Podcast · AI Kanava · 2025-11-10 · 1:01:15 · In Finnish
Economy, AI and Europe's Role in the AI Race — Sami Miettinen on AI Kanava
Published as “Vieraana: Sami Miettinen – Neuvottelija: Talous, Tekoäly ja Euroopan rooli AI-kilpailussa”
Hosted by: Lasse Mikkonen · Harri Juntunen
Lasse Mikkonen and Harri Juntunen recorded a special episode of AI Kanava at the Neuvottelija studio, having met Sami Miettinen at a private equity event and, within about ten minutes, agreed to found a vibe-coding subsection for private equity investors. The episode opens with a joke Miettinen generated himself: how did DeepSeek escape the Chinese room? It found a backdoor just a prompt away — and generated itself out.
An investment banker back at the keyboard. Miettinen studied at Tampere University of Technology, where object-oriented programming in Modula-2 was the newest thing, then moved to business school and deliberately left coding behind — Excel macros, then Visual Basic, neither worth relearning. AI brought him back. Going straight to the deep end with Codex in Visual Studio Code and Python produced too much hallucination, and pushing to GitHub from a command prompt felt like a thirty-year regression. The Swedish startup Lovable turned out to be the right-sized back door. A website that had taken half a year with two students on WordPress and PHP was reproduced in an hour from a single prompt, YouTube carousels included; the slowest part was working out the YouTube list API call. A custom CRM for Slush followed over a weekend, running on Supabase with a backend, a frontend and code he can read and push to GitHub without working at the code level.
The bug that explains the current state of the tools. Miettinen pressed a button offering improved security. It revoked his own admin update rights on one database — and did not tell the application code. Twelve prompts insisting the database be updated achieved nothing, and ChatGPT could not diagnose it either. The security feature and the coding agent were not talking to each other, and he had to work that out himself. Asking for his rights back fixed it; the detour consumed all his credits, and he upgraded a tier. Profit, as he notes, to Lovable’s business model.
What happens to SaaS when AI becomes the interface. Translink, where Miettinen is a partner, specialises in selling recurring-revenue software companies — ARR, low churn, gross margin above 90 per cent, Rule of 40. Coding himself has made him look harder at where the value in a median SaaS company actually sits: usually a database plus a usable interface onto it. The hosts have speculated on AI Kanava that enterprise software’s future is a generative user interface over pure data, and Miettinen has come to the same view from the user end — the world is full of poor UX built to an engineer’s logic. His example is using Perplexity’s Comet browser to reach a Statistics Finland database directly, bypassing the pull-downs entirely, with MCP removing the remaining friction.
He offers a Tolkien frame for it: in the UX world humans are hobbits, clicking with mice and dropping spreadsheets into download windows, while the Valar and Maiar of the AI layer move the pieces of the universe far more efficiently. The One Ring is your most trusted AI interface — and the question is whether it carries you into the dark world.
His commercial hope is narrower than his prediction: vertical B2B SaaS sitting behind a genuine moat should have at least a few good years, and dissolving the whole UX layer into protocol chatter between bots would take money away from a lot of Finnish coders and owners. The threat he names is uberisation — transactional API calls billed by the token rather than recurring subscriptions, with MaaS Global’s failed flat-rate model as the cautionary case. Translink’s quarterly SaaS reports already show the split: vertical SaaS trading at roughly 5x ARR, horizontal SaaS having fallen below 3.
The capital problem. Finland’s largest SaaS company, AlphaSense, runs at around $500 million ARR — and operates principally from the United States. Miettinen’s diagnosis is allocation. Solidium he calls close to a joke; the €270 billion pension sector is a separate monolith that has adopted best practice and therefore invests some 60 per cent in America, where capital allocation is most efficient. Total Finnish wealth is around €700 billion, most of it in homes. Against €540 billion of domestic revenue, even a 1x multiple implies about €154 billion of capital to run Finnish business — roughly two-thirds of which is currently foreign. The equity savings account passed one billion only last year, which is negligible in a global game. Meanwhile home-market bias moves startup headquarters: a company valued at one million in Finland is ten in Sweden and a hundred in the US, and Swedish household wealth per adult (around €300,000) sits between Finland’s €150,000 and America’s €502,000. His proposal follows Sweden’s: carve roughly 10 per cent of the pension contribution out for individuals to direct and legally own, so that domestic risk can be taken by someone with money genuinely at stake. Foreign owners, he notes, do not do due diligence on Finnish companies if no domestic investor will.
Three regulatory models. China’s totalitarian surveillance, from which nobody escapes; America’s deregulation, where half a trillion dollars buys the removal of the regulator from the middle; and Europe’s bottle-cap philosophy of building data protection 2.0 and calling it safety. Miettinen is sceptical — Europe lacks the capital to compete even unregulated, and adds both hands taped behind its back. Mikkonen counters that there is a European moment: 550 million citizens who want neither dependence on American hyperscalers nor Chinese totalitarianism, and an AI world grounded in individual rights remains possible. Both agree the open-source tradition — Torvalds, Linux, Git — is something Finland has given the world generously, while the productisation profits went to GitHub, and that Thiel’s preference for monopoly describes how the American model actually captures value. Mårten Mickos’s account of MySQL running from a $2 million seed round through Sun to Oracle is the same story told once more.
Negotiation, delegated. Miettinen updated Uusi neuvotteluvalta with Juhana Torkki in 2019 — some 30,000 copies sold — and interviewed Risto Linturi for it. He and Torkki now run client projects modelling how a company’s negotiation scripts should be built, where the value-creation points sit in the process and in the offering, and feeding that to a support AI. The further vision comes from the author of Rauhankone: negotiations of meaning, in which each side gives its fundamental interests to an AI and the machines converge on a deal for the humans to approve. Miettinen notes that the roles he has taught for years — the active beta negotiator, the observing alpha making critical calls, the scribe feeding both over Slack — are all automatable. Mikkonen presses on whether the human conversation that works the agreement has value of its own; Miettinen agrees it does for now, since a person must understand what they sign, and adds a sharper point: human trust is built on a track record, whereas an AI’s reputation is a function of code that can be refactored overnight.
The episode closes with one action each. Miettinen: coding has never been easier, and a visual MVP costs a few tens of euros — build the demo and show that to a developer instead of a PowerPoint. Juntunen: fix the capital shortage so demonstrations can be financed at all. Mikkonen: build your own support AI by giving a custom GPT or a free assistant your own documents as context, which unlocks a level without any investment.
Miettinen also mentions, in passing, a project of his own: having built neuvottelija.fi on Lovable, he has reserved neuvottelija.com for a site carrying transcripts and prompts aimed at machines rather than people — possibly with ugly UX, on the theory that what the models read may soon matter more than what humans see.
Topics: AI & Enterprise Tech · Investing & Markets · Finnish Economy & Policy · ← All media visits · Talks & Events · Markdown mirror: index.md