AI Consultant
Independent AI consultant for Swedish small and mid-sized businesses. I help you move from AI experiments to production-grade systems with measurable ROI — not more PowerPoint decks.
AI consultant is one of the most diluted titles in tech right now. What I actually do: I work alongside your team for 2–16 weeks, map where you're losing time or money, and build AI and automation systems that run in production when I'm done. No slides without code behind them. No reselling enterprise platforms you don't need. No AI agency with five juniors on the invoice. Just one senior consultant who has shipped data and AI for IKEA, Spotify, Husqvarna, Resurs Bank, and Stena Line — now focused entirely on helping SMBs make the leap from manual work to intelligent automation.
Your team built a GPT prototype that impressed leadership but never made it to production. Common cause: no clear measurement plan, no production environment, scope too broad. Fix: cut scope to a measurable task, build evals first, then solve.
Sales lives in HubSpot, support in Zendesk, finance in Fortnox, planning in Notion. Data is moved between them by hand. n8n or Make orchestrates the data flow — and an LLM makes the call where rules aren't enough.
Your company has 12 years of collective know-how scattered across email, Confluence, Slack, and the heads of two senior staff. A RAG implementation makes it searchable with source citations — without leaking data to vendors that don't meet GDPR.
Routine decisions (which customer, which quote, what level of support) happen manually and pull senior staff out of higher-leverage work. An agentic AI with clear escalation rules takes 80% of the calls in seconds and flags the rest.
We walk through your processes, systems, and data. I identify 2–3 areas with the highest ROI for automation or AI. Output: prioritized roadmap, build-vs-buy guidance, GDPR / EU AI Act notes. Fixed price.
We build the top-ranked solution together. I write the code, you test. Existing platforms first (n8n, Make, OpenAI, Claude); custom code only where it earns its keep. Weekly checkpoints.
The solution rolls out, docs are written, your team is trained. Self-hosted runtime means you own the system — no vendor lock-in. Optional retainer for continuous improvement.
Default: n8n self-hosted on Coolify for orchestration, Postgres + pgvector for retrieval, OpenAI or Claude for LLM calls. Everything runs in the EU. For Microsoft-heavy orgs: Power Automate and Azure OpenAI. For Google Workspace shops: Gemini and Apps Script. The stack is chosen for your existing environment, not what I happen to like.
My services: AI Readiness Assessment fixed price 25,000–60,000 SEK depending on scope. Workflow projects 50,000–150,000 SEK. Larger programs 200,000–500,000 SEK. Retainer typically 20–40 hours per month. I bill with caps — no open accounts that drift.
I'm one person. That means you get a senior delivery every time, no juniors on the invoice, fast decisions. Trade-off: I can't run five large projects in parallel. That constraint is deliberate. If you need 50 consultants from a pool, you should call someone else.
No. The default is to pick the LLM by use case and data sensitivity. For GDPR-critical data: Azure OpenAI with EU residency, Anthropic Claude on AWS Bedrock in the EU, or fully local Ollama on your own hardware. Model choice is configuration, not lock-in.
6+ years of data work for Swedish companies before founding VasthavM in 2022 — including IKEA, Spotify, Husqvarna, Resurs Bank, and Stena Line via Conversionista, Twoday, and Outfox Intelligence (Dentsu Aegis). The practice since then has focused on SMBs in Sweden and the EU.
Yes. Workflow runtimes run on EU servers (default: Stockholm or Frankfurt). LLM calls go through providers with EU residency options. EU AI Act classification is part of the readiness assessment — we know which systems qualify as high-risk before we build them.
Existing platforms first. n8n, Make, Power Automate, OpenAI, Claude — these solve 80% of cases faster and cheaper than custom code. Custom code goes in where the platforms leak: auth, evals, domain logic, RAG quality.
A typical first workflow is in production within 2–4 weeks of contract start. For larger programs (assessment + build + handover): 2–4 months. ROI examples: an e-commerce client went from 25 hours/week of manual work to 4 hours/week in 6 weeks. Another reduced support response time from 6 hours to 22 minutes in 8 weeks.
Book a free 30-minute call where we walk through where you stand, what's realistic, and what the next concrete step is. No sales scripts.
We use cookies for essential functionality and, with your consent, analytics to improve the site. Read the Privacy Policy.