Enterprise

Give every team AI agents, under rules IT sets

A workspace for each team, with AI agents your admins control, confidential AI for sensitive work and a rollout we plan with you.AI agents your admins control, and confidential AI for sensitive work.

Request a demo How we protect your data

Control

Your organisation decides

Separate workspacesGive each team or project its own workspace with its own members, channels and agents. Members and agents in one workspace can't see another's.Separate workspaces. Each team sees only its own.
Agent policyAdmins choose the models, tools and connectors agents may use.Agent policy. Admins choose models and tools.
One pool for the organisationEvery seat's included usage goes into one pool, and credits bought in advance top it up.One pool. All AI usage, shared.

HOLOH WorkspacesSearch settings

Agent policy

Applies to every workspace in Northwind. Workspace admins can narrow it, not widen it.

Frontier modelsPro and Max seats

Web searchAll agents

FilesFiles shared in the workspace

GitHubCoworker agents in #engineering

Before an agent actsMembers can change this per agent

Separate workspacesChecked on the serverOne usage pool

Security

How we protect your data

Planned for Enterprise: single sign-on with your own identity provider, an exportable audit log and your own retention settings. Ask us where each one stands.Single sign-on, audit log and retention settings are planned.

Request our security documents

Confidential AIIn a confidential workspace, prompts and answers are encrypted before they reach the model, which runs only on sealed hardware. HOLOH checks that hardware before every request and stops if the check fails.Confidential AI. Sealed hardware, checked every time.
Workspace boundariesAgents and members see only the workspaces they belong to. The rule is enforced on the server, on every request.Boundaries. Enforced on the server.
No training on your dataWhat your team sends to an agent is not used to train AI models.No training on your data.
EncryptionData is encrypted in transit and at rest.Encryption in transit and at rest.

	Standard workspace	Confidential workspace
Where the AI model runs	With the model providers HOLOH works with, under terms that rule out training on your data	On sealed hardware
Can the model provider read prompts and answers?	It processes them to answer, and may not keep them for training	No
Before each request	Access rules checked	Access rules checked, and the sealed hardware checked
If a check fails	The request is refused	The agent stops; it never falls back to an ordinary model

Sealed hardware means servers whose processors keep the model's working memory encrypted and walled off, so even the people who run the machines can't look inside. HOLOH checks this before it sends anything.

Rollout

Start with one team

Nothing to install

HOLOH runs in the browser; the desktop app is coming soon.

Teams stays

People on Teams keep it, and HOLOH members join their meetings.

A named contact

We plan the rollout with you before the first workspace is created.

Contact

Talk to us

Tell us about your team and what you need. We'll reply within one working day.

Email us at info@holoh.com