Nick Lafferty showed four things he built as Profound's founding marketing engineer. LinkedIn-to-pipeline attribution. A Google Ads agent on a cron job. AI-generated ads from a Figma MCP connection. A personal website that self-updates from AI search citation data. The title is made up. The output is real.
| Metric | Value |
|---|---|
| LinkedIn attribution tool build time | ~1 hour (airport lounge) |
| Meeting booked rate improvement (Default form) | 3x after switching to Default |
| LinkedIn UTM-tracked leads | 20% click directly |
| Self-reported LinkedIn leads | ~80% (open text form field) |
| AI search citations (personal site) | ~500K |
| Figma ad templates | 5 (swappable text) |
| Research deprecation window | 180 days |
| GitHub repo tenure | 8 years |
What to take away
- The title is invented. The role is real. Google's former AI general Marvin posted "Marketing engineers are the hire of 2026" on LinkedIn, unprompted. You can invent a job title if you're persuasive enough and the output backs you up.
- An hour in the airport lounge. The LinkedIn-to-pipeline attribution tool took Nick one hour to build flying back from Europe. He also bought Default on day two of their webinar and 3x'd his meeting booked rate from the form. That's the operating tempo.
- Human in the loop is still the right call. Nick's Google Ads agent monitors campaigns, surfaces recommendations, and delivers Slack buttons for one-click action. It does not touch the campaigns autonomously. He's explicit about this being a deliberate choice, not a limitation.
- MCP turns design tools into pipelines. By connecting Claude Code to Figma via MCP, Nick can generate production-ready LinkedIn ads from Figma templates without waiting for a designer. He demoed this live: ask Claude for Grateful Dead-themed ads, get five brand-consistent versions in seconds.
- Your website can update itself from AI search data. Nick's personal site uses a weekly GitHub Actions run to pull AI citation data from Profound's API, update his stats, deprecate research older than 180 days, and keep content current — without him writing anything.
- AI search changes the attribution model. The highest-quality leads are people who search your brand directly in ChatGPT, Gemini, or Google AI Mode. Being cited in AI responses is a new distribution channel, and it compounds with domain authority.
The role: what a marketing engineer actually is
Nick opened by defining the thing he invented.
"I'm a marketing engineer. We kind of just made this up, to be honest with y'all. Turns out you can just invent a job title and people just go with it if you're persuasive enough."
He's a paid ads marketer by training. He's had a personal GitHub repo for eight years. He runs his personal site off GitHub and Cloudflare Pages. When Profound needed someone to build marketing systems, he was the natural fit.
"Of everyone on my team, I was kind of the most natural person to adopt AI tools and AI systems."
Google's former AI general Marvin validated the concept publicly, posting unprompted on LinkedIn: "Marketing engineers are the hire of 2026."
That's the external signal. Internally, the definition is simpler: someone who builds AI agents to eliminate manual work for the marketing team, and can operate in the same environment as an engineer — terminal, GitHub, APIs.
Thing 1: LinkedIn-to-pipeline attribution (built in one hour)
Profound's growth loop is LinkedIn → leads → pipeline → revenue. Post on LinkedIn, get inbound. The problem was there was no way to close the loop between social performance and actual pipeline.
Nick connected LinkedIn and Twitter social metrics with HubSpot data in Claude Code, visualized when posts pop off and what happens to leads, and identified top contributors (the CEO, the CTO on Twitter, himself saying he didn't want to be a head of marketing — 133,000 impressions).
"This took me like an hour in the airport lounge when I was flying back from Europe last week."
The practical attribution answer was more interesting than the build: 20% of leads click directly from LinkedIn posts with UTMs. The other 80%? They fill out the sales form and self-report. Nick's Default form has a required open-text "How'd you hear about us?" field with a 5-character minimum. People type "LinkedIn."
"I force you to fill this out. This is a required, five-character minimum limit open text and you tell me LinkedIn."
Low-tech, but it works because he controls the form. He also runs Default for the webinar form. Bought it on day two, 3x'd the meeting booked rate.
Thing 2: Google Ads agent on GitHub Actions cron
Nick runs a marketing agency for Profound's Google Ads. He also wants to replace them. Mostly for the ego of the LinkedIn post, he admits.
The agent he built monitors campaigns on a schedule: search terms, new keywords, ad copy, geo and device splits, schedules, competitive intel. Everything gets packaged into a Slack message with action buttons. Nick reviews it and clicks the actions he approves.
"None of this takes action on my behalf in the platform. Mostly because I just don't trust it to fully have like a fully automated loop. I still do deeply believe in a human-in-the-loop element."
That constraint is worth noting. The agent is doing all the analysis and packaging the judgment. The human is doing the approval. Not because Claude can't take the action — because Nick doesn't want to outsource that call yet.
Thing 3: LinkedIn ad generation via Figma MCP
Profound's designers build product, not marketing assets. Nick has five Figma templates with swappable text. He connected Claude Code to Figma via MCP and built a workflow that generates brand-accurate ads on demand.
He demoed it live: asked Claude to write LinkedIn ad copy using lyrics from the Grateful Dead's "Althea," then generate the ads from the Figma templates. The ads appeared in the terminal, brand-consistent, ready to run.
"You can just make ads as long as you have the visual template. Everything else — CTA, text — you can change in Claude Code."
He won't run the Grateful Dead ads. But the scale unlock is real. One designer builds the template once. The marketing engineer scales output from it.
Thing 4: The self-updating website
This one runs in the background indefinitely.
He runs a personal website on Hugo (static site generator) hosted on GitHub + Cloudflare Pages. He uses it as a testing ground: validate things here, then deploy to the Profound site.
He wanted to surface how often his website and pages got cited in AI search results. Profound's platform tracks AI citations across ChatGPT, Gemini, Perplexity, and Google AI Mode. He connected a GitHub Actions cron that:
- Pings the Profound API weekly
- Pulls citation counts for his pages
- Updates his website stats automatically
But the deeper piece is the research freshness agent. Profound publishes research on which domains are most cited in AI search — billions of data points. That research lives in a Google Slides deck.
"I gave this to Claude and I was like, hey, I have this publicly sharable URL, can you turn this into markdown?"
Claude extracted the data from the live Google Slides URL, converted it to a markdown file in his repo. Now a GitHub Actions agent runs weekly, syncs the Profound research, and deprecates anything older than 180 days.
"Research gets old and stale in my industry very quickly. I don't want to cite a piece of research from a year ago because it's quite literally very out of date and often incorrect."
That markdown file is referenced by other agents that update and create content on his website. The site updates itself in the background.
On AI search and domain authority
Nick's personal site has around 500K AI search citations. Someone asked how.
"Every tech company I work at, I build a backlink to my website. Make a domain, write blogs for your company, and then just link back to your website and do that for 10 years and you get really good domain authority."
Loom. Profound. Every employer, a backlink. Author bios, inline links when relevant.
The AI search point: brands cited in AI responses are winning on the channel that matters most right now. Profound measures and improves that for companies. Nick's personal site is his own experiment in what it looks like for an individual.
What this means
If you want to understand what a marketing engineer actually does day-to-day, Nick's four builds are the clearest case study available. One hour builds a LinkedIn attribution tool. A weekend wires your website to AI search data. A Figma MCP connection turns five templates into unlimited ad variants. All of it runs on GitHub Actions crons — no dashboards, no manual work.
If you're thinking about AI search (AEO) for your brand, Nick's personal site with 500K citations demonstrates the compounding effect of domain authority plus AI citation coverage. Brands cited in AI responses are winning the distribution channel that's now beating organic search for high-intent queries.
If you're running Google Ads and want to move faster, the human-in-the-loop agent model is the right architecture: let Claude handle the analysis, surface Slack buttons for approval, and don't automate the final action until you trust the judgment. Nick chose this deliberately, not as a limitation.
Nick Lafferty on LinkedIn · @LaffertyN on X · nicklafferty.com · Profound
Watch the full GTM as Code event · Nandika Jhunjhunwala on account scoring for GTM agents · Hunter Rosenblume on RFP automation · Kathleen Booth on building competitive intelligence without engineers
Build marketing systems that compound over time
Deepline gives Claude Code the GTM data layer it needs. Enrich, score, personalize — from the terminal, on a schedule, without clicking through dashboards.