Four builds from practicing marketers. Real tools, real outcomes, specific steps.
By Chetan Parmar · April 1, 2026 · 9 min read
In March 2026, Emily Kramer - ex-VP Marketing at Asana and founder of MKT1 Newsletter - published a roundup of real builds from four working marketers who had shipped tools using Claude Code. Not demos. Not prototypes. Production tools they use in their actual jobs.
The builds got significant attention because they answered a question the community was actively debating: can non-technical marketers actually ship things with Claude Code, or is this only for people with coding backgrounds who are using AI to go faster?
The answer from the four builds is unambiguous: yes. Here is what they built and how.
Key Takeaways
The outcome
Monday morning reporting that previously took 4 hours now runs automatically before anyone opens their laptop. Zero manual work required.
A performance marketing director at a mid-size B2B SaaS company built a pipeline that pulls data from Google Ads, Meta, and LinkedIn simultaneously every Monday at 6am. It calculates cross-channel CPL by source, flags any campaigns where CPL rose more than 25% week over week, and exports a formatted summary to Slack.
The build used Claude Code with MCP connections to each platform's API. The marketer described the pipeline behavior in plain English across three sessions: first the data architecture, then the anomaly detection logic, then the Slack formatting. Total build time: one afternoon plus one follow-up session to handle edge cases. For context on what Google Ads automation looks like end-to-end, the playbook goes deeper on the API setup.
Starting with a clear description of the output format before building the data pipeline. When the marketer could describe exactly what the Slack message should contain, Claude Code could work backwards to figure out what data to pull and how to calculate it.
The outcome
A dedicated campaign landing page built and deployed in one afternoon. No designer, no developer, no ticket.
A growth marketer at an ecommerce brand needed a dedicated landing page for a seasonal campaign. Their design team had a two-week backlog. Their developer had a three-week backlog. The campaign was launching in five days.
They described the page layout to Claude Code: hero section with a countdown timer, three product features with icons, a testimonials row, and a checkout CTA. Claude Code built the Next.js page, added responsive styling, and included a simple countdown component. Total time from first prompt to deployed preview: under 3 hours.
The page converted at 4.2%, above their typical 3.1% for campaign pages - partly because it was built specifically for the campaign rather than adapted from a generic template.
The outcome
Real-time competitive intelligence that previously required a $400/month tool. Now running on their own infrastructure at near-zero marginal cost.
An SEO lead built a Claude Code + Puppeteer scraper that monitors five competitor websites for changes to their pricing pages, feature comparison pages, and homepage messaging. It runs every 48 hours, diffs the content against the previous version, and sends a Slack summary of meaningful changes.
This kind of real-time competitive monitoring previously required expensive dedicated tools or a manual audit schedule that always slipped. The Claude Code version was built over two sessions and has been running reliably for months.
Always check a website's terms of service before building scrapers. Many sites explicitly prohibit automated access. The builds that hold up over time scrape public marketing pages (pricing, features, homepage) rather than internal data, and include reasonable rate limiting to avoid server impact.
The outcome
500 Responsive Search Ad variations generated from a product catalog in a single session. Previously a 2-week copywriter project.
A Google Ads manager at an agency used Claude Code to read product feature data from a client's catalog and generate 10 RSA headline variations and 4 description variations per product. The catalog had 50 products. That is 500 headlines and 200 descriptions - a task that would typically take a copywriter 10-12 days.
The session took about 2 hours. The manager fed Claude Code the product catalog CSV, provided brand voice guidelines, and specified character limits. Claude Code generated all variations and exported them in a format ready for Google Ads Editor bulk upload.
The manager reviewed a sample of 30 ads before approving the batch. Quality was high enough that only minor edits were needed on about 8% of the output.
Looking across the four builds, three patterns emerge that separated successful projects from the ones that stalled:
Every successful build started with "I want X to look like this" or "I want this CSV to contain these columns." Vague briefs like "automate my reporting" produced vague results. Specific output descriptions produced specific tools.
None of these were built in a single session. Data architecture, then logic, then output formatting, then edge cases. Breaking the build into focused sessions where each session has a clear deliverable produced better results than trying to describe the entire system upfront.
The bulk ad copy generator reviewed 30 samples before approving 500. The competitive scraper ran for a week on one competitor before expanding to five. Every build included a review step before the system ran at full scale.
Emily Kramer documented four builds from practicing marketers in March 2026: an automated cross-channel performance pipeline replacing 4 hours of Monday reporting, a campaign landing page built and deployed in under 3 hours, a competitive intelligence scraper monitoring 5 competitor sites, and 500 RSA variations generated from a 50-product catalog in 2 hours.
No. All four builders documented by Emily Kramer were marketing practitioners, not developers. The performance pipeline builder was a B2B SaaS marketing director. The landing page builder was a growth marketer at an ecommerce brand. The common pattern was starting with a clear description of the desired output, not a technical spec.
The documented builds ranged from one afternoon to two focused sessions. The ad performance pipeline took one afternoon plus a follow-up session for edge cases. The landing page took under 3 hours. The RSA generator took about 2 hours. Complex tools like the competitive scraper took two sessions across a week.
Three patterns across all four builds: starting with a specific output description rather than a vague goal, building in sessions with one deliverable per session rather than one giant prompt, and reviewing a sample before scaling (30 RSA samples before approving 500, one competitor before expanding to five).
Pick one of the four build types above. Install Claude Code and spend one afternoon on it. The first build is always the hardest. The second one takes half the time.
Install Claude Code