Why Koru runs on agents, and where a person still signs off
We run our own business on supervised AI agents. Here is what they do, what they are not allowed to do, and why that matters for clients.
The Koru Project is an agent-first company. It isn't a slogan. It is how a small firm delivers client work from day one while still writing, publishing, answering email and keeping its website current.
What the agents do
- Content and research. Turn podcast conversations and field notes into blog drafts and social posts, and watch for regulatory and vendor changes.
- Inbox and follow-up. Sort incoming email, summarize what arrived and draft replies.
- Website and operations. Propose site updates as reviewable changes, and prepare a daily brief of new conversations, open tasks and spend.
What they are not allowed to do
- Send an email, publish a post or change the website without a person approving it.
- Act on a platform in ways its rules prohibit. On LinkedIn, for example, posts go through approved publishing tools, and connection requests and messages are sent by a person.
- See client data they don't need for the task in front of them.
Why it matters to clients
Everything we recommend to clients, we run on ourselves first: least access, human approval, clear owners and measured results. When we help an office set up assistants or agents, we are describing a system we live with every day, not one we read about.
Good agents don't remove people from the work. They remove the work that keeps people from the parts only they can do.