Governed AI operations · on infrastructure you own

We design, deploy and operate AI-driven organizations.

Agent companies are easy to start and hard to govern. We design the organization around your goals, stand it up on infrastructure you control, and tell you, decision by decision, what is genuinely enforced and what is only written down.

A brief turning into an organization chart

The next company runs on people and agents.

The shift is already underway.

The businesses pulling ahead are already working this way. Orchestrating people and agents together is a new discipline, and the advantage goes to whoever learns it first.

Launching agents is not managing them.

The tools will start agents for you. None of them tell you who is accountable when an agent makes a call nobody approved. That gap is where the damage happens.

We design the organization, deploy it on infrastructure you own, and help operate it. You get the upside without giving up control, and evidence, rather than assurance, that it is governed.

Governed, and provably so.

Guardrails, approval gates and a fail-closed posture from the first agent, plus a document stating which of them the platform actually enforces.

On infrastructure you control.

Run it where your data has to live, credential-isolated and egress-controlled. Your credentials and your control stay with you.

Senior operators.

People who run agent companies every day, not junior contractors and not set-and-forget software. They keep it healthy and recover it fast when something breaks.

Open, and yours.

Built on open source and delivered as configuration you own. Inspect it or move it whenever you want. No black box, no lock-in.

01

Design the right organization for your goals

A working session maps where agents create real value in your business, and turns it into a deployable organization built to your goals rather than a template: org design, agent mandates, governance and guardrails. You leave with a company definition and a verification report.

02

Deploy on your infrastructure

We stand it up governed, credential-isolated and observable, on your own infrastructure, so nothing sensitive leaves your control.

03

Operate senior operators running it with you

Oversight, supervision and weekly sparring, with sane cost defaults so it never runs up a surprise bill. The work that currently requires judgment, done by people who have judgment.

Every engagement is scoped to the organization.

Contact us for a proposal
Example organizations

What this looks like, built for you.

The domain matters less than the shape: defined decision boundaries, and output someone can check. Some of the domains we work in:

Research & innovation discovery

Continuous literature and competitor scanning, technology and market screening, experiment design, technical due diligence drafts. Coverage no single hire could sustain, reporting to a person who decides what matters.

Growth & marketing operations

Ongoing research and monitoring, content production on a schedule, editorial review, search coverage, newsletter and distribution, analytics rolled back into what to make next. A coordinated operation with named owners.

Product & software development

Feature discovery and prototyping, spec drafting, incoming issues sorted, implementation, review passes, test coverage, release notes. Work moving from idea to shippable diff, merged only when you choose to.

Every organization we build is designed from your brief and your constraints. These are shapes, not a menu.

A document that says what actually holds.

Most governance lives in prose. Board approval before a new agent exists is a real control: the platform refuses to create the agent. A boundary written into an agent's instructions is a sentence it may or may not follow. In a design document the two look identical, and almost nobody can tell them apart until something goes wrong. The engine's governance report sorts every decision into one of five enforcement classes, decision by decision.

02 Sep 2026
Verified against Paperclip v2026.831.1
hiring-approval
Whether a new agent can be created without board approval
Source: .paperclip.yaml
Enforced
goal-ownership
Which goals each agent is accountable for
Source: AGENTS.md
Reaches the agents as context
escalation-path
When a customer refund gets escalated to a person
Source: refund_agent.md
Words, not switches
secret-binding
Whether a secret is bound to one agent or to the whole company
Provisional

Decisions it can't settle are listed for re-checking. Where something can't be verified, the report says so instead of guessing.

Who it's for

Built for how you work.

For leadership and business owners

Agent capacity without an agent-ops hire. You keep the infrastructure, the data, and the senior seat that decides what ships.

  • Full data control
  • Senior oversight
  • Governed by default

For growth and marketing leaders

A function you can explain to your board without hedging. Someone signs off before anything publishes, and you can point to who.

  • Runs on your calendar
  • You keep editorial say
  • Accountable to a name

For product owners and software teams

The work fits your existing review discipline, not a new one bolted on. Merge stays a decision your team makes, every time.

  • Your review process, kept
  • Nothing merges unwatched
  • Fits existing tooling
Book a working session

We build and operate on Paperclip, the leading open-source platform for agent companies.

Paperclip runs teams of AI agents as a company: org structure, goals, budgets, approval gates and task tracking. It is self-hosted, so it runs on your own infrastructure and your data stays local. We're specialists in it.

Designed, not templated.

Every company is designed around your goals, with org structure and governance built in, not a downloaded template with the names swapped.

Custom modules.

We extend the platform with the skills, integrations and modules your operation needs, beyond what ships by default.

Run to a reliability standard.

We handle the integration work that trips up naive deployments, so it holds up in production instead of falling over.

Verified, not asserted.

Our governance model is derived from what the platform actually enforces and confirmed against running instances, which is why the report can tell you a decision is prose rather than quietly counting it as governance.

We run our own company on Paperclip. We operate what we sell.

"Senior operators" means someone specific.

You're buying judgment, not a brand. Here's who.

Kimmo Nurmisto

Kimmo Nurmisto · Founder

Fifteen years building growth systems for technology companies. Now designing and running agent companies in production, and building the tooling that makes the work repeatable.

Track record

SaaS, deep tech and infrastructure, from pre-seed to listed. Creator of the open-source CLI that generates Paperclip companies. Senior Lead at Innovation Fund Sitra, driving AI-led productivity across the Finnish public sector.

How I work

Operator, not agency. Exec-level seat.

Book a working session.

Book a working session