Cornerstone OnDemand cuts database diagnosis time 78% with Orion AI on Amazon Bedrock

From reactive firefighting to coordinated agents

Cornerstone OnDemand, Inc. describes itself as a global leader in workforce readiness solutions, serving 140 million users across 186 countries. Its Enterprise DataOps team used to work reactively, according to the AWS post. Investigating a database performance problem took about 45 minutes per incident and was spread across several tools and system views.

The post lists four recurring pain points:

• Database performance investigations took approximately 45 minutes per incident.
• Database lifecycle workflows required 10 or more manual steps, from establishing connections to communicating status updates.
• Reporting between the site reliability engineering (SRE) team and the data team ran on a 15-minute lag.
• Redundant and overlapping alerts created noise that buried the signals that mattered.

As a result, engineers spent their time on manual coordination instead of resolving issues.

How Orion AI works

Orion AI is a multi-agent system in which a coordinating agent, the meta-orchestrator, delegates work to specialized child agents in a hub-and-spoke topology. Engineers use it through a web application, where they ask questions and approve actions. The agents cover infrastructure monitoring, database diagnostics, database lifecycle operations, customer analytics and knowledge-driven support. In total, Orion AI uses 13 domain-specific agents.

Amazon Bedrock provides managed access to foundation models, and Strands Agents, an open source agent orchestration framework from AWS, handles orchestration. The services run as containers on Amazon Elastic Container Service. Two tenets guided the build: data privacy using AWS controls under the shared responsibility model, and deep integration with existing operational tools.

For routing, the system tries keyword matching first. According to the post, approximately 80 percent of queries resolve through an in-memory keyword fast-path in under a millisecond. Ambiguous requests fall back to semantic search powered by Amazon Titan Text Embeddings V2. Where direct tool invocation isn't appropriate, Amazon Bedrock Knowledge Bases retrieves operational documentation so responses are grounded in approved procedures.

Measured results

Cornerstone reports the following outcomes:

• Database diagnosis time: from 45 minutes to 10 minutes, 78% faster.
• Manual lifecycle steps: from 10+ steps to 1 interaction, a 70% reduction.
• SRE-to-data-team reporting lag: from 15 minutes to instantaneous.
• Redundant alerts: a 65% reduction (median).

Previously, engineers connected to the affected SQL Server instance, queried system views for blocking chains and wait types, cross-referenced logs to isolate long-running queries and correlated the evidence to form a root cause hypothesis. Now three specialized agents share the investigation, each owning a distinct stage. Orion AI also identifies the root cause, recommends a fix and creates a populated Jira ticket assigned to the right on-call engineer, collapsing a four-step manual handoff into a single interaction.

On alerts, the post says Orion AI uses deduplication, threshold filtering and cross-signal correlation. For every 10 alerts previously generated, only 3 to 4 now reach engineers, and alerts go directly to the responsible teams without an intermediate alerting layer.

Design principles other teams can reuse

AWS highlights three decisions it says can apply to other multi-agent projects:

• Split agents by domain, not by task complexity. Each agent carries a narrow set of tool integrations for one operational domain, which keeps the model's context focused and improves tool selection accuracy.
• Default to keyword routing, fall back to semantic search. This keeps latency low without sacrificing correctness on harder queries.
• Scope conversational memory to sessions and bypass it for live metrics. Operational questions always read current system state, which helps prevent stale data from contaminating real-time diagnostics.

In the Strands Agents implementation, the hub is a single agent that holds no domain tools, only routing and control-flow tools. The spokes are independent agents, each with its own model, domain tools and localized system prompt.

What it means

The case gives teams running database operations a concrete reference for agent-based automation, with figures reported by the company itself. From the engineer's perspective, the work is reduced to prompting Orion AI and validating the findings it reports back. The post also notes that a three-person team delivered the system in six months, which indicates the scale of effort Cornerstone needed for this approach.