Finoa
Role: Founder's Associate, CEO Office · 2025
Building the content infrastructure for a company that had no marketing team — from idea capture through to published post.
- Content output, same headcount
- 3xContent output, same headcount
- Less time per post
- 90%+Less time per post
- Team capacity freed per year
- 180+ hrsTeam capacity freed per year
- Total system cost per year
- €780Total system cost per year
Context
Finoa is a BaFin-regulated digital asset custodian in Berlin, serving pension funds, asset managers and banks alongside crypto-native clients. During my internship there was no dedicated marketing team, no blog on the Finoa Consensus Services side of the business, no first-page Google ranking and no presence in AI-generated search answers. Content happened ad hoc or not at all.
Challenge
Turn the ambition to be a visible thought leader into a system a lean team could actually run — without hiring a marketing function to run it.
Three months, no marketing team, no budget to hire one, and no prior experience with the tooling — I learned n8n from scratch to build this.
Approach
- 01Sat in twice-weekly growth meetings to find where marketing actually broke down: manual coordination, no infrastructure, and no blog on the FCS side at all.
- 02Launched the FCS blog and defined its three content categories — news, announcements and research — so readers could navigate it and the automation had a schema to key off.
- 03Designed the Marketing OS as four modular n8n workflows — topic collection, draft generation, review, publishing — each able to run and be debugged independently.
- 04Learned n8n from scratch, tracked the build in Jira sprints, and demoed progress to the founders every other day.
Decisions
n8n over a closed automation platform
Open source, granular enough to build custom logic, and it integrated with the Google stack the company already ran on. It could be extended later without a rebuild.
Google Sheets as the front end, not a custom interface
The team already worked in Sheets every day. Adoption mattered more than polish, and building UI would have spent the three months on the wrong problem.
Human review kept in the loop, not full automation
Finoa is a regulated custodian. Compliance and brand voice are not things to hand to a model unsupervised, so every draft passes a person before it publishes.
Firecrawl MCP tested, then left out
It produced richer source material for drafts, but connections were unstable and retrieval incomplete. Documented as a roadmap item rather than shipped into something the team would depend on.
System
Prototype walkthrough
One idea in. Three drafts out. A person approves before anything publishes.
- 01
Collect
Someone adds an idea, or a scraper finds one and scores how relevant it is. Approved ideas get tagged.
- 02
Write
The tags pick the prompt. Blog, LinkedIn and Twitter drafts come out, each in the right voice.
- 03
Review
Drafts go to a Google Doc. Reviewers comment, and the system acts on the comments.
- 04
Publish
Approved posts get scheduled. The dashboard shows what's where.
Everything carries an ID, so any post traces back to the idea it came from.
Outcome
Delivered a working prototype in three months and handed it over with documentation. In its first months the FCS blog launched and became the second most viewed page on the site, and posts started generating discussion on social channels — early signals rather than a full year of data. The system runs on €780 a year.
- Marketing OS on n8n — four modular workflows
- FCS blog launch and content structure
- Metadata-keyed prompt library
- Comment-driven review loop
- AI tooling evaluation — Base44, Firecrawl MCP, DataForSEO
What's next
- Pipe DataForSEO keyword data into the pipeline so topic selection follows search demand rather than instinct.
- Revisit Firecrawl MCP once it stabilises, for deeper source material.
- Push approved posts straight to the CMS instead of scheduling them by hand.
Work
1 images — click to enlarge
Next