Your information.
Clear boundaries.
The platform applies access checks to research requests, source reads and saved results. Your company’s connections, research environment and data-handling requirements are scoped before rollout.
Approved people. Approved sources.
Server-side checks bind access to a signed-in identity, company membership and permitted sources.
Boundaries between companies.
Company-specific research workers and storage boundaries separate investigations, credentials and retained evidence.
A record of the work.
Saved source evidence, citations and access records support review of how research was requested and delivered.
Identity and permissions are checked on the server.
The managed workspace uses an identity provider for sign-in and server-managed sessions. Each request checks the user’s current membership and permissions. Knowing a company or task identifier does not grant access to it, and a prompt cannot assign a new role or add an approved source.
Saved research is checked against its owner and current source permissions before it is returned. Member suspension and session revocation take effect on subsequent requests. Billing access is separate from access to research content.
Browser sessions use secure, HTTP-only cookies, with checks to prevent another website from submitting authenticated changes. Sign-in tokens are not stored in browser local storage.
Separate research environments and protected credentials.
The managed architecture assigns each company its own research worker, private working storage and source configuration. The shared management database also enforces company boundaries at the database level, alongside application access checks.
Provider credentials are held in protected server-side configuration, outside the agent’s research workspace. The management interface does not receive the source or model-provider credentials. Source readers use the permissions configured for that connection and check approval before reading data.
What happens to your data during research.
A question becomes a recorded research task. The agent reads the approved sources selected for that task and retains the material used as evidence in the company’s research environment. Completed answers, citations and task records are saved so authorised users can return to them.
Direct access to a source does not mean that no copy is kept. Questions, retrieved evidence, generated answers and artifacts can all contain company information. They need to be covered by the agreed retention, deletion and backup arrangements.
The research tools are read-only. Reading a document or message does not grant permission to edit it, send a reply or change the source system. Any action-taking integration requires a separately agreed scope and controls.
AI processing has its own data boundary.
The configured AI provider processes the question and relevant retrieved material to perform the investigation. Separating company research environments does not mean the model runs entirely inside your own infrastructure.
Before rollout, we agree the model-provider route and review which data it receives, processing locations, provider retention and training-use terms. Those details depend on the selected service and account configuration; we do not make a blanket no-retention or no-training claim for every deployment.
Source content can contain misleading instructions as well as useful information. Tool permissions and approved source boundaries are enforced outside the model. Generated answers remain subject to human review appropriate to the decision being made.
Access records and operational visibility.
The management service records research requests, acceptance and answer access. Routine operator views show task and service metadata rather than client questions and answers. Operator actions have separate permissions and audit records.
Source availability, approval state and recorded delivery are monitored separately. An unavailable source or uncertain task outcome is surfaced for investigation rather than treated as proof that the information does not exist.
Data protection is completed for your deployment.
The workspace is served over HTTPS. Remote database connections require verified TLS as part of the installation requirements. Infrastructure controls must then be confirmed for the environment chosen for your company.
Before sensitive data is connected, we agree and verify the applicable requirements for:
- Hosting and processing regions, including the AI-provider route.
- Encryption at rest, secret storage and credential rotation responsibilities.
- Retention of questions, source evidence, answers, artifacts and access records.
- Deletion and export procedures, including retained copies and backups.
- Backup protection, restoration checks and incident responsibilities.
We do not claim a certification or compliance standard on this page. Discuss your requirements with us so the controls and evidence needed for your engagement are explicit.
The public website assistant is separate.
The guide on this website uses a small collection of published company information. It cannot access client knowledge, internal systems or action tools. Its answers link to the relevant page.
Server-side bot verification, request limits and strict input limits protect the service. The model can select approved answers; it cannot publish arbitrary text or run commands. Please do not submit confidential material. See our privacy notice for processing details.