Customer story · Public sector
A public-sector organisation
Ticketing and service automation for a public-sector organisation
Sovereign service operations inside the organisation's environment. Requests are validated, triaged by a locally hosted model, tracked against SLA policy and escalated to people whenever the model is unsure.
Executive summary
The situation
Citizen and employee requests needed a clear route through service teams without sending their content to an external model. Validation, prioritisation and SLA policy had to work together inside the organisation’s environment.
What was built
An intake form, multi-user dashboard and read-only support assistant sit above Nexus workflows. Local AI proposes triage decisions with confidence scores; rules persist tickets, notify teams and track service policy.
Why it matters
Gives service staff a shared view of requests and their priority. Low-confidence decisions reach a person, so AI assistance does not mean handing over control of service delivery.
- Primary users
- Service teams and human reviewers
- Core systems/data
- Requests, ticket data and local models
- Human control point
- Human review of low-confidence triage
- Nexus role
- Validation, triage, routing and SLA tracking
The challenge
Service requests that cannot leave the building.
A request form captures the problem, but does not by itself make the submission useful work. It still needs enough information, a route to the relevant team and a priority staff can understand. Citizens and employees depend on that handoff.
For this organisation, AI assistance also had to work inside its government/private-cloud environment. The challenge was to connect intake, triage and SLA tracking without separating automation from the people accountable for the service.
What we built
An intake-to-resolution service flow with local AI.
The business experience starts with an intake form and continues in a multi-user dashboard. Between them sits a Nexus workflow that validates the submission, asks a locally hosted model to classify and prioritise it, and records a confidence score. Low confidence is a reason to ask a person, not push a doubtful classification further down the process.
The ticket is persisted and the relevant team notified. Encoded SLA policies give the dashboard a basis for tracking requests. Staff can also ask a support assistant about ticket data, but answering a question does not give that assistant permission to change the work. Its access is read-only, and no external LLM API is required.
How the workflow runs
Step by step.
01
Intake
A citizen or employee submits a request through the intake form. The workflow checks the request for required information before it is accepted.
02
Triage
A local model classifies and prioritises the request and records a confidence score. Low-confidence decisions go to a human reviewer instead of being applied automatically.
03
Persist and notify
The ticket is stored and the responsible team is notified. The dashboard shows each request against the SLA policy that applies to it.
What the business team sees
One dashboard for service teams.
Service staff need to see requests, not the mechanics of model execution. The dashboard brings priority, SLA position, assignments and items awaiting review into a shared view. The support assistant offers another way to read ticket information, while changes remain with staff and the workflow rules they own.
- Incoming requests by priority
- SLA position per ticket
- Low-confidence triage review queue
- Team assignments and notifications
- Read-only support assistant
How Nexus works underneath
Service rules connect the form to the team.
Nexus carries validation, confidence-based routing, ticket persistence, notifications and SLA tracking between the business screens and the ticket store. The model’s role is narrower: it proposes classification and priority, but does not own service policy. Read-only access likewise keeps the assistant in an information role rather than a record-changing one.
Architecture
How the layers fit together.
Business team
Intake form and service dashboard
Requests · SLA tracking · review queue · support assistant
Nexus orchestration
- 01Validate request
- 02Local AI triage
- 03Confidence routing
- 04Persist ticket
- 05Notify team
- 06SLA tracking
Connected systems and data
- Ticket store
- Locally hosted models
- Team notifications
Human control and audit
People decide when the model is unsure.
Every triage decision carries a confidence score. Below the configured threshold, the request goes to a human reviewer, and the reviewer's decision is what the ticket follows.
Because the assistant only reads, it cannot alter tickets, priorities or assignments. Changes remain with service staff and the workflow rules they own.
What this changes
A shared service view. Gives teams one place to inspect requests, priority and SLA position.
Uncertainty has a route. Low-confidence triage becomes review work rather than an unquestioned decision.
AI stays within the environment. Local models support assistance without requiring an external model API.
Your process
Discuss a similar workflow
Tell us about the documents, systems and decisions involved. We will discuss where Nexus orchestration fits and where your team stays in control.