Call center ai · Production

Spare Labs deploys Retell AI voice agents to automate public transit call centers

The problem

Transit agency call centers serving paratransit riders with disabilities were chronically understaffed and outsourced to third parties, resulting in long hold times that prevented riders from booking trips or planning their day.

First attempt

Spare's 2023 attempt to integrate AI voice models did not meet quality requirements and the project was shelved.

Workflow diagram · grounded in source
1
Rider initiates call
trigger
“a phone call is the only way to book a trip, check an ETA, or get an update”
2
Configurable agent handles call
ai_action
“the team built a single, configurable agent that can be turned on for any transit agency in a click. The agent already knows transit flows, ride-booking logic, and the tool calls required to reach into Spare's platform.”
3
Platform data lookup
integration
“The agent integrates with a single system of record: Spare itself. Because Spare is the operating system for the transit agency, every data point the agent needs—ride status, driver location, account info, service changes—is already avai…”
4
Human agent fallback
routing
“Most agencies deploy the agent as a first line of defense, with a fallback to human agents for edge cases.”
5
Prompt and model iteration
feedback_loop
“The lift came from steady iteration on prompts, tool calls, and voice quality paired with rapid model improvements from Retell.”
Reported outcome

Within the first week of deployment, Spare's voice agents handled around 30% of inbound calls, immediately cutting hold times.
Average call resolution now sits close to 70%, reaching close to 100% for simpler operations. The AI product line grew over 10x year-over-year, and agents successfully absorbed a record snowfall service surge that would have overwhelmed a human call center.

Reported metrics
Inbound calls automated in first weekaround 30%
Average call resolution rateclose to 70%
Call resolution at top-performing customersclose to 100%
AI product line YoY growthover 10x growth
Show all 5 reported metrics
inbound calls automated in first weekaround 30%
average call resolution rateclose to 70%
call resolution at top-performing customersclose to 100%
AI product line YoY growthover 10x growth
hold time reductionimmediately cutting hold times
Reported stack
RetellSpare
Source
https://www.retellai.com/case-study/how-sparelabs-put-ai-voice-on-the-front-line-of-public-transit-and-watched-it-take-off
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Within the first week of deployment, Spare's voice agents handled around 30% of inbound calls, immediately cutting hold times.

What tools did this team use?

Retell, Spare.

What results were reported?

Inbound calls automated in first week: around 30%; Average call resolution rate: close to 70%; Call resolution at top-performing customers: close to 100%; AI product line YoY growth: over 10x growth (source-reported, not independently verified).

What failed first in this deployment?

Spare's 2023 attempt to integrate AI voice models did not meet quality requirements and the project was shelved.

How is this call center ai AI workflow structured?

Rider initiates call → Configurable agent handles call → Platform data lookup → Human agent fallback → Prompt and model iteration.

Related call center ai cases