All case studies
Government · Transit·2024
National Urban Transit Network

AI service analytics that cut ticket resolution by four-fifths.

Service analytics and automation across a major urban transit system. Inbound calls triaged, routed, resolved or escalated with a full audit trail.

Engagement
Duration
8 months
Squad
Delivery partner · 4 engineers · 2 AI engineers · 1 operations lead
Status
Live in production
80%
Faster resolution
70%
Calls automated
11
Languages
Real-time
Dashboards
The problem

What we walked into.

A major urban transit network was being buried under its own call volume. Pass extensions. Route queries. Refund requests. Grievances. The support floor was staffed for peak load and still falling behind. No one could see, in real time, what the network was actually asking the support team to do.

Our approach

What Safemode did.

  1. 01

    Stood up a multilingual voice front-end that triages every inbound call across 11 languages, with fall-back to a human agent where confidence drops below a policy threshold.

  2. 02

    Automated the three highest-volume call reasons end-to-end (pass extensions, fare queries, refund status) with policy-as-code governing what could be resolved without a human.

  3. 03

    Built a real-time operations dashboard so the network's operations team can see volume, reason mix, containment rate and escalation patterns by station and by hour.

  4. 04

    Wired a full audit trail (who asked what, what the agent did, what policy fired) back into the transit network's core systems.

Outcome

What moved.

  • Average ticket resolution time fell by roughly 80% across the top three call reasons.

  • Roughly 70% of calls are now contained entirely inside the AI layer without a human touch.

  • The operations team sees demand shifts in minutes rather than at the end-of-month report.

  • Human agents have been redeployed against the harder, higher-trust escalations the network actually needed them for.

10 · Start here

Let’sbuildyoursystemnext.

Thirty minutes with someone who’d be doing the work. No slide deck, no intake form. We’ll tell you what’s feasible, where you’ll hit friction, and what we’d pick up first.

Response
< 24 hours
First read
No NDA needed
Bangalore / Remote
UTC ±12