Is it safe to use AI on clients' financial data in bookkeeping?
By Fidelis Solutions · Published June 10, 2026
The confidentiality of client financial data is a professional obligation, not just a preference. Before introducing any AI tool into a bookkeeping workflow, a firm needs to know exactly where the data goes, who can access it, whether it is used to train any model, and what the retention policy is.
Fidelis Ledger — For Professionals, built by Fidelis Solutions, runs its AI inference layer on AWS Bedrock under a zero-data-retention (ZDR) agreement. That means:
- No client prompts or outputs are retained by the model provider after the inference call completes.
- No client data is used to train or improve any foundation model.
- Data stays within the AWS infrastructure boundaries; it is not shared with third parties.
Beyond the inference layer, the platform enforces row-level security so that each client's data is isolated from other firms' clients. Staff access is scoped to the clients they are assigned. All data in transit is encrypted; data at rest is encrypted using AWS KMS.
The platform does not provide tax advice or legal advice — it categorizes transactions and drafts entries for your licensed professional to review. The professional judgment layer remains with your firm.
For firms evaluating the full tool stack, see what to look for in an AI bookkeeping tool for QuickBooks Online. For firms weighing the ownership question, see how keeping bookkeeping in your firm compares to third-party services.
If you want to walk through the data handling posture in detail before a demo, reach out to Fidelis Solutions.