IskayPet cuts answer time to 7 seconds with Amazon Bedrock

IskayPet's starting point was a familiar corporate problem: answers to everyday operational questions lived in hundreds of Microsoft SharePoint documents spread across four languages and three countries, and finding the right information took a long time. Rather than migrate that content, the team used the native SharePoint connector in Amazon Bedrock Managed Knowledge Bases, which syncs incrementally and can extract text from images embedded in documents. That choice kept the existing repositories in place and avoided a costly migration.

The design decision that shaped the rollout was to create a dedicated knowledge base per assistant, each scoped to specific SharePoint sites or libraries, rather than one shared index with country and area tags. According to the post, this keeps each body of knowledge isolated end to end and lets new areas or countries be added by pointing a new assistant at the right content instead of re-architecting. Access control follows the same logic: an employee sees only the assistants whose documentation they are entitled to read, and retrieval enforces document-level permissions natively.

On the build side, IskayPet used Kiro with Claude Haiku 4.5 to generate the frontend and provision the AWS services, which the post says helped a small team move from concept to production quickly. The team is now adding Amazon Bedrock Guardrails for content filtering and responsible AI policies, and is evaluating prompt caching and batch inference as usage grows toward the expected roughly 100,000 queries per month.