New perspective · 14 Sep 2026Why AI agents stall after the pilot — and what production actually requires→
Optivaco
Insights / Case studies / Aerospace & Defense
A Fortune 500 aerospace manufacturer

A maintenance knowledge assistant for MRO technicians.

INDUSTRYAerospace & Defense
CAPABILITIESGenerative AI · Data & AI Platforms · Managed Pods
ENGAGEMENTDelivery partner + managed pod
DURATION6 months to production
01 — Challenge

Maintenance knowledge was locked in decades of manuals and service records.

02 — Approach
01Assess the knowledge baseCatalogued manuals, service bulletins, and records by format, sensitivity, and ownership.
02Architect for securityDesigned retrieval with document-level access control inside a secure cloud tenant.
03Build with techniciansDeveloped the assistant against real diagnostic tasks, with technicians reviewing every release.
04Staff and runA dedicated pod owns content ingestion, evaluation, and ongoing improvement.
03 — Tech stack
AWS GovCloudAmazon BedrockAnthropic ClaudeOpenSearchSAPPythonOpenTelemetry
04 — Outcomes
35%reduction in time-to-diagnosis
120Kdocuments made searchable
92%of queries answered with a cited source
Illustrative figures, pending client approval.
Your program

Have a pilot that should be in production?

Tell us where it stalled. We’ll tell you what it would take.