Growth / AI-Enabled Marketer

Synthesized from ten live postings at Anthropic, Vercel, Databricks, Robinhood, Chime, Duolingo, Notion, Klaviyo, Gusto, and Webflow, read from employer careers boards on August 9, 2026. Not any one company's job description; the center of mass.

You are no longer measured on the campaigns you ran. You are measured on the efficiency of a system that runs them, on whether the machine's output was good enough to ship, and on whether you can defend the arithmetic when the measurement is contested.

You work with data science and analytics as first-class partners (nine of ten postings), with engineering and marketing ops on tracking and build-versus-buy, with creative and content where your role shifted from briefing to running the gate, and with finance and executives via written, decision-ready recommendations. Seven of ten ask for 5+ years; Klaviyo accepts 3 to 6 traded against a hard requirement that you have run LLM tooling in production. Nine of ten state no degree requirement.

Ten shifts show up across the postings. The first two are the reason this ladder exists.

Ad platforms are named by brand in two of ten postings. Marketing automation suites in one. The SEO toolchain in one. Three postings name no tools at all. What employers name is arithmetic (CAC, payback, LTV, incrementality) and behaviors (built agents, killed a test that failed its criterion, made the case in an imperfect measurement environment).

So no rubric here awards points for a platform. We score the payback math with the gross-profit figure intact, the experiment with its success criterion stated before the data, the gate that would have caught the bad content, and the case you make to a CFO.

• Anthropic

  $255K to $320K

  SEO Lead: the highest-paying posting in the sample, and it names AI answer engines as a discovery surface.

• Vercel

  $150K to $220K

  Three growth reqs open at once. “Have built agents, automations, or workflows that changed how you operate.” “Ability to scale taste with AI.”

  Tier 1

• Databricks

  No range posted

  Director, Generative Engine Optimization: grow share of citations across priority AI answer engines. A net-new function.

• Robinhood

  $150K to $225K

  Staff Performance Marketing Manager owns the martech roadmap, traditional-versus-agentic decisions, and guardrails so agentic systems behave predictably.

  Tier 1

• Chime

  $150K to $208K

  Senior Lifecycle Marketing Manager; two lifecycle reqs open at once; analytics stack named.

• Duolingo

  $163.2K to $244.8K

  Director of Growth Marketing Strategy: the widest band in the sample and a classical JD. AI is mentioned once, as a screening disclosure.

  Tier 2

• Notion

  $165K to $190K

  B2B Growth Marketing Manager; four growth roles across seven postings; excellent classical JD with no AI expectation.

  Tier 2

• Klaviyo

  $124K to $186K

  Lead GTM Data Operations Analyst, AI Workflows: 3 to 6 years accepted, traded against running LLM tooling in production. Confidence scoring, versioned prompts, QA sampling gates.

  Tier 1

• Gusto

  $108.6K to $161K

  Senior PMM, GTM Experimentation: “making the case for paid social investment in imperfect attribution environments.” The lowest top-of-band in the sample.

• Webflow

  $131.5K to $180K

  Agent-driven campaign orchestration; the required-qualifications block is a list of behaviors: think in systems, not campaigns.

  Tier 1

Provn is not affiliated with any employer listed. Descriptions summarize each company's public job posting as read in August 2026.

1. Practice20 min total

   Find the Broken Step

   Wickfield's funnel, six steps. Find the one that is actually broken, and do not confuse it with the one that looks worst.

   Isolates: Can you look at a funnel and find the step that is actually broken, and not confuse it with the step that looks worst? 3 to 4 min video.

2. Easy28 min total

   Where the Next $300K Goes

   Five channels, current CAC and payback for each, $300K to allocate. Decide with the payback arithmetic intact and say what you would cut.

   Isolates: Can you allocate money across channels with the payback arithmetic intact, and say what you'd cut? 4 to 5 min video.

3. Medium40 min total

   The Content Engine Is Winning and Losing

   An agent pipeline publishing sixty pages a week; traffic up, conversion down. Diagnose the machine and design the gate that would have caught it.

   Isolates: Can you diagnose a machine that is producing more and converting less, and design the gate that would have caught it? 5 to 6 min video.

4. Hard40 min total

   Twelve-Month Payback

   The board wants twelve-month payback across all channels. Rebuild Wickfield's growth economics against the target and defend the arithmetic to a CFO and a skeptic.

   Isolates: Can you rebuild a company's growth economics against a hard target and defend the arithmetic to a CFO and a skeptic? 6 to 8 min video.

Every tier is scored on the same five dimensions. AI leverage and automation is weighted 25 at every tier, higher than on most hubs, because half of the postings make building with AI a hard requirement.

What Transformative looks like

On Where the Next $300K Goes, Transformative allocates with payback computed per channel on the $107.64 gross-profit figure, states which number it does not trust and why (last-touch on paid social, for example), cuts a channel and names what that costs, and sequences the spend so the riskiest assumption is tested first. Adoptive allocates with the arithmetic shown and names a cut; that is the advance bar. Capable allocates proportionally to current volume and calls it a plan. Below Capable does the arithmetic on gross profit but cannot commit to a cut or name a channel to defund. Not Yet allocates on revenue instead of gross profit.

Communication is judged on whether the person you are explaining to could act on your video alone. Never on accent, pace, or filler words.

1. Paid social looks like our best channel on last-touch. Should we double it?

   Not on last-touch alone. Say what incrementality evidence you would get (a geo holdout, a spend-down test), what payback looks like on gross profit rather than revenue, and what you would do while the test runs. The strong answer names the number it does not trust.

   Where it comes from: Gusto: imperfect attribution environments. Seven of ten name attribution and incrementality.

2. The content pipeline publishes sixty pages a week. Traffic is up 40% and trial conversion is down. What happened?

   The machine is optimizing for the thing you measured. Diagnose which pages converted before and which do not now, what changed in the generation step, and design the gate: what is checked automatically, what a human samples, and the kill criterion. The strong answer says what it would stop publishing this week.

   Where it comes from: Klaviyo: QA sampling gates, versioned prompts. Robinhood: where human judgment stays in the loop.

3. Describe an experiment you killed.

   Name the success criterion you set before the data came in, the result, and what you concluded when it was ambiguous. Generic answers about testing culture score poorly.

   Where it comes from: Nine of ten name experiment design with pre-stated criteria; the recurring word is rigorous.

4. How would you get Wickfield cited in ChatGPT and Perplexity answers about small-business bookkeeping?

   Answer-first pages that resolve a real question, extractable structure, entity consistency, and brand mentions off-site, then measure share of citations directly rather than rankings. The strong answer treats it as a channel with its own economics.

   Where it comes from: Databricks: Director, GEO. Anthropic and Webflow name AI answer engines as surfaces.

5. Tell me about an agent or automation you built that changed how a team operates.

   Name it, say what it replaced, how you gated its output, and where it failed. This is also the shape of the mandatory AI question in your video.

   Where it comes from: Vercel: have built agents, automations, or workflows that changed how you operate. Five of ten require it.