DATA AND ACCESS
Use the minimum information needed.
Begin discovery with synthetic or appropriately redacted examples. Before connecting business data, agree the data categories, permitted use, authorized users, approved environments and transfer method. The public enquiry form is not a confidential upload channel.
MODEL AND PROVIDER CHOICES
Agree where information is processed.
Provider selection considers data-use terms, retention, geography, quality, latency and operating cost. Document the chosen configuration before production data is connected. We do not promise that every provider has the same training, retention or residency policy.
TOOLS AND AUTHORIZATION
Separate a proposed action from permission to execute it.
Define allowed tools, scoped credentials, approval gates and logging. Server-side authorization remains necessary when an AI model proposes an action. MCP and other integrations receive a workflow-specific access and threat assessment.
QUALITY AND RELEASE
Make acceptance repeatable.
Define expected and unacceptable results using representative examples. Review failure handling, data isolation and access checks alongside interface behavior. Document known limitations, release responsibilities and rollback or recovery steps appropriate to the system.
HANDOVER AND RETENTION
Record who owns and maintains each part.
Agree repository access, service accounts, source and asset ownership, documentation, support and data return or deletion at the end of the engagement. Backup and restore responsibilities are explicit deliverables when included; availability and recovery guarantees are not assumed.
SCOPE OF THIS PAGE
Project arrangements are agreed in writing.
This page describes a delivery approach, not a certification, executed NDA or universal compliance claim. The signed project agreement and chosen providers define the applicable commitments. Website enquiries are handled under the separate privacy notice.