Your day starts with decisions
The inbox is sorted before you arrive. What needs you is at the top, and the rest is filed, drafted or already handled.
AI virtual employee — Daneel
Hand over the work. Keep the decisions.
Daneel is an AI virtual employee. They read, draft, file and chase, so your people can spend their day on the decisions only they can make.
You agree what they may take on. You see everything they do. They are given more as they prove themselves, like any new colleague.
Built on AILANG · Private by default · Promoted on their record
What changes when Daneel joins the team.
Nothing drops
Every request gets a reference, and every follow-up happens on the day it was promised.
Sensitive work gets delegated
Client material and confidential documents, the work you would never paste into an AI tool, handled inside limits you set.
You can show what happened
Every action and every refusal is on the record. When a client or a regulator asks, you have the answer.
A clear process for what you hand over.
Delegation fails when nobody agreed what was delegated. So Daneel starts with a job description. With the Decision Budget, we map the decisions in your work and agree which ones Daneel takes, which stay with your people, and what they must never do.
Each kind of work then sits on a rung. That work moves up when Daneel’s record earns it, and back down in one step whenever you choose.
- 0
Held
Your people decide.
- 1
Instrumented
You see what they would have done.
- 2
Shadowing
They work alongside your team to compare.
- 3
Proposals
They do the work. You press send.
- 4
Automated
They act within your limits, and ask outside them.
Why AI employees need care.
An AI employee, by definition, has all three parts of what the security researcher Simon Willison calls the lethal trifecta. Put them together, and one well-crafted email can turn a helpful assistant into a leak.
Your private data
They read your mail, your documents and your clients’ files. That is part of their job.
Content from strangers
Every email, attachment and web page they read can carry hidden instructions.
A way out
They can send email and share documents, so whatever they read can leave.
Most AI employees guard against this with instructions to the model, and instructions can be argued with. Daneel’s limits are built into the program itself, written in AILANG and proven before it runs. No message can talk them into writing to someone they should not.
Built on AILANGTrust you can see.
You do not need to understand how Daneel thinks. You need to see what they do.
They declare their authority
If something is not written down as allowed, Daneel cannot do it.
They show their work
Every action goes on a record they cannot rewrite.
They say “I don’t know”
Outside their limits or beyond their knowledge, they stop and ask.
They stay private
On your hardware or in a private seat with us. Your material leaves only by routes you have agreed.
The thinking behind this: “Can I trust AI?” is the wrong question.
Your employee, not our software.
They work at your address, learn how your firm does things, and follow the rules you set. Ours is called Daneel. You can call yours whatever your team will actually say out loud.
They earn their place when…
- Your best people spend their days on work beneath their judgement.
- Confidentiality is the reason you have not adopted AI already.
- You would delegate tomorrow if you could watch them first.
- You need to explain the arrangement to a client, a board or a regulator, and be believed.
What would you hand over first?
Tell us about the work that fills your team’s week. We will show you how Daneel would take it on, what they would be allowed to do, and how you would see them working.
Delivered by Sunholo / Holosun ApS, Copenhagen. We build AI systems and the controls that make delegation practical.