How can I use generative AI on documents while controlling where data is processed?
Use FormKiQ's AI Processing and Analysis module with an approved Amazon Bedrock model, execution role, and processing geography. Customer-managed storage gives you deployment control, but it does not mean inference runs inside your own AWS account or always stays in one region. Review inference profiles, logs, outputs, and external integrations as part of the data boundary.
Choose an approved processing path
The module invokes configured models through your AWS account permissions. AWS operates Bedrock's model deployment accounts; a customer-owned FormKiQ stack is not a promise that model computation occurs inside the customer's account.
Select the model or inference profile deliberately. Cross-region inference can route requests outside the source region. Inspect the permitted destination regions and organizational policies before using sensitive documents. Review Bedrock data protection and cross-region inference for the AWS processing model.
Validate one document analysis
Create a named prompt, use fictional content, and inspect the resulting summary or extracted fields. Define an explicit missing-evidence result and a human review path. Document text must be treated as evidence rather than trusted instructions to run tools or disclose data.
Protect prompts, completions, OCR text, logs, and cached results as document-derived data. Keep source and prompt/model versions where reproducibility matters. An external AI client connected through MCP has its own processing boundary to review.
Edition and configuration
AI Processing and Analysis is a selected Advanced/Enterprise extension. Model access, inference profiles, IAM permissions, regional availability, implementation, and AWS usage are separate configuration and cost decisions.