Document Processing
Databricks needed a model that could supervise enterprise agent workflows involving scanned PDFs, legacy files, and long-context documents without the parsing and retrieval errors that frequently break production systems.
Reported impact
Databricks improved the reliability of enterprise document agents, surpassed prior benchmark performance, and gave customers a production path for GPT-5.5-supervised workflows that require stronger parsing and grounded retrieval.
Databricks validated GPT-5.5 on OfficeQA Pro, its benchmark for complex enterprise document tasks, and then exposed the model for production use inside customer agent workflows. The benchmark stresses parsing, retrieval, and grounded reasoning across scanned PDFs, legacy files, and long-context enterprise documents, where a single extraction error can derail downstream execution. After GPT-5.5 reduced errors by 46% versus GPT-5.4 and became the first model to exceed 50% accuracy on OfficeQA Pro, Databricks made it available through AI Unity Gateway for workflows built with AgentBricks and the Agent Supervisor API. In that architecture, GPT-5.5 supervises specialized agents handling parsing, retrieval, and execution so customers can run more reliable multistep document workflows with fewer search detours and less manual intervention.
Scanned PDFs, legacy files, and long-context documents enter agent workflows built with AgentBricks. GPT-5.5, delivered through AI Unity Gateway and coordinated by the Agent Supervisor API, supervises parsing, retrieval, and execution steps before grounded answers or task completions are returned.
"Databricks now makes GPT‑5.5 available through AI Unity Gateway"
"GPT‑5.5 reduced errors by 46% compared to GPT‑5.4 and became the first model to surpass 50% accuracy on OfficeQA Pro."
"In these systems, GPT‑5.5 orchestrates parsing, retrieval, and execution across specialized agents."
"OfficeQA Pro evaluates how models handle parsing, retrieval, and grounded reasoning across workflows involving scanned PDFs, legacy files, and long-context documents"
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
OpenAI Sample Sep 3, 2026 Invideo The AI agent must translate the editor’s direction into a sequence of steps, choose the right tools, execute the work, and verify the result. Sanket says GPT‑6 Astra can plan these operations in fewer reasoning steps… The AI agent must translate the editor’s direction into a sequence of steps, choose the right tools, execute… GPT-6 Astra OpenAI Sample Aug 20, 2026 Stampli Stampli needed to launch complex new products and keep product knowledge current without waiting on scarce design, contractor, and operations bandwidth. 243 modeled active role-hours compressed to about 77 for the Deep Finance launch ChatGPT Work · Codex · Deep Finance OpenAI Sample Jun 12, 2026 Preply Preply wanted to personalize language learning more deeply and reduce repetitive tutor work without replacing the human relationship that makes one-to-one teaching effective. 4.7/5 satisfaction rating Brand Voice GPT · ChatGPT Enterprise · Codex 