Aderant has put an AI ticket-triage system into production for its SierraOps team, using Amazon Nova Lite through Amazon Bedrock to gather context, classify support requests and route work. In its first 2.5 weeks, the company says the system reviewed 109 tickets with approximately 96% routing accuracy, while recovering an estimated 8 to 14 engineering hours each week.
AWS AI Watch analysis
What happened
The system reviews newly submitted, unassigned tickets during scheduled hourly runs on business days. It pulls together information from Jira, Confluence, Amazon Athena and Microsoft SharePoint, then recommends a team assignment and a starting point for resolution.
Aderant says the workflow can also reassign misrouted tickets, notify the relevant Microsoft Teams channel, post useful knowledge into tickets and apply predefined actions for issues that can be handled by support staff. Lower-confidence classifications are sent for human review rather than being waved through by an overconfident robot with a clipboard.
The company estimates the system costs less than US$30 a month to run, with Bedrock inference costing less than US$1 a month during the initial period. AWS published the account as a guest post co-written by Aderant's Angela Mapes and Adam Walker.
Key findings
- 109 tickets analysed
The early production sample covered 2.5 weeks of live operation from 30 June to 17 July 2026. - Approximately 96% routing accuracy
Aderant counted four misroutes by comparing the system's assignment with the team that ultimately resolved each ticket. - 8 to 14 engineering hours recovered weekly
The estimate is based on a previous manual triage burden of 15 to 25 minutes per ticket. - Human review remains in the loop
Tickets below a configured confidence threshold are flagged rather than acted on automatically. - No client matter data processed
Aderant says the analyser uses operational ticket data, metadata and internal knowledge, not client application business data or client matter data.
Why it matters
This is a useful example of where enterprise AI is becoming less theatrical and more operational. The system is not attempting to solve every support problem. It handles the repetitive investigative work that delays an engineer before resolution can begin: finding the right context, identifying ownership and locating relevant documentation.
The claimed savings are meaningful for a 38-person team operating across 268 client environments, but they are still early results from one company's deployment. A 96% routing rate over 109 tickets is encouraging evidence, not a universal benchmark for AI support systems.
Our read
The strongest part of Aderant's approach is the division of labour. Nova Lite is used to assemble and recommend, while monitoring, confidence thresholds and human review provide brakes when the system is less certain. That is a more credible route to automation than asking a model to impersonate an entire operations department from day one.
The real test will be whether accuracy holds as ticket types change, whether the knowledge loop stays current and whether engineers trust the recommendations enough to keep using them. For teams considering similar systems, the practical lesson is to measure misroutes and time saved before celebrating the dashboard. A small, governed workflow may beat a grand AI transformation plan hands down.
What to watch
- Whether Aderant reports results over a longer period and a larger ticket sample.
- How often low-confidence cases are escalated to human reviewers.
- Whether the recurring-issue knowledge loop reduces repeat investigation work.
- Whether the workflow expands into deeper client-history analysis or semantic knowledge matching.
Discussion spark: Should companies automate ticket routing once the early accuracy looks strong, or should they require longer-term evidence before allowing AI to act on operational work?
Sources and evidence
- Aderant builds intelligent ticket triage with Amazon Nova (24 September 2026, 16:06 UTC)
not affiliated with or endorsed by Amazon Web Services (AWS)