---
title: "AI, Work and Capital: How an Agent Breaks the Left–Right Game"
titleOriginal: "Tekoäly, työ, pääoma"
description: "Sami Miettinen's Hämeenlinna keynote at Kasvuareena 2026: a working AI agent is simultaneously labour and capital, which scrambles the old work-versus-capital, left-versus-right divide. Illustrated with Samantha, the OpenClaw agent."
role: "keynote"
event: "Kasvuareena"
date: 2026-06-04
location: "Verkatehdas, Hämeenlinna"
language: "fi"
slides: https://www.neuvottelija.com/events/ai-work-capital-hameenlinna.pdf
canonical: https://www.neuvottelija.com/events/ai-work-capital-hameenlinna/
---

# AI, Work and Capital: How an Agent Breaks the Left–Right Game

*Keynote delivered in Finnish under the title **"Tekoäly, työ, pääoma"** ("AI, Work, Capital") at **Kasvuareena** in Hämeenlinna on 4 June 2026. This is an English summary of the argument; the [original slide deck is available as a PDF](/events/ai-work-capital-hameenlinna.pdf). Because the talk was built to be delivered live around visuals, the summary captures the throughline rather than transcribing every slide.*

## The core idea

A capable AI agent is, at the same time, **work** and **capital**. It does labour — it books meetings, drafts, analyses, codes — yet it is owned, like a machine or a share of stock. That dual nature scrambles the familiar political frame of **work versus capital, left versus right**. An agent sits on both sides of the line at once, and that is what makes it hard to slot into the old argument about who creates value and who captures it.

The talk uses one running distinction between what **agents** do and what **humans** do, and asks where the value — and the ownership of that value — actually lands as agents take on more of the work.

## Samantha, made concrete

Rather than argue in the abstract, the keynote introduces a real agent: **Samantha**, part of the OpenClaw home lab. She runs locally on a **Mac Mini (M3, 16 GB)**, combining a frontier model with a smaller local model and **three tiers of memory** (a working/context memory, a graph memory, and a longer-term "Karpathy" memory), and now also does high-end image and video generation.

What makes the point land is the range of roles she already fills, for example:

- booking client meetings for Translink Corporate Finance
- acting as secretary for a shareholder group
- working alongside a mixed team of people and AI "workers"
- serving as a general problem-solver and stock-picker for research
- keeping an EU AI Act usage policy updated week to week

The team slide makes the same argument visually: a professional services firm's roster, with an AI teammate listed alongside the human analysts. The question the audience is left with is not "will agents arrive" but "how do we own and organise them."

## Why it matters for owners

If an agent is both labour and capital, then the interesting question for a business owner is **ownership**: who owns the agent, the data it learns from, and the workflows it runs. That is where the talk connects to the classic themes of capital and productivity — and where a small country's productivity could shift quickly if agents are adopted and owned well. The provocation, only half in jest, is that a well-run agent doesn't take a side in the work-versus-capital debate so much as **break the game** it is played on.

*A short technical appendix closed the talk with the specific model stack used on the day.*

Slides: https://www.neuvottelija.com/events/ai-work-capital-hameenlinna.pdf

---

Cite as: Sami Miettinen, Neuvottelija — AI, Work and Capital: How an Agent Breaks the Left–Right Game, Kasvuareena, 2026-06-04, https://www.neuvottelija.com/events/ai-work-capital-hameenlinna/.