01 / The thesis
Capability no longer has to grow in step with headcount.
Building a company has often meant adding people, handoffs, meetings, and management layers as the work grows. Agents create another possibility: a small team can coordinate a much larger amount of work, with software taking on parts of research, production, analysis, and routine operations.
That is the idea behind the Foundry’s provocation: “One person can now run what used to take twenty.” It is a direction to explore, not a staffing ratio or a promise for every business. The practical question is which work can change—and what must remain firmly in human hands.
02 / The agent economy
From using tools to directing work.
By “agent economy,” I mean a way of working where software can pursue defined tasks, use tools, and coordinate steps within explicit boundaries. People set the purpose, supply the context, evaluate the result, and remain accountable for the consequences.
My thesis is that advantage will increasingly come from how well a company organizes that relationship. Buying access to a model is only a starting point. The harder work is making company knowledge usable, redesigning workflows, assigning ownership, and earning trust through reliable performance.
03 / Human responsibility
Delegate the work. Keep ownership clear.
Agents need a defined remit: what they may access, which actions they may take, when a person must approve, and how failure is handled. A named human owns the business outcome. Important decisions need review, and consequential actions need appropriate controls.
Judgment, taste, relationships, and responsibility matter more when execution becomes easier. A company still has to decide what is worth doing, whose problem it solves, and what a good result looks like.
04 / The way of working
Start small enough to learn. Build well enough to trust.
- Start with a business constraint. Find the delay, cost, quality problem, or unmet customer need. Establish a baseline before introducing automation.
- Design the whole workflow. Define the inputs, source of truth, agent tasks, human decisions, permissions, and handover points.
- Prove it in real work. Test representative cases, including failures. Review output quality, reliability, usage, and cost. Keep a practical fallback.
- Give the team ownership. Document the system, coach the people using it, and make decisions and performance visible.
- Expand when the evidence supports it. Scale what earns its place. Change or stop what does not.
05 / The economics
Count the value that survives contact with reality.
Generating more output is useful only if someone needs it. I look at the total cost of a workflow: models, integrations, review, maintenance, mistakes, and the effort required to change how people work.
The outcome may be faster delivery, better service, lower operating cost, or capacity for something the business could not previously attempt. The measure belongs to the business problem. It should not be chosen to make the technology look successful.
06 / Why the Foundry exists
The company should serve a life worth living.
I co-founded Haltian and helped grow it past €19M in revenue and 100 people. I know the demands of building an organization. From Levi in Finnish Lapland, I am exploring another model: ambitious work supported by capable systems, with space for health, family, and life outside the company.
Herman Foundry is where I put this thinking into practice. Through a few focused leadership collaborations, I also help established companies apply it to their own circumstances. The aim is durable capability, commercial value, and an organization that can keep improving.