IT Operations Automation · Security and Compliance
Okta Inbox needed a more scalable and lower-overhead way to power natural-language application access requests without maintaining separate vector infrastructure or slowing down product iteration.
Reported impact
Okta simplified its identity-governance search stack, improved developer focus on product innovation, and is on track to reduce operating costs by 30% while preserving a strong customer experience.
To modernize Okta Inbox, Okta replaced a home-grown vector workflow that depended on custom machine learning models and a dedicated Kubernetes cluster with MongoDB Atlas and MongoDB Vector Search on Atlas. The product helps end users request access to applications in natural language, so the system has to understand phrases that describe the same tool in very different ways. Previously, embeddings and search logic were managed separately, which increased infrastructure cost and reduced room for experimentation. Okta moved embeddings into Atlas collections and used MongoDB Vector Search on Atlas to query them directly alongside application and ticketing data. This gave the engineering team a single operational data layer for structured records plus semantic retrieval, enabling automated app matching and ticket creation from chatbot requests. By removing extra infrastructure, the team could iterate faster while keeping the platform resilient as adoption grew.
User requests arrive through chatbots, embeddings and application metadata are stored together in MongoDB Atlas, MongoDB Vector Search on Atlas retrieves the best matching app from those embeddings, and Okta Inbox automatically raises the access ticket for downstream processing.
“It receives thousands of requests through our chatbots, and we no longer need to manually process thousands of text embeddings to resolve those queries.”
“MongoDB Atlas provided one efficient data layer that simplified these operations.”
“MongoDB Vector Search on Atlas was the answer to our problems. It simplifies a lot of the work that goes into making Okta Inbox super user-friendly for customers.”
“We’re also on track to reduce operating costs by 30% compared to the self-hosted solution.”
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
Google Cloud Sample Modified · Oct 9, 2026 Replenit Legacy marketing tools often treat individual shoppers as spreadsheet averages, relying on broad customer segments and predictable, rigid automation rules. As a result, brands frequently blast customers with irrelevant,… 360% business growth in one quarter with Google Cloud BigQuery · Cloud TPUs · Gemini Google Cloud Sample Modified · Oct 9, 2026 Riafy Technologies Riafy Technologies began its journey with a clear mission: to build adaptable and accessible AI systems that operate natively at a global scale. Today, the company's consumer-end products support 125 million people… Cut incident diagnostics from 40 minutes to under 1 minute Cloud Run · Cloud SQL · Gemini Models 
OpenAI Sample Oct 9, 2026 Sophos Sophos needed to investigate and respond to a growing volume of cyber threats faster, while scaling its expertise across every protected customer without matching growth in scarce analyst headcount. 52% of MDR cases resolved end-to-end by AI OpenAI Daybreak · OpenAI models · Sophos Fusion 
