Connection and security
What the agent touches, and what stays with you.
We agree the agent’s access, approval steps and data handling with you. The setup depends on your tools and requirements.
- 09:12Agent → YouDraft quote
- 09:14You → AgentApproval
- 09:16Agent → CustomerApproved quote
- 09:18Agent → YouSend confirmation
Rights
Four levels, tool by tool.
For each connected tool, we agree together on what the agent can do. It has only the rights the task needs, and you choose which actions need approval.
Read
It looks, without changing anything.
For a quote: The customer’s request, their record and your price list.
Write
It creates or completes, within the agreed tool and scope.
For a quote: The quote created as a draft in your management tool.
On approval
It prepares, then waits for your go-ahead before acting.
For a quote: The quote sent to the customer after your approval.
Never
Ruled out at scoping, and written into its rules.
For a quote: Inventing a price that is not in your price list.
The framework
What frames the agent, from day one to switch-off.
Limited access
We choose the connections and permissions needed for the task, based on what your tools support.
Example: An “Agent” account in your CRM, read-only.
Autonomy agreed with you
You choose which actions require human approval. Routine actions can be authorised in advance, within the agreed limits.
Monitoring agreed together
We agree what to keep in the log to review actions and find the relevant approvals.
Where the data goes
Hosting and services are chosen during scoping. We specify which data passes through them and how long it is retained.
Example: In your environment or on an external service, depending on the project.
Access you can revoke
You can revoke the agent’s access. We show you how to stop the system and disable its connections.
Example: Disable the “Agent” account or revoke its connection to the CRM.
A specific requirement on your side?
Contracts, compliance, internal rules: we set them out during scoping, before connecting anything.