- Braze cut content translation cycles by 98% after consolidating four separate content workflows onto a single Sanity.io content model, replacing a manual copy-paste process that consumed hundreds of hours per learning path.
- Sanity’s agent infrastructure powers Braze’s custom AI agents for automated summaries, FAQ generation and AEO metadata, with the core automation built in 30 lines of code; Sanity’s own case study cites “0 engineering tickets” needed to launch a new agent.
Four separate content workflows, hundreds of hours of manual translation work per learning path, and a search integration that took weeks to update: Braze replaced all of it with a single Sanity.io content model and a 30-line automation. The 98% reduction in translation cycle time, detailed in Sanity.io case study material published June 29, 2026, is the sharpest number in what amounts to a content operations rebuild.
Previously, Braze’s content teams managed learning paths, blog posts and marketing pages across multiple platforms, each with its own rigid structure. Content could not be reused across contexts, translation required manual copy-paste work across dashboards, and generating structured AEO metadata fell to writers producing it by hand. Engineering tickets were the bottleneck for anything that crossed system boundaries.
Braze’s response was to implement Sanity as a centralised content layer, covering search, translation, automation and internal AI agents across the organisation. The company refers to this as “AI Content Operations.” Sanity’s API-first Content Lake allowed flexible content modelling, while its GROQ query language simplified delivery to downstream tools.
AEO Metadata on Publish
The first workflow to change was AEO metadata generation. When a blog post publishes, a Sanity Function now triggers an AI agent that automatically produces a summary, a FAQ set and a machine-readable description for schema markup. The entire automation runs in 30 lines of code, removing a task that previously required manual writer input for every piece of content.
Translation Without the Dashboard Work
The translation overhaul replaced a costly third-party integration with a custom plugin built on Sanity Functions. Editors select target languages and a translation model; the job runs automatically and ties into Content Releases for human review. The step that previously consumed the most time, configuring each lesson, course and URL inside the learning management system dashboard by hand, is now handled by a function that fires on publish. According to Sanity.io’s case study, the new pipeline cut translation shipping time by 98%.
Customer Success Gets Its Own Agent
Braze built a dedicated AI agent for its customer success team using Sanity Context, pulling data from Salesforce, Gong call recordings, support tickets and product request forms. Customer success managers can now query it for relevant case studies and customer insights in seconds, without filing an engineering ticket. The agent’s configuration lives in Sanity documents, which means non-technical teams can build and deploy new agents independently. That last point has real operational weight: the engineering bottleneck that previously gated every new integration is removed from the content team’s workflow entirely.
Search Rebuilt in 10 Minutes
The previous learning content search relied on a rigid LMS API that returned only basic metadata. Braze’s web engineering team replaced that integration with a single GROQ query, written in roughly 10 minutes, that pulls full lesson content and converts Sanity’s Portable Text into embeddable strings. The result, per the case study, was higher-quality embeddings and more relevant search results. Retrieval quality depends directly on how well the underlying content is modelled, which is why structured content architecture has become central to agent performance.
The federated search improvement illustrates the compound effect of the whole migration: a single queryable API now handles what previously required multiple platform-specific integrations, each maintained separately. Whether the same content model scales as Braze adds languages, products and agent use cases is not something the case study addresses.
