AI Marketing
Independent consultant for AI marketing and AI agency engagements. I build automated campaign optimization, generative content at scale, and RAG-based segmentation — wired into your CRM, ad platforms, and content stack. No creative briefs that die in Trello, no AI pilots without measurable results.
AI marketing isn't pasting a ChatGPT prompt into a blog post template, or having an AI agency sell you an "intelligent dashboard" no one opens. It's wiring LLMs, automation, and your own data so that campaigns are optimized, content is produced, and leads are routed faster than a manual team can manage. My practice focuses on what's measurable: lower CPA, higher conversion rates, more relevant content per SEK of ad spend. Stack: GPT-4 / Claude for content, n8n and Make for automation, Meta Ads + Google Ads APIs for bidding, Postgres + pgvector for RAG segmentation, GA4 + GTM for measurement. Everything runs in the EU with traceable logs — GDPR and EU AI Act compliance is built in, not bolted on.
The marketing team adjusts bids, audiences, and creatives by hand across Meta, Google, and LinkedIn. Tired teams miss patterns in the data. Fix: an AI-driven optimization loop that reads GA4 + platform APIs, identifies underperforming ads, and auto-adjusts bids or pauses them — escalating to a human for larger budget changes.
You know what you should be publishing (Swedish SEO articles, LinkedIn posts, email sequences) but never get to it. An LLM pipeline with RAG against your product docs and tone of voice produces drafts your marketer polishes — instead of writing from scratch. 5–10× faster with quality intact.
Inbound leads from forms, demos, and event registrations land in the same email queue. AI routing classifies lead size, intent, and industry, automatically drops them in the right nurture sequence in HubSpot/Mailchimp, and pings sales when a hot lead arrives.
You have years of customer data in CRM, transaction history, and support tickets — but ads run on simple demographic segments. RAG-based segmentation finds patterns (customer journey, vertical niches, churn risk) and builds audiences that actually convert.
Before automating anything: what are you already measuring, what's missing, where does the data live? I review GA4, GTM, CRM, ad accounts, and content stack. Output: a clear map of data points, measurement gaps, and where AI has the highest leverage. Fixed price.
We pick one concrete problem (campaign optimization, content, lead routing, or segmentation) and build the pilot. All with evals — no black boxes hallucinating expensive ad decisions. You test it in parallel with your current process.
Pilot moves to production with monitoring, alerting, and a runbook you own. Then we extend to the next use case. Self-hosted runtime means data and prompts don't live inside an agency tool you don't control.
Default: n8n self-hosted for automation and routing, OpenAI/Claude for LLM calls, Postgres + pgvector for RAG segmentation, GA4 + GTM for measurement, Meta Ads API + Google Ads API for bidding. For Microsoft-heavy orgs: Power Automate + Azure OpenAI. For HubSpot-based marketing teams: Operations Hub + custom webhooks. Everything runs in the EU. Model choice depends on data sensitivity — for GDPR-critical customer data: Azure OpenAI EU residency or Claude on AWS Bedrock EU.
An AI agency typically sells packages: a dashboard, a monthly report, an "AI strategy" that looks great in a deck. I build systems that run in production when I'm done. One person, senior delivery, code you own. Trade-off: I won't take five large clients at once and I don't sell off-the-shelf "AI packages" — every engagement is scoped against your problem.
Measurement and data mapping fixed price 25,000–60,000 SEK. Pilot on one use case (campaign optimization, content pipeline, lead routing): 75,000–200,000 SEK depending on number of platforms and evals. Larger programs (3–5 use cases + production rollout): 250,000–500,000 SEK. Retainer for ongoing optimization: 20–40 hours per month. No open accounts.
No. For Swedish-language quality, Claude and GPT-4 are comparable; the choice depends on data sensitivity. For GDPR-critical customer data we run Azure OpenAI with EU residency or Claude on AWS Bedrock EU. For public content production with no customer data the standard API is usually fine. Model choice is configuration, not lock-in.
Yes, with the right setup. What doesn't work: raw LLM output without sources, factual grounding, or editing. What works: a RAG pipeline against your product docs and tone of voice + a human editor who polishes. Result: 5–10× faster production with quality that actually ranks — not AI spam Google deindexes.
Customer data stays in the EU (servers in Stockholm or Frankfurt, LLM calls via EU residency providers). Lead and campaign data is logged with full traceability. Profiling is classified per EU AI Act — we know at build time whether anything qualifies as high-risk. Nothing goes to third parties outside the EU without a clear legal basis.
6+ years as a data and BI consultant at Conversionista, Twoday, and Outfox (Dentsu Aegis) with direct work for IKEA, Spotify, Husqvarna, Resurs Bank, and Stena Line. A big focus has been campaign measurement, attribution, audience segmentation, and A/B testing — now combined with AI and automation for Swedish SMBs.
First pilot in production within 3–4 weeks of contract start. Measurable effects usually show 4–8 weeks in (lower CPA on automated bidding, faster content production, better lead conversion). One e-commerce client went from 15h/week of manual campaign tuning to 2h/week with 23% lower CPA in 8 weeks.
Book a free 30-minute call where we walk through where AI has the highest leverage for you — campaigns, content, leads, or segmentation. No sales scripts, no off-the-shelf packages.
Cookies. Optional analytics only with your consent. Privacy