← All posts

What 3,821 Job Postings Say About How GTM Teams Are Told to Use AI

September 29, 2026 · Marketing

I wanted to better understand how companies are asking GTM teams to use AI. There’s all this talk from every corner of GTM about AI. Some good. Some not so good.

I’ve talked with people in the trenches, but my network, like all networks, can be a bubble sometimes. My first instinct was to do a survey. Again, network bubble risk. So I explored other avenues, one of which was reviewing job postings to see how companies are mentioning AI for various GTM roles. What are they asking sales and marketing folks to do with AI?

I pulled 3,821 marketing and sales postings from 352 tech companies with public job boards between September 24 and 26, 2026. I sorted every sentence that mentions AI by what it asks of the person in the role: nothing at all, selling AI, being open to it, doing existing work faster, building something new, or governing what AI produces. The full dataset is at the bottom if you want to dive into the weeds or do your own analysis.

What job postings ask each role to do with AI

Each posting counted once, at the most it asks of the role. Click any part of a bar to read what those postings say.

Loading the data…

Here are some findings that caught my attention.

1. Most mention AI. Few ask the role to use it.

I would have thought more than half of these postings would ask the person in the role to actually use AI, given how much GTM talks about it and all the push for AI-forward candidates. It’s 24%.

78% mention AI somewhere, and most of those mentions are about the company.

What the posting does with AI Share of postings
Asks the person in the role to use it 24%
Only describes the company or its product 27%
Mentions AI only as the thing the role sells or needs to understand 18%
Says everyone at the company uses AI, then asks nothing of this role 9%
Mentions AI only in a hiring-process notice 2%
Doesn’t mention AI 20%

2. Plenty of generating, not much governing

130 of the 352 companies ask at least one marketing or sales role to generate customer-facing material with AI. Stuff like drafts, copy, decks, proposals and personalized outreach. One demand gen posting spelled out the whole list: “Use AI to accelerate research, campaign briefs, landing page drafts, creative variants, webinar ops, nurture, reporting, and analysis.”

26 of those 130 companies ask anyone to govern what AI produces. 7 give that job to product marketing.

Across all the postings it runs about six to one. 207 ask someone to generate with AI. 34 ask someone to govern it.

3. Marketers get asked. Sellers mostly don’t.

37% of marketing postings name a specific AI task. For account executives it’s 8%. SDRs come in at 13% and presales at 18%.

Compare inside each company instead of across the pool and marketing still comes out about 30 points ahead.

When a sales posting does name a task, it’s vague like this one: “Use AI tools to speed up account research, prospecting, and deal preparation.”

4. Transformation vs acceleration

Of the marketing postings that name an AI task, 53% ask the person to build something or change how the work gets done. That covers workflows, agents and whole functions redesigned around AI. Product marketing sits at 48%. Marketing ops is at 78%.

On the sales side it’s about one in five. The other four ask for the same work, faster.

The rebuilt version of sales work barely shows up. 4 of 2,818 sales-side postings mention using AI on a proposal. None of them mention building from product marketing’s material.

5. Slop prevention as an individual responsibility vs system

65 sentences call out the failure directly. Eve: “We have a hard line against AI slop, and we’re looking for someone whose instincts are audience-first to keep content valuable.” Bureau wants copy “without enterprise jargon or AI slop.” D3 put it in a heading: “Clear, fast writing. Short sentences, no AI slop filler.”

114 postings ask anyone to judge the quality of AI output. 67 of them put it on the individual checking their own work. 47 ask for a standard or a system other people work from.

The postings that get it right

Ten product marketing postings ask the PMM to own what AI works from. They’re worth reading if you’re writing one this quarter.

  • Listen Labs: “A single source of truth: Build the AI tooling our go-to-market team needs to self-serve the most up-to-date assets, messaging, and positioning.”
  • Intercom: “maintain the core ‘brain’ of PMM context that feeds our AI-first approach”
  • IonQ: “Own and continuously evolve Space Solutions AI-enabled go-to-market systems, including domain-specific agents, knowledge repositories, competitive intelligence, product information, proof points, ICPs, and approved messaging”
  • Serval: “agents that draft cheatsheets, certification scripts, and training modules from product specs, Gong insights, and your messaging frameworks, with you setting the standards”
  • RapidSOS: “Develop and evangelize AI workflows, prompt libraries, and content templates that expand what the broader PMM and GTM team can produce independently”
  • ZoomInfo: “Design AI-powered workflows, structured product intake, reusable templates, and scalable frameworks that help PMs consistently produce high-quality messaging, release communications, and launch assets.”

Brand postings ask for it more often, 8% against 5% for product marketing. Mercury wants someone to “build AI-ready brand guidelines the team can design against.” Gusto wants someone who “sets the creative standard for what AI output is and isn’t acceptable for the Gusto brand.”

One sales-side posting made the list. Superhuman’s AI GTM role has to “Set the standards, guardrails, and reporting that keep AI-generated outreach compliant, on-brand, and measurable.”

My thoughts

At this point, it’s too early to claim that sales and marketing talent has to be “AI-native” (whatever that means). From job postings, we’re a long way away from companies having clear standards and expectations for how those roles should use AI. Not the first time that the discourse is ahead of the reality, and probably not the last.

Does this change how I will continue to approach AI myself? Nope. If anything it signals an even bigger opportunity to rethink how GTM works. Thinking outside the box is what got me into tech marketing. The box isn’t there anymore, and that’s what excites me the most.

Methodology info

The frame. 352 technology companies with public job boards on Greenhouse, Lever or Ashby. I left out companies that build AI models to avoid skewing the results. The companies came from three places: a list of security vendors, established software companies found through Wikipedia and careers pages, and companies found by searching for open marketing roles. That last group over-represents companies hiring marketers. Rates differ by source, so these numbers describe these 352 companies, not tech as a whole.

The roles. Job families come from titles. Audits of random title samples tightened the rules until misfiled titles fell to a few percent. Growth engineers, pricing strategists, and finance and HR roles with “GTM” in the title came out.

The coding. The categories came from an exploration sample of 165 postings, and I locked them before analyzing the rest. I used AI to build the collection scripts and to analyze every AI-related sentence against those categories, one sentence at a time. No number above includes the exploration sample.

The audits. Two of them failed, and both got fixed.

  • A full re-read of every “build or redesign” code found that about one in five single-sentence cases were really “use AI to be more efficient.” Each one got recoded.
  • The first filter for “generating customer-facing material” was 55% accurate. It counted “insight generation” and “generate pipeline.” The rebuilt version checks out at about 90%.
  • A blind check of marketing and sales sentences, with the job title hidden, found the original coding leaned slightly toward marketing. It’s worth about 2 points on the sales comparison, which is why I call the gap “about 30 points.”
  • Checks for duplicate, empty and stale postings and for company boilerplate each moved the headline numbers by a point or less.

Every correction is in the dataset, with the reason.

The limits. A job posting is what an employer wrote, not the actual job being done. No second rater checked the coding independently. It’s one snapshot, so it says nothing about trends.

Download the data

The postings themselves belong to the employers, so the files carry each posting’s link and every sentence the analysis uses. The coding is CC BY 4.0. Check my work. Tell me what you find.