AI-First Marketing Generalist & Agent Manager

Company: Harmony Reports to: Gadi BenMark, CMO Team: You, plus a workforce of AI marketing agents (paid media, copy, research, orchestration). Location: New York City. Strong preference for in-office 4 days a week. Level / comp: Flexible, calibrated to the person. We'll scope seniority and compensation to the right builder rather than force-fit a band.

The one-line version

We sell AI agents. We run our marketing on them too. We need the human who commands that agent workforce, owns the output, and moves fast enough that the CMO can stay on strategy.

Why this role exists

Harmony is an AI-native company. Our marketing should be the proof.

Today the CMO directs the agents, runs paid, builds the decks, writes the keynotes, and ships the banners. That does not scale. This role takes the operating layer off the CMO's plate: you run paid media across every channel, you produce the brand and content output AI-first, and you make the agents themselves better over time.

This is not a MarTech or Marketing Operations role. That function is owned. This role owns creative, content, and paid growth, executed through AI agents.

What you'll own

1. Email marketing, run agentically

• Own the whole email program:

  • Launch and product announcements

  • Webinar and event invites

  • Nurture and follow-up sequences

  • Newsletters

• Agents draft, segment, and quality-check the work.

• You set the strategy, judge the output, and own the results.

• Run a real test discipline: judge the subject line on opens and the body on clicks, never in the same split.

• Own deliverability and list health:

  • Sender reputation

  • Domain rotation

  • Inbox placement

• Report in pipeline and meetings, not just opens.

2. Content syndication

• Own the program end to end:

  • Vendor selection

  • Lead specifications and ideal customer profile (ICP) filters

  • The assets that go out

• Build the follow-up engine that turns a syndicated lead into a conversation, with every email carrying the bridge from the source to Harmony.

• Judge every vendor and every asset on cost per completed meeting, not lead volume.

3. Conversion rate optimization (CRO) across the website and landing pages

• Own conversion on the website and on every campaign landing page, including the ones Tal sends paid traffic to.

• Run a continuous test program: hypothesis, build, ship, read, repeat. Agents generate and quality-check the variants.

• Own the conversion mechanics:

  • Forms and field counts

  • Calls to action

  • Page structure and proof placement

• Report what moved the number and what did not.

4. Website content, built for search engine optimization (SEO), answer engine optimization (AEO), and generative engine optimization (GEO)

• Create and refresh the website content:

  • Product and platform pages

  • Use cases and guides

  • Comparison pages

  • Knowledge articles and the blog

• Write for three readers:

  • The buyer

  • The search engine

  • The AI answer engine that quotes us

• Track where Harmony shows up in AI answers for the questions our buyers ask, and feed what you learn back into the pages.

• Work from the keyword and topic plan already in place, and extend it with the buyer language from our sales calls.

5. Advanced paid channels

• Launch and run connected TV (CTV) as the first advanced channel, then other channels beyond search and social as the data supports.

• Own the full test:

  • Audience and budget

  • Creative, including agent-generated video

  • Measurement, including whether the channel adds results we would not have gotten anyway

• Start small, prove it, then scale or kill it.

6. Build and improve the agents

• Build new agents, not just run them:

  • Instructions and skills

  • Workflows

  • Connections to our tools

• Brief agents the way you would brief a sharp teammate: what, when, and how, not line-by-line prompts.

• Share what you build with Tal, and take what he builds, so the whole system compounds.

• Close the loop: capture what worked and promote it into the system.

7. Work directly and closely with the CMO

• Own planning and execution end to end.

• Report crisply, in numbers:

  • What shipped

  • What is working

  • What you would change

• Bring decisions to the CMO, not tasks.

The work you'll hand off to agents (a starter set)

Real jobs you would delegate and direct in your first 90 days. Each is written so a stranger could check it is done, which is exactly how you will brief the agents.

• An email engine: a flagship nurture rebuilt as segmented flavors, each with two subject lines and two bodies, drafted by agents from real buyer language in our sales calls, with a test plan attached.

• A weekly email health readout: how the program is performing from first open to pipeline, ending in one recommended action.

• A syndication scorecard: every vendor ranked on cost per completed meeting, plus a follow-up sequence where each email carries the source-to-Harmony bridge.

• A ranked CRO test backlog: built from analytics and on-page behavior, with the first tests live inside 30 days.

• Landing pages run through a "copy tournament": multiple versions judged by adversarial agents before anything ships.

• A website refresh pass: top pages rewritten with a clear definition in the first 50 words and a comparison table, so they work for search and for AI answers.

• A monthly AI-answer visibility check: the questions our buyers ask, run across the major AI answer engines, with where we appear and where we do not.

• A CTV test plan: one page that says who we reach, what we spend, and when we kill it. Then launched.

• A quarterly "kill Harmony" red-team: an agent playing our top competitor's CMO, hunting the holes in our website positioning.

• One funnel view: first touch to pipeline across every channel you own, so each one is judged on the same scoreboard.

The operating method behind all of it: you brief work as a goal with a checkable "Done when..." bar, not a vague task. Vague briefs loop forever. Checkable ones get finished.

Who you are

• You've built things on the internet. Side projects, a newsletter, a viral experiment, a micro-tool, a channel you grew. You make things because you can't not.

• You hand off whole jobs, not just tasks. The bar is not "AI helps me work faster." It's that you delegate entire jobs to agents that overdeliver, then direct and judge the output. You manage AI like a manager, not a user.

• You're a true generalist. Comfortable being part-strategist, part-copywriter, part-designer, part-analyst in the same afternoon. You'd rather own a channel end-to-end than hand off between silos.

• You're fluent in paid. You've run real budgets across more than one channel and can read a performance report and act on it.

• You get inside the head of our buyers. You can think like a stretched IT or HR leader, feel their pains and pressures, and use that to judge what will actually land before you spend a dollar or ship a slide. Customer-centric thinking is how you separate the agent output that works from the output that only looks good.

• You have taste. You know the difference between AI output and good AI output, and you close that gap by hand.

• You move at hacker speed. You ship the MVP, measure, and iterate. You don't wait for permission or perfect.

• You write well. Clear, concise, no fluff. You can make a slide, a banner, and a LinkedIn post all sound like us.

Nice to have

• B2B SaaS or developer / IT-audience experience.

• Hands-on with no-code and automation tools (Zapier, Make, n8n, Retool, or similar) and with LLM tooling.

• Design sensibility in a deck tool and a basic creative tool.

• Comfortable behind a camera and in a video-editing tool, and willing to travel to customer sites for shoots.

• You've stood up a paid channel from zero.

How we'll evaluate you (the work sample beats the resume)

• Show us something you built and grew yourself. The scrappier and more self-directed, the better.

• Show us a marketing agent, workflow, or automation you set up, and what it now does without you.

• A short live exercise: take one Harmony asset (a deck slide, an ad, a banner) and make it better with agents in front of us. We care how you direct the work, not just the result. Bonus if you frame it as a goal with a checkable "Done when..." bar.