For IT teams
Technical overview.
How a deployment of our AI agents is built, in terms your IT and security teams can check. The details for your project go into your proposal.
Your data stays where you choose: on your own servers, in your own cloud account, or with a contracted model provider. We confirm which in writing.
How a deployment is built.
- Agents with one job each
- Each agent has one role, such as reading, checking or drafting. The rules for that role are agreed with you.
- The tools each agent may use
- An agent can reach only the systems and actions its job needs. Everything else is out of its reach.
- The approval queue
- Actions that need a person wait in a queue with the agent's notes, until someone with the right role approves or holds them.
- The log
- Every action is written down with its time, its source and who approved it. Your team can read it and export it.
- The model
- Runs in the delivery option you choose. The agents, the approvals and the log are built around it the same way, whichever model is used.
Connecting to your systems.
Our integrations page has the detail.
- We use each system's API where it has one, with an account limited to what the agents need.
- Without an API, we read its database with read-only access where possible, or work from its exports.
- Writes happen only through approved steps, and each one is logged.
- If a system has no way in, we tell you before work starts.
Where it runs, and how it's kept safe.
- Where it runs
- Your own servers, your own cloud account, or a contracted model provider. Where your AI runs explains each option.
- Security and safety
- Data location, training, access, approvals and the off switch. Security & safety covers each one.
What you receive at launch.
- An architecture overview: the agents, their tools, the approval queue and the log.
- The delivery option, with the model named and where it runs.
- A runbook: how to run, pause, switch off and restart each agent.
- The access list: who on each side can reach what.
- A sample of the log, in the format you can export.
Questions from IT teams.
What runs on our side?
It depends on the delivery option, and the proposal lists it. On your own servers, everything runs inside your network. In your cloud account, everything runs inside that account. With a contracted provider, the model runs with the provider and the rest sits where we agree in writing.
How do the agents sign in to our systems?
With accounts you create and control, limited to what each agent needs. You can revoke them at any time.
Can we review each agent's instructions and rules?
Yes. Each agent's rules, approval points and tools are written down and shared with your team.
Which model do you use?
We name the exact model, and where it runs, in your proposal and contract.
Programs we're part of
Bring your IT team to the call.
We'll go through the architecture, the delivery option and the access your project needs.
Talk to usYou decide. We answer for it.