---
title: "Best Practices in AI Work | Anssi Nurminen, Lasse Mikkonen | Neuvottelija 408"
titleOriginal: "Tekoälytyön huippukäytännöt | Anssi Nurminen Lasse Mikkonen | Neuvottelija 408"
episodeNumber: "408"
guest: "Anssi Nurminen, Lasse Mikkonen"
datePublished: 2026-09-17
duration: "1:01:45"
youtube: "https://www.youtube.com/watch?v=UaG4uuMYDi8"
originalLanguage: "fi"
topics: ["ai_enterprise_tech"]
subtitleMethod: "none"
provenance: "Owner page assembled from YouTube metadata, the neuvottelija.fi episode record and the publisher's own chapter marks, translated one for one. The summary was written from the uploader's own Finnish caption track. No English caption track exists, so no transcript is published here. Disclosure carried from the episode: the host sits on the board of Fredman Group, where Kespro, Anssi Nurminen's employer, is a partner."
fiCanonical: "https://www.neuvottelija.fi/fi/episodes/854-tekoalytyon-huippukaytannot-anssi-nurminen-lasse-mikkonen-ne/"
canonical: https://www.neuvottelija.com/podcast/episodes/ep408-tekoalytyon-huippukaytannot-nurminen-mikkonen/
---
# Best Practices in AI Work | Anssi Nurminen, Lasse Mikkonen | Neuvottelija 408

English subtitles are not available on this page. This is an episode summary, not a curated transcript.

## Chapters

- [00:00](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=0s) Guests Anssi Nurminen and Lasse Mikkonen
- [00:09](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=9s) The AI working group and its 14 AI workers
- [00:48](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=48s) Anssi's background and the Anneli agent
- [01:31](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=91s) Lasse's path from DevOps to agentic work
- [02:33](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=153s) Ossi and Dude handle the receipt drudgery
- [03:51](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=231s) How Anneli makes videos from WhatsApp
- [05:12](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=312s) Teaching Anneli through feedback
- [06:32](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=392s) What MCP actually is
- [06:56](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=416s) Skills for an AI worker
- [07:46](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=466s) Video as a way into the restaurant trade
- [08:48](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=528s) Meta Ray-Ban glasses in use
- [10:05](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=605s) Image search through the glasses camera
- [10:39](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=639s) Digital footprint and anonymity
- [11:57](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=717s) Making videos while wearing the glasses
- [12:38](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=758s) Continuous recording and strengthening memory
- [13:47](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=827s) Finnish is not supported in the glasses
- [16:21](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=981s) Why augmented reality is the more interesting one
- [17:07](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=1027s) Video models and China's lead
- [18:21](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=1101s) Finland's platform-economy debate is behind
- [19:07](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=1147s) Is it worth waiting for a laid table
- [21:34](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=1294s) Vibe coding versus real competence
- [23:19](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=1399s) Delegating to agents and what breakage teaches
- [24:47](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=1487s) The harness keeps the agent on track
- [26:19](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=1579s) When an agent leaked its reasoning into the group
- [27:49](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=1669s) What other skills have been built
- [28:39](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=1719s) Dude's case memory and meeting notes
- [30:23](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=1823s) A restaurant profitability skill as a joint project
- [31:01](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=1861s) Long-form video and character consistency
- [33:28](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=2008s) Silicon Valley's sense of time versus Europe
- [35:04](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=2104s) Why it is worth trying at any level
- [36:19](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=2179s) Security and the Red Queen effect
- [39:03](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=2343s) Mac Mini, Nvidia and local compute
- [41:13](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=2473s) Pseudonymisation before sending to the cloud
- [43:00](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=2580s) Europe's second chance in models
- [44:21](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=2661s) A monthly subscription is a cheap investment in yourself
- [44:56](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=2696s) Taking an agent to the company IT department
- [47:05](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=2825s) The Hello Humans podcast simulator
- [48:29](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=2909s) Crowdsourcing and trading skills
- [50:03](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=3003s) Memory structure as the year's main theme
- [52:00](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=3120s) Graphify and a custom case template
- [54:35](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=3275s) Memory already works in ChatGPT and Claude
- [56:13](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=3373s) Skill updates based on the conversation
- [57:01](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=3421s) A Finnish-language skill removes the slop
- [57:50](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=3470s) Bots talking to each other in a liminal state
- [58:48](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=3528s) Kesko's AI strategy and training
- [59:50](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=3590s) The race between America and China in models
- [1:01:15](https://www.youtube.com/watch?v=UaG4uuMYDi8&t=3675s) Closing words and Sisäpiiri

## Summary

**Three AI workers, three architectures.** Anssi Nurminen and Lasse Mikkonen belong to the same fourteen-person AI working group as the host, alongside fourteen AI agents that the members can address across one another. Nurminen — speaking as a private individual, although he works in marketing at Kespro — runs **Anneli**, a cloud agent that turns WhatsApp photos, videos or plain text into conceptualised videos through MCP connections, mainly to Scenario. Mikkonen, a long-time DevOps practitioner who now builds agentic systems for companies, runs **Ossi** on the lightest rentable AWS Linux instance as pure orchestration, steered over WhatsApp, and **Dude** on a home machine with a 16 GB Nvidia card. The host's **Samantha** runs on a Mac Mini M4.

**Three terms the rest depends on.** MCP is, in the host's analogy, the USB of the database world. A *skill* is a way of working taught to an agent rather than a program. And the **harness** is the part of the system that keeps the agent on track and checks the result — Mikkonen equates it with automated testing in DevOps. What is new is that the supervision is itself built as skills.

**What the agents actually do.** Mikkonen's examples are deliberately mundane: reconcile a credit-card bill by reading email, fetching the receipts and matching the PDF from the accounting system line by line — with the VAT step still missing, which he says openly. Dude's largest skill is a case memory of projects, investments and meetings that removes note-taking. Nurminen's distinction is the key claim: almost anyone can do one-shots with AI, but repeatable, conceptualised processes are markedly harder.

**Memory is the real subject.** The host describes a layered model — a lossless context expander, a wiki-like long-term memory in Google's Open Knowledge Format, and a graph memory at the bottom. Mikkonen starts from the other end, with a graph memory built on the open-source Graphify and his own case template, and separate vector databases for large document sets. The most usable advice follows at once: ChatGPT, Claude and Gemini already have memory, steered in plain language.

**Three disagreements, left open.** Mikkonen runs inference in the cloud, Nurminen stays in the cloud for access to the newest video models, and the host buys his own hardware for independence should models be export-restricted. Nurminen says he has no worry even if everything leaked, because his use is pure hobby and learning; the other two hold client material. And Nurminen argues against waiting for finished tools — the jump is brutal and you miss the journey — while Mikkonen makes the same point about code: to an outsider, vibe coding and agentic engineering look alike, and the difference is security and scalability.

**Meta Ray-Ban glasses** are demonstrated and fail usefully: face recognition does not identify the host, biometric locking is on the device, and Finnish transcription is not supported at all. Continuous recording as memory augmentation is raised as a use case; the privacy of bystanders is raised and not resolved.

## A note on the source

The summary rests on the uploader's own Finnish caption track, which is far cleaner than machine captions. Product and model names are given as spoken in the episode; this page does not verify their features.

## Watch

The recording lives on the Neuvottelija channel: [Tekoälytyön huippukäytännöt | Anssi Nurminen Lasse Mikkonen | Neuvottelija 408](https://www.youtube.com/watch?v=UaG4uuMYDi8).
A Finnish edition of this episode is published at [www.neuvottelija.fi](https://www.neuvottelija.fi/fi/episodes/854-tekoalytyon-huippukaytannot-anssi-nurminen-lasse-mikkonen-ne/).

## In depth

The Neuvottelija AI editions carry a long-form write-up of this episode:
[English](https://www.neuvottelija.com/ai/ep408-tekoalytyon-huippukaytannot-nurminen-mikkonen/) · [suomeksi](https://www.neuvottelija.fi/tools/ep408-tekoalytyon-huippukaytannot-nurminen-mikkonen/).

## Explore the ideas in depth

- [Enterprise AI Agents: Economics, Governance and the Shift From Tools to Workers](https://www.neuvottelija.com/guides/enterprise-ai-agents-economics-and-governance/): A guide to enterprise AI agents — the real cost model behind agent work, why owning your own stack is becoming a strategic question, a working governance framework with approval gates and audit trails, the agent risk matrix, the EU AI Act timeline as it stands, and who captures the productivity gains. Grounded in a real multi-agent lab, two 2026 keynotes, and Neuvottelija conversations.
- [Nordic SaaS Valuation & M&A: How AI Reprices Software Companies](https://www.neuvottelija.com/guides/nordic-saas-valuation-and-ma/): How software companies in the Nordics are valued and sold as AI moves inference into the cost of goods sold — where multiples stand in mid-2026, the metrics that get repriced, why vertical SaaS defends its premium, AI due diligence, and the shift from seats to outcomes. Grounded in Translink's SaaS valuation work and Neuvottelija conversations.

---

Cite as: Sami Miettinen, Neuvottelija — Best Practices in AI Work | Anssi Nurminen, Lasse Mikkonen | Neuvottelija 408, https://www.neuvottelija.com/podcast/episodes/ep408-tekoalytyon-huippukaytannot-nurminen-mikkonen/, 2026-09-17. For quotes include episode 408 and timestamp.