From near-invisible on LinkedIn to 5,279 impressions in 46 days

Walma builds AI document tools for municipalities, housing companies and enterprises. The product was strong, but the company was close to invisible online. In the 46 days before this period, its LinkedIn page got about 90 impressions, so a buyer who looked Walma up found very little. It mattered because Walma sells trust. In a market full of AI hype, a company that wants to be taken seriously by public-sector and enterprise buyers has to show up where those buyers look.

I built a content strategy around five themes that Walma's buyers care about: the EU AI Act, AI ROI, agentic AI, data sovereignty and AI literacy. Every theme connected to something Walma was actually building or advising on, so the content was specific rather than generic. I wrote posts for the company page and for the leadership team's profiles, and optimized headlines, wording and hashtags for how people search inside LinkedIn. I also ran an SEO audit of Walma's website, with recommendations on schema markup and crawlability.

Five content pillars, one per theme above, each written for the questions a municipality, housing company or enterprise buyer would ask before choosing an AI vendor. Posts for both the company page and the leadership profiles, so the same message reached buyers from two directions. Website content for walma.ai, plus the EU AI Act and data-sovereignty material that fed the LinkedIn work.

For a couple of months I was Walma's main AI engineer, working directly with the CTO at a team of five. I shipped features and fixes to Noda, a multi-tenant AI assistant that answers questions about internal documents with citations, and to Ocle, the processing engine behind it. The work included an invoice reviewer for a Swedish municipality that checks supplier invoices against contracts, a fix for product mix-ups in a support assistant for a security hardware manufacturer, a SQL agent with database connections, document generation to PDF, PowerPoint and Word, and a skills system for the agent. I also moved the Ocle workers to Azure AI Foundry behind an LLM-provider abstraction. Walma had no documentation, so I wrote the first onboarding guides, a testing and development guide and READMEs for four repositories, and set the rule that docs are updated with every feature change.