win. Measuring inputs against outcomes is still how you improve. A revenue-centric culture still wins both the short and the long game. A lot changed in the how; nothing changed in the what. Hold onto that, because everything else in this piece hangs off it.

Do you still need an ICP in the age of AI?

Yes — and arguably more than ever, because AI amplifies whatever you point it at. Understanding your customer is never finished, and the best marketing comes from them, not from you: who bought, why they bought, how they experienced the value, asked again and again. It's about what they think is compelling, not what you think is compelling. If you haven't talked to your customers lately, you haven't nailed it — most founders are sure they have, and most haven't.

Here's where AI earns its keep: getting the right data is the whole game, and that's exactly where AI helps most. The ICP doc and the plan are the easy part. The hard part — the listening, the synthesis — is where the leverage now lives.

And here's the trap. I call it the howitzer problem. AI amplifies your aim. Point it at the wrong clients — the unprofitable ones, the high-churn ones, the ones who beat up your CS team — and you'll just get very good at winning more of them. Garbage in, garbage out — you just win more of the wrong clients. Aiming matters as much as the tool.

How do you know you're actually post-product-market-fit?

You're post-PMF only when three things are simultaneously true — and plenty of teams who think they're past this aren't.

First, three to five clients have hit outsized value: NPS 9–10, they'd pay 100% more, and they'd refer you. Second, those clients are commercially viable — profitable, not a resource drain, and your client-success team doesn't hate them. Third, you've scaled acquisition of clients like them in at least one repeatable way. All three, or you're not there yet.

Why this matters: where you are on the PMF journey decides what to do now, and where you're going decides whether you survive. Pre-PMF, the job is to prove outsized value with a meaningful number of viable clients. At PMF, it's to build a repeatable way to acquire more of your ICP. At GTM-fit, acquisition scales and the job becomes growing and defending it. Misread your stage and you'll automate a motion you haven't earned the right to scale.

How do you build a moat when AI can clone your product?

You name your moat or you build one — because AI now lets a point solution get gobbled up in minutes, and "death by a thousand cuts" is happening to the big incumbent platforms in real time. The question to sit with: what stops two smart people in a basement in San Francisco from rebuilding your product this weekend?

There's a clean way to find out. It's a test I learned from Raif "Rafe" Barbaros, Partner at Mistral Venture Partners — the red-team test. Task your sharpest developers with rebuilding your product over a weekend using AI. It does two things at once: it shows you your real exposure, and it surfaces your next product leap — sometimes a full replatform. Run it on yourself before the basement does.

As for what actually holds, the moats fall into a few types. Deep, outsized value: if a client would pay 100% more in a second, you're sticky — be indispensable. Hardware: a physical component two devs can't spin up over a weekend (this used to be a valuation discount; now it's a moat). And licensing or regulation: a legal or regulatory lock on the category. If you have no product moat, distribution can be your moat — so 10× it. And think in multiples, not 15–30% increments. Incremental thinking is loser talk when you're competing against everyone who has access to the same tools you do.

Does measurement still matter when AI runs the motion?

It matters as much as it ever did — what gets measured gets managed. A CRM married to cost data tells you the ROI of every revenue play. You forecast and measure inputs against outcomes so you improve, deliberately, every cycle. Skip it and AI just helps you do the wrong thing faster, and more confidently.

What's changed is the capture, not the principle. I almost never open my CRM anymore — Claude plus a set of connectors keeps it current and sends me a daily diff of what moved. I spend my time having conversations and writing, not living in a pipeline view. But the CRM still earns its keep: it structures the data so the AI can actually reach and leverage it. I'm not a believer that the CRM goes away. Whether any one specific platform survives is a fair question — but the structured layer underneath doesn't. You just don't live in it anymore. Automate the capture and the measurement gets better, not just faster.

How do you map a GTM motion for the AI age?

You break your motion into its parts and decide, stage by stage, who owns it, who executes, what technology runs it, and how that technology is actually used. The work didn't disappear when AI arrived — it slid to the right, toward software, automation, and AI. Every task sits somewhere on a spectrum: human only (judgment, relationships, the first discovery call), tech-enabled human (a person leaning on software), human-in-the-loop (AI drafts and acts, a human reviews and approves), and fully automated (software runs it end-to-end). Here's how that maps across a real motion:

Prioritize. Not everything matters equally at your stage — automate the high-volume, top-of-funnel work first. Contract scalability can wait when you're early.

Take one row all the way down — lead outreach — and you see what this looks like in practice. A B2B SaaS team I work with runs Clay, Smartlead.ai, and an LLM to produce roughly 1,600 personalized drafts a day, opening outbound as a real channel off the side of one person's desk. But "human-in-the-loop" here is not a cursory glance. The human has to know what great looks like (the taste) and prompt for it (the skill) — both, or you just get mediocre at scale. You don't set the campaign up anymore; you have to be able to QA it.

None of this works without the data underneath it, and the best data is your own calls. I have two non-negotiables when I start with a client: the whole revenue team on one AI/LLM tool — the same one — and every client call recorded. "Our clients won't agree to be recorded" is the usual objection; I was selling to Schedule-1 banks in fintech and still got roughly 80% of calls recorded. 80% beats 0%. Pay the $15 per seat per month — you're a tech company, it'd be hypocritical not to use the tools. The payoff: one recorded call automatically drafts the follow-up email and the proposal, creates the next-step tasks, and updates and moves the deal in your CRM. All of it, off one recording.

What marketing roles do you need now?

You don't need five specialists. You need one T-shaped operator.

The old model was a paid-search specialist, a paid-social specialist, an analytics specialist, a content specialist, and an ops/automation specialist. I used to advocate for exactly that — there's video of me poo-pooing generalists. Today the shape is one AI-augmented operator who is T-shaped and tool-obsessed: knows enough to QA every channel, lives and breathes the tooling, and drives adoption across the team. The role isn't "knows how to set up the Google Ads campaign." It's "knows what a campaign, an ad group, and a keyword are, and can QA what the AI builds." The speedboat, not the tanker.

The math is the part founders underrate: one operator at roughly 80% efficacy across five areas beats a single specialist at 100% in one — call it 400% more output. The one caveat: if a single channel is make-or-break for you — events, or paid, say — keep the deep vertical of the T planted right there.

Where do you start Monday morning?

Where you start depends on where you are — and the rule underneath both paths is the same: figure out if it works, then automate for scale. Don't automate a guess.

If you're pre-PMF, lock the three fundamentals before you automate anything. Get three to five clients to outsized value, interview them, and draft the ICP — AI helps, but you decide. Stand up the simplest measurement that proves what's working. And automate nothing you haven't first done by hand.

If you're post-PMF, map your motion stage by stage on the template above. For each task, decide its spot on the spectrum, from human to automated. Get the motion right first, then automate it to scale. And hire a T-shaped GTM/RevOps operator who owns the tooling and knows what great looks like.

FAQ

Michael Gaudet is the founder of Eighty Twenty CMO, a fractional CMO practice for venture-backed B2B SaaS companies. He was employee #63 at Benevity through its $1B+ exit, has delivered 20 full 12-week engagements, serves as Executive in Residence at Co.Labs, and holds an MBA from the Haskayne School of Business.