Aderant Cuts Ticket Triage Time 8-14 Hours Weekly Using Amazon Nova Lite

Aderant, a business management software provider for the legal industry, has deployed Amazon Nova Lite through Amazon Bedrock to automate support-ticket triage for its 38-person SierraOps team, which manages Expert Sierra across 268 client environments globally.

According to a joint case study published by Aderant and AWS, the resulting system — called the Intelligent Ticket Analyzer — reviewed 109 tickets during its first 2.5 weeks in production (June 30 to July 17, 2026) and achieved approximately 96% routing accuracy, with four misroutes. Aderant estimates the system recovers 8-14 engineering hours per week, or 32-56 hours per month, based on a measured pre-automation baseline of 15-25 minutes of manual triage per ticket.

How it works

The system runs as a serverless workflow: a single AWS Lambda function, triggered hourly on business days by Amazon EventBridge, processes unassigned tickets in Aderant's CloudOps queue through five stages:

1. Identify — pull unassigned tickets from Jira
2. Enrich — gather client metadata from Amazon Athena, documentation from Confluence and SharePoint, and comparable resolved tickets from Jira
3. Classify — send the ticket and assembled context to Amazon Nova Lite via the Amazon Bedrock Converse API for structured classification
4. Act — post analysis, reassign misrouted tickets, notify Microsoft Teams channels, and apply predefined actions
5. Observe and improve — log metrics to Amazon CloudWatch and route lower-confidence classifications to human review

High-confidence routing decisions execute autonomously; tickets below a configured confidence threshold are flagged for manual review. Aderant's team reviews routing corrections weekly to refine prompts and logic.

Why Nova Lite, according to Aderant

Aderant says it evaluated multiple foundation models through Bedrock using real ticket data before selecting Nova Lite, citing three factors:

• Operational specificity: Nova Lite reportedly extracted more specific resolution steps and correlated past issues more effectively than other evaluated models on identical ticket content, environment data, and runbooks.
• Native AWS integration: The model works directly with the Bedrock Converse API, IAM permissions, and the existing AWS SDK, avoiding an additional vendor contract.
• Cost at scale: Total system operating cost stayed under $30 per month, with Bedrock inference costs under $1 per month — low enough, Aderant says, to justify analyzing routine tickets rather than reserving automation for high-priority cases only.

Data boundaries

The analyzer operates only on internal operational ticket data and metadata. According to Aderant, it does not access, process, or store client matter data or client application business data.

What this means

This is a deployment case study, not a model release — Amazon Nova Lite itself is unchanged. The notable figures here are operational rather than architectural: a 268-client-environment support operation running full ticket classification for under $1 in monthly inference cost illustrates how cheap small multimodal-capable models have become for structured, repetitive enterprise tasks that don't require frontier reasoning.

The 96% routing accuracy claim comes from a 2.5-week sample of 109 tickets — a genuinely early result, as Aderant itself notes, not a validated long-term benchmark. The four misroutes in that window matter less than whether the confidence-threshold system continues to catch edge cases as ticket volume and variety grow. The broader signal for enterprise buyers is the widening gap between frontier-model pricing for complex reasoning and the near-negligible cost of running smaller models like Nova Lite for high-volume classification and routing work — a gap AWS is clearly positioning Nova to exploit against competitors' equivalent lightweight tiers.