Sami Porkka
I look at a business process, figure out where AI and automation actually fit, and build the connective tissue.
Integrations, workflows and the guardrails that make them safe to run on real data. Right now that means revenue operations, with AI connectors live on production CRM data and every model call kept inside the EU.
About Me
I’m a product manager in revenue operations. The platforms change from one company to the next, so what I keep is the pattern. Map the process, find the leverage point, ship something that works, and make it compliant by design.
Before this I ran technical projects, owned CRM and customer data, and co-founded an indie game studio, which is why I prototype instead of writing requirements docs.
Selected work
Longer write-ups, each with what’s live and what isn’t.
AI access to CRM data, kept in the EU
A custom MCP server that gives an AI assistant read and write access to our CRM. Live on production data.
A lead pipeline, rebuilt
The first design turned out to be impossible on that platform, so I rebuilt it. Verified end to end in test.
Self-serve warehouse data
Plain language in, CSV out, consent filters on every list. First three colleagues onboarded.
MCP servers
Three built, two of them public on GitHub, and two deployed.
Apps I’ve shipped
Three apps, designed, built and shipped solo with Claude Code. Two are live and Well Planned is retired, with a write-up.
DebtFree AI
Debt payoff planner. Enter your debts, compare avalanche vs snowball methods.
Well Planned
Checklist app for web and Android, built around phases and reminders. Retired in October 2026 and written up as a case study.
Stickieview
Sticky-note dashboard for the desktop. Local, offline, no account. Linux and Windows.
Latest articles
-
Well Planned, the checklist app I built and then retired
A checklist app built solo with Claude Code on Next.js, Expo and Supabase, with freemium billing and reminders, and why I retired it.
-
MCP servers I’ve built and deployed
Three MCP servers I built, two of them open source on GitHub, and two I deployed. Designed so an agent’s mistakes stay small, visible and reversible.
-
A lead pipeline, and the first design that turned out to be impossible
A website-to-CRM lead pipeline whose first design turned out to be impossible on the platform, and the rebuild on scheduled API polling.
HIDDEN: Let’s build something
If you have a process that needs AI and automation wired into it, let’s talk.