For the stewards of creative work
Intelligent agency for artificial intelligence.
You hold it in trust. That changes every answer.
Estates, foundations, and family trusts. Museums, theaters, orchestras, libraries, and archives. Universities, publishers, and galleries. Every decision here is really two: is this good for the work, and would the person whose name is on it have agreed?
What you hold is singular. What you build around it should be too.
You carry two obligations at once.
The work of the people who came before, held in trust on behalf of those who are not here to speak for themselves. And the working lives of the people coming next.
Most AI conversations serve one of those and quietly cost you the other. A vendor who has never had to answer to a board or a family will not see the difference, and will not understand why you said no.
Rights work precedes build work. Always.
Everything we build runs on a consent scope somebody actually granted — traced, recorded, and legible to your board or an estate. Where the scope does not cover it, the answer is no, and we say so early.
Most institutions have never had this written down in one place. What you hold, what you hold on someone else's behalf, and what you are permitted to do with either. That record is the deliverable, and it is useful whether or not anything gets built afterward.
We will tell you to wait. Some archives are not ready. Some rights questions need answering before anything gets built. We would rather lose a project than hand you a problem.
Bring your board.
Every use authorized, credited, and recorded. Defined scope, a real end date, and an off switch that works.
The questions a trustee asks are the questions we start with: what did we permit, who approved it, where is that written, and what happens if we change our minds. If those cannot be answered plainly, the project is not ready, whatever the demo looked like.
We stay in the room. One team from the first conversation through the first month after launch. No handoff to a junior bench, no learning your collection on your dime.
This does not have to come out of program budget.
If the right source is a capacity-building grant or a designated gift, we help build the request: the case for support, the project budget, and the scope in the language a funder expects.
Nothing rolls forward automatically. Every stage is quoted on its own and approved on its own, and nothing is priced on a percentage of something you cannot yet see.
Retrieval, not generation.
Your material stays outside the model itself. It is retrieved at the moment of the question rather than absorbed into training.
Two consequences a trustee will care about. Every answer traces to a source, so the rights position stays auditable. And a change of scope is effective immediately, because there is nothing to unlearn.
Stated plainly, because the category is full of people who will not.
We don't imitate.
Nothing we build generates content in anyone's style.
We can't stop others from scraping you.
That's litigation, not software.
We don't clear your rights.
Identifying and sequencing the rights work is ours. Clearing it is your counsel's.
We don't train on your material.
No fine-tuning, no weight updates, no derivative model artifacts.
Your data stays yours. Never pooled, never sold, and it leaves with you.
Somebody is going to ask. Come back with an answer, not a policy.
We would like the first AI agreement in your field to be one you wrote.
Start with the Readiness Review. Commit to nothing else. One conversation, no deck. Tell us what you are looking after and what is worrying you, and we will tell you honestly whether we are useful.