Dispatch  /  The crew
CREWJob roles

Ten agents. Ten job descriptions.

An agent with a vague role improvises, and improvisation is the thing you are buying this desk to eliminate. Each agent is written as a role your supervisor could hand to a new hire — owns, decides alone, must escalate, measured on.

10 at go-live25 at ceiling3 autonomy classes
CREW-01

Dispatch AI  Supervised

Booking intake, allocation and ETA

Takes ride requests from phone, app and web; matches the rider record; places, amends and cancels bookings in the operator's dispatch system; answers allocation and ETA questions.

Decides alone

Place, amend and cancel bookings within operator policy. Re-allocate a car. Quote an ETA. Read prior trip and access notes.

Must escalate

Anything involving a fare change, a driver complaint, a route the rider disputes, or a booking the system refuses twice.

Measured on

Booking accuracy, ETA quote accuracy, first-response time on live ride issues.

CREW-02

Watch AI  Supervised

Live queue and anomaly monitoring

Watches the live board: queue depth, unallocated rides, cars not moving, ETA slippage, no-car conditions in a zone, and channel backlogs. Raises before anyone complains.

Decides alone

Re-check allocation, prompt a proactive rider message, raise a no-car alert to the operator, flag queue ageing to your supervisor.

Must escalate

Any pattern suggesting a systemic outage, and any anomaly with a safety signature — handed straight to Safety AI and then to a human.

Measured on

Proactive contact rate, queue ageing, time from anomaly to alert.

CREW-03

Rider Support AI  Draft only

Rider-side queries and complaints

Pre-ride and in-ride questions, lost property, service complaints, refund requests and ride history queries across app, chat, email and voice.

Decides alone

Nothing autonomously. It drafts the reply and the proposed action with the policy citation attached.

Must escalate

Everything reaches a rider only after release. Complaints alleging driver misconduct go to Safety AI first, not to a reply queue.

Measured on

Draft acceptance rate, edit distance on released replies, rider CSAT on resolved contacts.

CREW-04

Driver Support AI  Draft only

Driver-side queries and app issues

App and device problems, document expiry reminders, trip and earnings queries, cancellation disputes, shift and incentive questions. Drivers are a customer here, not a resource.

Decides alone

Nothing autonomously. It drafts recovery steps, opens reconstruction cases and prepares document reminders.

Must escalate

Account status questions of any kind, earnings disputes above the ceiling, and anything a driver frames as unfair treatment.

Measured on

Draft acceptance rate, driver repeat-contact rate, time to first response.

CREW-05

Onboarding AI  Supervised

Driver documents to activation

Collects licence, insurance, permit and vehicle documents; checks completeness and expiry; chases every 48 hours; prepares the activation pack when the checklist closes.

Decides alone

Chase, check completeness, flag expiry, order the pipeline, prepare the pack.

Must escalate

Any judgement on whether a document is genuine, any background or right-to-work determination, and the activation decision itself.

Measured on

Time to activation, pipeline ageing, document rejection rate at operator review.

CREW-06

Fare AI  Draft only

Fare disputes and adjustments

Reconstructs disputed fares from trip data, diversions, waiting time, tolls and meter faults; prepares an adjustment with the evidence attached.

Decides alone

Nothing autonomously. It prepares. Under the ceiling, your supervisor releases; above it, a named approval is required in code.

Must escalate

Every adjustment above $25. Any dispute alleging deliberate overcharging. Any pattern suggesting a fare-structure fault rather than a one-off.

Measured on

Reconstruction accuracy, dispute reopen rate, adjustment value against benchmark.

CREW-07

Payments AI  Draft only

Payouts, wallets and settlement

Driver payout and wallet queries, incentive calculations, settlement reconciliation against the operator's ledger, and short-payment identification.

Decides alone

Nothing autonomously. It reconciles and drafts.

Must escalate

Any funds movement, any payout hold or release, and any discrepancy above the ceiling.

Measured on

Reconciliation accuracy, unexplained variance, query resolution time.

CREW-08

Safety AI  Escalate only

Incident triage and immediate handover

Watches for trip anomalies, SOS signals, accident notifications and conduct complaints. Classifies, collates trip record, route, both parties and history — and stops.

Decides alone

Nothing. It has no reply tool and no closure tool. Triage and handover are its entire function.

Must escalate

Everything, within 60 seconds, at any hour, to a named human. There is no threshold and no exception.

Measured on

Time from signal to human, classification accuracy, false-negative rate (target zero).

CREW-09

Fleet AI  Supervised

Fleet partner accounts and compliance

Fleet partner account setup, vehicle register maintenance, inspection and compliance expiry tracking, and fleet-level reporting to the operator.

Decides alone

Maintain the register, track and flag expiry, prepare fleet reporting, chase compliance items.

Must escalate

Any decision to suspend a fleet or a vehicle, and any commercial term with a fleet partner.

Measured on

Register accuracy, compliance expiry lead time, fleet reporting timeliness.

CREW-10

Quality AI  Supervised

Rating patterns and coaching flags

Watches rider ratings, complaint themes and cancellation patterns by driver and by zone; raises coaching flags to the operator early rather than at review time.

Decides alone

Compile patterns, raise coaching flags, produce zone and driver reporting.

Must escalate

Any recommendation affecting a driver's standing, earnings or account. Those go to the operator through your supervisor, never direct.

Measured on

Flag precision, lead time before an issue becomes a complaint.

Beyond ten

Specialists, added when volume justifies them.

Bought in blocks of five for ₹1,85,000 when allocated volume grows. A block adds about ₹1,06,000 a month after its own running cost, repaying itself in under two months. You never carry crew you are not being paid to run.

SPEC A

Time-zone specialists

Agents tuned to a specific market's idiom, address formats and service expectations — a US desk and a UK desk do not sound the same, and riders notice.

SPEC B

Airport & venue desks

Dedicated handling for airport terminals, stadiums and event venues, where access rules, pickup bays and surge behaviour are their own discipline.

SPEC C

Corporate & account rides

Business accounts, billing codes, cost-centre allocation and monthly invoicing for operators with a corporate book.

SPEC D

Claims support

Evidence collation for insurance and damage claims — collation only. No claim is ever accepted, denied or commented on by any agent.

SPEC E

Bench auditor

An internal agent that samples the crew's own output ahead of the weekly QC pass, catching drift before it reaches a scorecard.

SPEC F

Operator reporting

Client-facing performance packs, anomaly narration and month-end reporting for every operator you hold.

Roles on paper are one thing. Watch them run.

A ride-along shows the crew working the live queue, the review lane filling, and Safety AI refusing to touch a flag it has just raised.