How Keelwater handles client data.

No as a default operating practice. Sensitive client workflows should use business/enterprise-grade accounts and approved tools with documented privacy controls. The exact toolset depends on the workflow and the firm's requirements.

Keelwater prioritizes providers and account configurations that contractually restrict model training on business data where sensitive client information is involved. We document the actual provider and data flow rather than relying on a blanket "AI is secure" claim.

We design around least privilege: only the systems, clients and actions required for the workflow. Where practical, integrations use dedicated service accounts, API applications and scoped permissions rather than shared employee credentials.

Tax-return information requires workflow-specific analysis. Keelwater maps where the information is used or disclosed and designs applicable disclosure/consent steps into the workflow where required. We do not treat one generic consent form as a substitute for reviewing the actual data flow. Final legal/compliance determinations remain with the firm and its qualified advisers.

No. Keelwater designs human review gates into professional workflows. AI can gather, classify, draft, route and surface exceptions; the responsible accounting professional retains judgment and sign-off.

Usually not. Keelwater's default is to connect and improve the systems that already work. Replacement is considered only when a tool is itself the constraint.

Production workflows should preserve enough event history to determine what the system did, what source it used, what changed and where a human approved or overrode the action. The exact logging layer depends on the systems involved.

Data retention and deletion should be defined per implementation. Keelwater should minimize unnecessary copies of client data and document what persistent data stores, if any, are created outside the firm's existing systems.