Download Your Guide to AI-Powered Taxi App Development
                     Ratings as at October 2026
V T  D I G I T A L  P L AY B O O K  ·  O C T O B E R  2 0 2 6
The AI Ride
Hailing Playbook
20 AI Features Ranked by Cost
and Payback
For Australian taxi and rideshare operators adding AI
to an app they already run, and founders building one.
VT Digital · vtdigital.com.au October 2026
H O W  T O  R E A D  T H I S  P L AY B O O K
B E F O R E  YO U  S TA R T
I N  T H I S  P L AY B O O K
How to read this playbook 03 All 20 features at a glance Ranking
05 Matching and dispatch 6 · 15 · 10
Every feature carries five tags. They’re our assessment of
07 Arrival times and demand 5 · 2 · 13
the work involved, so use them as a starting point for your
own numbers. 09 Driver supply and fares 12 · 9 · 7
11 Pricing and payments 18 · 14 · 3
13 Fraud and safety 4 · 16 · 20
15 Support 1 · 8 · 17
17 Drivers and compliance 11 · 19
19 Where to start for your Stage guide
stage
20 Before you build Checklist
21 Talk to VT Digital Next step
TAG W H AT  I T  T E L LS  YO U R AT I N G S
Cost band How big the build is compared with Low means an API or vendor tool plugged
the other 19 features into your app. Medium means a model
Cost Medium trained on your own data inside one part of
the app. High means models working across
several systems in real time
Effort How much time your own team will Light, Moderate or Heavy
spend on data, operations and
Effort Light testing
Payback How soon the feature should move a Short shows up within a pilot of a few weeks.
number you already track Medium takes two or three months. Long
Payback Short needs a full year of data, or pays back in risk
avoided
Build call Whether to build it, buy it or Build, Buy or Mix
combine the two
Build call Mix
Privacy flag Whether the feature is likely to fall Shown only where it applies
under the Privacy Act’s automated
Privacy flag decision rules from 10 December
2026
READING THE DOTS Low · Light · Short Medium · Moderate High · Heavy · Long
No dollar figures per feature Features are numbered by rank
API pricing changes often and your data drives the Page 3 ranks all 20, pages 5 to 18 cover each one,
cost. For whole-app ranges, see our guide to app and page 19 shows where to start.
development costs in Australia.
vtdigital.com.au 02
A L L  2 0  F E AT U R E S  AT  A  G L A N C E
T H E  R A N K I N G
All 20 features at a glance
PAY BAC K  AC R O S S  T H E  20 Long · 19–20
Ranked by payback first, then by cost and effort. The features
near the top are quicker and cheaper to prove. Short · 1–8 Medium · 9–18
BUILD
RANK FEATURE COST EFFORT PAYBACK PAGE
CALL
01 Support ticket triage and auto-replies  Low  Light  Short Buy 16
02 Smart pickup points  Low  Light  Short Mix 8
03 Cash shortfall detection  Low  Light  Short Build 12
04 Promo and duplicate account detection  Low  Light  Short Mix 14
05 Arrival time correction  Medium  Moderate  Short Build 8
06 Acceptance-aware matching  Medium  Moderate  Short Build 6
07 Upfront fare accuracy  Medium  Moderate  Short Build 10
08 Refund and fare dispute agent  Medium  Moderate  Short Mix 16
09 Event and weather demand signals  Low  Light  Medium Buy 10
10 Heading-home matching  Low  Light  Medium Mix 6
11 Driver earnings explainer  Low  Light  Medium Build 18
12 Driver positioning prompts  Low  Light  Medium Build 10
13 Short-term demand forecasting  Medium  Moderate  Medium Build 8
14 Payment risk scoring  Medium  Moderate  Medium Mix 12
15 Peak and queue dispatch  Medium  Moderate  Medium Build 6
16 GPS spoofing and fake trip detection  Medium  Moderate  Medium Mix 14
17 Voice and phone booking  Medium  Moderate  Medium Buy 16
18 Dynamic pricing with caps  High  Heavy  Medium Build 12
19 Automated decision log  Low  Moderate  Long Build 18
20 Trip safety alerts  Medium  Moderate  Long Build 14
D E P E N D E N CY N OT  O P T I O N A L
Driver positioning prompts (12) need demand The decision log (19) ranks low on payback but isn’t
forecasting (13) running first. optional. If AI in your app makes decisions about
riders or drivers, start it before 10 December 2026.
vtdigital.com.au 03
PAY B A C K  M A P
T H E  R A N K I N G ,  M A P P E D
Where the payback lands
Ranked by payback first, then by cost and effort. The features near the top are quicker
and cheaper to prove.
Short 8 features Medium 10 features Long 2 features
Shows up within a pilot of a few weeks Takes two or three months Needs a full year of data, or pays back in
risk avoided
01 Support ticket triage and 09 Event and weather 19 Automated decision log
auto-replies demand signals Low · Moderate · Build
Low · Light · Buy Low · Light · Buy
20 Trip safety alerts
02 Smart pickup points 10 Heading-home matching Medium · Moderate · Build
Low · Light · Mix Low · Light · Mix
03 Cash shortfall detection 11 Driver earnings explainer N OT  O P T I O N A L
Low · Light · Build Low · Light · Build The decision log (19) ranks low on
payback but isn’t optional. If AI in
04 Promo and duplicate 12 Driver positioning your app makes decisions about
account detection prompts riders or drivers, start it before 10
Low · Light · Mix Low · Light · Build
December 2026.
05 Arrival time correction 13 Short-term demand
Medium · Moderate · Build forecasting
Medium · Moderate · Build D E P E N D E N CY
06 Acceptance-aware Driver positioning prompts (12)
matching 14 Payment risk scoring need demand forecasting (13)
Medium · Moderate · Build Medium · Moderate · Mix running first.
07 Upfront fare accuracy 15 Peak and queue dispatch
Medium · Moderate · Build Medium · Moderate · Build
08 Refund and fare dispute 16 GPS spoofing and fake
agent trip detection
Medium · Moderate · Mix Medium · Moderate · Mix
17 Voice and phone booking
Medium · Moderate · Buy
18 Dynamic pricing with
caps
High · Heavy · Build
Privacy flag Cost · Effort · Build call shown under each feature Select any feature to jump to it
vtdigital.com.au 04
S E C T I O N  0 1
01
Matching and dispatch
Dispatch decides which driver gets which job, so small improvements
here show up in almost every trip.
F E AT U R E S  I N  T H I S  S E C T I O N
06 15 10
Acceptance-aware Peak and queue dispatch Heading-home matching
matching
Medium cost · Medium payback · Low cost · Medium payback · Mix
Medium cost · Short payback · Build Privacy flag
Build Privacy flag
Privacy flag
vtdigital.com.au 05
M AT C H I N G  A N D  D I S PAT C H
S E C T I O N  0 1  ·  F E AT U R E S  6 ,  1 5 ,  1 0
I N  T H I S  S E C T I O N
Matching and dispatch 06 Acceptance-aware matching Short
Dispatch decides which driver gets which job, so small 15 Peak and queue dispatch Medium
improvements here show up in almost every trip.
10 Heading-home matching Medium
06 Acceptance-aware matching
Cost Medium Effort Moderate Payback Short Build call Build Privacy flag
Picks the driver most likely to accept and arrive quickly, using pickup time, which way the driver is
heading and how they’ve responded to similar jobs. Nearest car is only one of the inputs.
DATA  YO U  N E E D
WATC H  F O R
Accepted, declined and cancelled jobs with
This decides who gets work, so be ready to explain
timestamps, plus driver GPS
allocation job by job. The Fair Work rideshare
W H AT  TO  M E A S U R E application is still open
Cancellations after a driver is assigned, and time
to accept
15 Peak and queue dispatch
Cost Medium Effort Moderate Payback Medium Build call Build Privacy flag
At airports, stadium exits and the Friday late rush, the app holds requests for a few seconds and
matches them as a batch. That usually beats first-come, first-served when demand spikes.
DATA  YO U  N E E D
WATC H  F O R
Booking and driver location history from your
Airport pickup zones and rank rules, which vary by
busiest periods
airport and state
W H AT  TO  M E A S U R E
Average pickup time and unfilled requests during
peaks
10 Heading-home matching
Cost Low Effort Light Payback Medium Build call Mix Privacy flag
Offers drivers finishing a shift trips that run in the direction of home, so the last leg isn’t driven empty.
DATA  YO U  N E E D
WATC H  F O R
Driver-set destinations, plus trip origins and
Drivers will ask why they did or didn’t get a trip, so
destinations
show the reason in the driver app
W H AT  TO  M E A S U R E
Empty kilometres at the end of shifts, and driver
retention
TRY THIS WEEK Pull last month’s declined jobs and sort them by driver. If a handful of drivers
decline most of the work, start with acceptance-aware matching (6).
vtdigital.com.au 06
S E C T I O N  0 2
02
Arrival times and demand
Riders cancel when the arrival time they were shown turns out to be
wrong. Drivers sit idle when they're in the wrong part of town. Both
problems come down to prediction.
F E AT U R E S  I N  T H I S  S E C T I O N
05 02 13
Arrival time correction Smart pickup points Short-term demand
forecasting
Medium cost · Short payback · Low cost · Short payback · Mix
Build Medium cost · Medium payback ·
Build
vtdigital.com.au 07
A R R I VA L  T I M E S  A N D  D E M A N D
S E C T I O N  0 2  ·  F E AT U R E S  5 ,  2 ,  1 3
Arrival times I N  T H I S  S E C T I O N
and demand 05 Arrival time correction Short
02 Smart pickup points Short
Riders cancel when the arrival time they were shown
turns out to be wrong. Drivers sit idle when they’re in the Short-term demand13 Medium
forecasting
wrong part of town. Both problems come down to
prediction.
05 Arrival time correction
Cost Medium Effort Moderate Payback Short Build call Build
A model learns the gap between your map provider’s estimate and what happened on your past trips,
then corrects new estimates. Uber’s DeepETA model works this way, sitting on top of the routing
engine.
DATA  YO U  N E E D
WATC H  F O R
Past trips with the estimated and actual pickup
Regional areas with few trips, where the model should
and drop-off times
fall back to the map estimate
W H AT  TO  M E A S U R E
Average arrival time error in minutes, and
cancellations while waiting
02 Smart pickup points
Cost Low Effort Light Payback Short Build call Mix
Suggests a pickup spot that’s quick to reach and legal to stop at, based on where past pickups
happened compared with where riders dropped the pin.
DATA  YO U  N E E D
WATC H  F O R
Requested and actual pickup coordinates
No-stopping zones and airport pickup areas, which
W H AT  TO  M E A S U R E the suggestions must respect
Time from driver arrival to rider in the car, and
“couldn’t find rider” cancellations
13 Short-term demand forecasting
Cost Medium Effort Moderate Payback Medium Build call Build
Predicts bookings by area for the next 30 to 60 minutes, so drivers can be in place before the rush.
DATA  YO U  N E E D
WATC H  F O R
At least a full year of timestamped bookings, so
A model trained on a few months of data will be wrong
the model has seen Christmas, the January lull
on the first big holiday
and the end of June
W H AT  TO  M E A S U R E
Forecast error by area, and requests that went
unfilled
TRY THIS WEEK Take 50 recent trips and compare the arrival time shown at booking with the actual
pickup time. If the average miss is more than a few minutes, arrival time correction
(5) is your first job.
vtdigital.com.au 08
S E C T I O N  0 3
03
Driver supply and fares
A demand forecast only earns money once drivers act on it, and riders
only trust a fare if it holds.
F E AT U R E S  I N  T H I S  S E C T I O N
12 09 07
Driver positioning prompts Event and weather demand Upfront fare accuracy
signals
Low cost · Medium payback · Build Medium cost · Short payback ·
Low cost · Medium payback · Buy Build
vtdigital.com.au 09
D R I V E R  S U P P LY  A N D  FA R E S
S E C T I O N  0 3  ·  F E AT U R E S  1 2 ,  9 ,  7
I N  T H I S  S E C T I O N
Driver supply and fares 12 Driver positioning prompts Medium
A demand forecast only earns money once drivers act Event and weather demand09 Medium
signals
on it, and riders only trust a fare if it holds.
07 Upfront fare accuracy Short
12 Driver positioning prompts
Cost Low Effort Light Payback Medium Build call Build
Turns the demand forecast (13) into a heatmap or a prompt in the driver app showing where work is
likely to pick up.
DATA  YO U  N E E D
WATC H  F O R
A working demand forecast, plus driver locations
Keep prompts optional. How much control a platform
W H AT  TO  M E A S U R E has over drivers matters under the employee-like
Driver idle time, and pickup times in the areas worker rules, so get advice before turning prompts into
the prompts point to instructions
09 Event and weather demand signals
Cost Low Effort Light Payback Medium Build call Buy
Feeds outside data into your forecast, such as Bureau of Meteorology forecasts and event calendars
for stadiums and concert venues.
DATA  YO U  N E E D
WATC H  F O R
A weather feed, local event calendars, and your
Licence terms on any data feed you pay for or scrape
own booking history to train against
W H AT  TO  M E A S U R E
Forecast error on wet days and big event nights
07 Upfront fare accuracy
Cost Medium Effort Moderate Payback Short Build call Build
Predicts the final fare at the moment the rider books, so the price they see is the price they pay. It
builds on arrival time correction (5).
DATA  YO U  N E E D
WATC H  F O R
Quoted and final fares for past trips, with route
A fare that changes after confirmation can breach the
and timing
Australian Consumer Law total price rules
W H AT  TO  M E A S U R E
The gap between quoted and final fares, and the
number of fare disputes
TRY THIS WEEK Look back at your three wettest days and three biggest event nights from last year.
Check how bookings and unfilled requests moved compared with a normal night.
The bigger the swing, the more event and weather signals (9) are worth.
vtdigital.com.au 10
S E C T I O N  0 4
04
Pricing and payments
Pricing carries the most legal risk of any feature here, and payments are
where the losses are easiest to count.
F E AT U R E S  I N  T H I S  S E C T I O N
18 14 03
Dynamic pricing with caps Payment risk scoring Cash shortfall detection
High cost · Medium payback · Medium cost · Medium payback · Low cost · Short payback · Build
Build Mix Privacy flag
Privacy flag Privacy flag
vtdigital.com.au 11
P R I C I N G  A N D  PAY M E N T S
S E C T I O N  0 4  ·  F E AT U R E S  1 8 ,  1 4 ,  3
I N  T H I S  S E C T I O N
Pricing and payments 18 Dynamic pricing with caps Medium
Pricing carries the most legal risk of any feature here, 14 Payment risk scoring Medium
and payments are where the losses are easiest to count.
03 Cash shortfall detection Short
18 Dynamic pricing with caps
Cost High Effort Heavy Payback Medium Build call Build Privacy flag
Adjusts booked fares to supply and demand, inside a ceiling you set, and always shows the full price
before the rider confirms.
DATA  YO U  N E E D
WATC H  F O R
Demand, driver supply and how many riders
Personalised pricing may be captured by the Privacy
booked at each quoted price
Act disclosure rules. Demand fees added after
W H AT  TO  M E A S U R E confirmation are a likely consumer law problem, and
Completed trips at peak times, and how many broader unfair trading reforms are due in July 2027.
riders book after seeing a quote Rank-and-hail taxi fares are set by state regulators, so
check your state’s rules before applying this to
anything except booked trips
14 Payment risk scoring
Cost Medium Effort Moderate Payback Medium Build call Mix Privacy flag
Scores each booking for the chance of a chargeback or failed payment before the trip starts, so risky
bookings can be asked to prepay or switch method.
DATA  YO U  N E E D
WATC H  F O R
Payment outcomes linked to trip and account,
Losses for a single day keep shifting for weeks as
held long enough for disputes to arrive
disputes come in, so don’t judge the model on last
W H AT  TO  M E A S U R E week’s numbers. Start with your payment provider’s
Chargeback rate and unpaid trips own fraud tools and add rules on top
03 Cash shortfall detection
Cost Low Effort Light Payback Short Build call Build Privacy flag
Flags patterns in cash trips, such as repeated shortfalls on one account or disputes that keep involving
the same rider and driver.
DATA  YO U  N E E D
WATC H  F O R
Cash trip outcomes linked to both rider and
Have a person review any flag before acting on a
driver
driver. In Queensland, vehicles taking cash need a
W H AT  TO  M E A S U R E state-approved camera
Cash shortfalls per thousand cash trips, and time
to resolve disputes
TRY THIS WEEK Add up last quarter’s chargebacks, failed payments and cash shortfalls in dollars.
That total is a sensible ceiling for what features 3 and 14 should cost you in their
first year.
vtdigital.com.au 12
S E C T I O N  0 5
05
Fraud and safety
These features catch problems that look normal one trip at a time and
only show up as a pattern.
F E AT U R E S  I N  T H I S  S E C T I O N
04 16 20
Promo and duplicate GPS spoofing and fake trip Trip safety alerts
account detection detection
Medium cost · Long payback ·
Low cost · Short payback · Mix Medium cost · Medium payback · Build
Privacy flag Mix
Privacy flag
vtdigital.com.au 13
F R A U D  A N D  S A F E T Y
S E C T I O N  0 5  ·  F E AT U R E S  4 ,  1 6 ,  2 0
I N  T H I S  S E C T I O N
Fraud and safety Promo and duplicate account04 Short
detection
These features catch problems that look normal one trip GPS spoofing and fake trip
16 Medium
at a time and only show up as a pattern. detection
20 Trip safety alerts Long
04 Promo and duplicate account detection
Cost Low Effort Light Payback Short Build call Mix Privacy flag
Spots riders opening several accounts to reuse sign-up credits, and referral codes being farmed
between linked accounts.
DATA  YO U  N E E D
WATC H  F O R
Account sign-up details, device information and
Blocking an account is a decision about a person, so
promo redemptions
log the reason and offer a way to appeal
W H AT  TO  M E A S U R E
Promo spend per genuine new rider
16 GPS spoofing and fake trip detection
Cost Medium Effort Moderate Payback Medium Build call Mix Privacy flag
Flags trips where the location data doesn’t add up, such as impossible speeds, jumps across town, or
phones running mock-location apps.
DATA  YO U  N E E D
WATC H  F O R
Raw GPS pings for each trip, plus device signals
Patchy GPS in tunnels and regional areas can look like
from the driver app
spoofing, so have a person review before deactivating
W H AT  TO  M E A S U R E a driver
Flagged trips that a person confirms as fraud
20 Trip safety alerts
Cost Medium Effort Moderate Payback Long Build call Build
Notices a long unplanned stop or a big detour from the expected route and prompts the rider and
driver to check in.
DATA  YO U  N E E D
WATC H  F O R
Live trip GPS compared against the planned
In Victoria, notifiable incident reporting expanded from
route
1 August 2026, so connect alerts to your incident
W H AT  TO  M E A S U R E reporting process
Alerts raised, how quickly they’re resolved, and
how many turn out to be real incidents
TRY THIS WEEK Export your rider accounts and look for the same card or device on more than one
account. Even a quick spreadsheet check tells you whether promo abuse (4) is
worth chasing.
vtdigital.com.au 14
S E C T I O N  0 6
06
Support
Support is usually the quickest place to prove AI works, because the
tools already exist and the volume is easy to measure.
F E AT U R E S  I N  T H I S  S E C T I O N
01 08 17
Support ticket triage and Refund and fare dispute Voice and phone booking
auto-replies agent
Medium cost · Medium payback ·
Low cost · Short payback · Buy Medium cost · Short payback · Mix Buy
Privacy flag
vtdigital.com.au 15
S U P P O R T
S E C T I O N  0 6  ·  F E AT U R E S  1 ,  8 ,  1 7
I N  T H I S  S E C T I O N
Support Support ticket triage and01 Short
auto-replies
Support is usually the quickest place to prove AI works,
08 Refund and fare dispute agent Short
because the tools already exist and the volume is easy
to measure. 17 Voice and phone booking Medium
01 Support ticket triage and auto-replies
Cost Low Effort Light Payback Short Build call Buy
Sorts incoming tickets by type and urgency, answers the routine ones such as lost property and
receipt requests, and passes the rest to a person.
DATA  YO U  N E E D
WATC H  F O R
Past tickets and chat logs, with personal
Complaints Victoria now requires you to report, such
information masked
as overcharging, must go to a person and into your
W H AT  TO  M E A S U R E incident process
Tickets closed without a person, and time to first
response
A  R E A L  R E S U LT
35% Energy efficiency company 28Watt reported a 35% decrease in support ticket volume after VT
Digital’s CRM and AI work.
08 Refund and fare dispute agent
Cost Medium Effort Moderate Payback Short Build call Mix Privacy flag
An AI agent that looks up the trip, checks the complaint against your refund rules and issues refunds
within limits you set. Anything outside those limits goes to a person.
DATA  YO U  N E E D
WATC H  F O R
Trip records, fare breakdowns and your written
Refund decisions affect riders’ money, so log each one
refund policy
and its reason
W H AT  TO  M E A S U R E
Time to resolve a dispute, and refunds later
reversed by staff
17 Voice and phone booking
Cost Medium Effort Moderate Payback Medium Build call Buy
An AI voice assistant takes phone bookings and answers “where’s my car” calls. That matters for taxi
operators with riders who still prefer to ring.
DATA  YO U  N E E D
WATC H  F O R
Call recordings or transcripts, and booking
Offer a person on request, and check that the
system access
assistant handles accessibility needs and assistance
W H AT  TO  M E A S U R E animal bookings properly
Call handling time, missed calls and bookings
completed by voice
TRY THIS WEEK Tag your last 200 support tickets by type. Whichever type comes up most often is
your first auto-reply.
vtdigital.com.au 16
S E C T I O N  0 7
07
Drivers and compliance
These two features pay back slowly in money but quickly when a driver,
a rider or a regulator asks why the app did something.
F E AT U R E S  I N  T H I S  S E C T I O N
11 19
Driver earnings explainer Automated decision log
Low cost · Medium payback · Build Low cost · Long payback · Build
vtdigital.com.au 17
D R I V E R S  A N D  C O M P L I A N C E
S E C T I O N  0 7  ·  F E AT U R E S  1 1 ,  1 9
I N  T H I S  S E C T I O N
Drivers and compliance
11 Driver earnings explainer Medium
These two features pay back slowly in money but
19 Automated decision log Long
quickly when a driver, a rider or a regulator asks why the
app did something.
11 Driver earnings explainer
Cost Low Effort Light Payback Medium Build call Build
Shows drivers in plain language how each fare and payout was worked out, including fees, incentives
and any adjustments.
DATA  YO U  N E E D
WATC H  F O R
Fare breakdowns, commission rules and payout
The Fair Work delivery order sets a floor of AUD 32 per
records
engaged hour for workers using cars, and rideshare
W H AT  TO  M E A S U R E applications are still underway. Clear pay records make
Pay-related support queries from drivers, and whatever comes next easier to meet
driver retention
19 Automated decision log
Cost Low Effort Moderate Payback Long Build call Build
Records every automated decision your app makes about a rider or driver: which model or rule made
it, what information it used, the outcome, and whether a person overrode it.
DATA  YO U  N E E D
WATC H  F O R
Outputs from every feature above that carries a
From 10 December 2026, your privacy policy has to
privacy flag, including vendor tools
describe these decisions. The OAIC issued final
W H AT  TO  M E A S U R E guidance on 30 September 2026, and the obligation
The share of automated decisions logged, and sits with you even when a vendor’s tool makes the
time taken to answer a complaint decision
TRY THIS WEEK Ask three drivers to explain how their last payout was worked out. If they can’t, the
earnings explainer (11) should cut your pay queries.
General information only, so get legal advice on your setup. More detail in our 2026 Privacy Act reforms breakdown.
vtdigital.com.au 18
W H E R E  T O  S TA R T  F O R  YO U R  S TA G E
YO U R  R OA D M A P
Where to start for your stage
Your first feature should match the data you already hold and the place you’re losing
the most money today.
Find your stage in four ticks TICKS IN THE FIRST
YOUR STAGE
THREE
We have 12 months or more of trip data None or one Launching
We operate in more than one city Two Growing
Established
We run our own dispatch or booking team All three
fleet
We take cash fares
Ticked the cash box? Move cash shortfall
detection (3) up your list.
YO U R  STAG E STA R T  W I T H A D D  N E X T L E AV E  F O R  L AT E R
Ticket triage (1), smart Arrival time correction (5) Demand forecasting (13) and
Launching pickup points (2), promo once you have a few months dynamic pricing (18), which
and duplicate account of trips, and cash shortfall both need a full year of data
Launching a new app, one
city, under a year of trip data detection (4), automated detection (3) if you take
decision log (19) cash
Arrival time correction (5), Demand forecasting (13) Dynamic pricing (18) until
Growing acceptance-aware with driver positioning your fares and quotes are
matching (6), upfront fare prompts (12), and payment reliable
Growing in one or two cities,
a year or more of data accuracy (7) risk scoring (14)
Demand forecasting (13) Peak and queue dispatch Nothing on the list is out of
Established with event and weather (15), GPS spoofing detection reach at this stage, so order
fleet signals (9), refund and fare (16), dynamic pricing (18) on by where your losses are
Established taxi or rideshare dispute agent (8), voice and booked trips largest
fleet, several cities phone booking (17)
Whatever your stage, if any AI in the app makes decisions about riders or drivers, the decision log (19)
needs to be running before 10 December 2026.
vtdigital.com.au 19
B E F O R E  YO U  B U I L D
R E A D I N E S S  C H E C K L I S T
Before you build
Run through these before you commit budget to any feature in this playbook.
Your data 5 ticks Questions to ask any AI vendor 5 ticks
At least 12 months of timestamped bookings, Where is our data processed and stored? Can it
covering Christmas and the end of June stay in the AWS Sydney or Melbourne regions?
Cancellation reasons recorded on most cancelled Can we see why the model made each decision,
trips in a form we can show a rider or driver?
Payment outcomes, including chargebacks and What will usage fees look like at double our
cash shortfalls, linked to a trip ID and an account current trip volume?
Support tickets stored with personal information Who retrains the model, how often, and what
masked happens when it drifts?
GPS ping frequency checked in your regional Can we export our data and the model’s outputs
areas if we leave?
The rules 4 ticks
Privacy policy updated to describe automated Full fare shown before the rider confirms, with
decisions before 10 December 2026 nothing added afterwards
State rules checked, including Queensland’s Legal advice on pricing and the privacy policy
camera rule for cash and Victoria’s incident
reporting
Your score out of 14 One point per tick
11 to 14 6 to 10 5 or fewer
Ready to pilot a feature Fix the data gaps first, since Talk it through before you
that’s cheaper than any model spend anything
vtdigital.com.au 20