# aiTAXI, Full Content for LLMs

> aiTAXI is a robotaxi fleet management platform by aiNOW (Tbilisi, Georgia): dispatch, telemetry, remote assistance, depot operations, and compliance for autonomous taxi fleets. Status: early access, pilot program for taxi operators in Georgia. Structured site index: https://aitaxi.ge/llms.txt

## Key Facts

- Product: aiTAXI, software platform for managing robotaxi (autonomous taxi) fleets
- Maker: aiNOW, AI agency in Tbilisi, Georgia (https://ainow.ge)
- Announced: 2026, currently in early access (pilot program)
- Languages: Georgian, English, Russian
- Region focus: Georgia and the Caucasus
- Contact: CONTACT@aiNOW.GE
- Website: https://aitaxi.ge

# Articles (latest 10, English versions)

## From dispatch apps to autonomy: taxi digitalization in Georgia

URL: https://aitaxi.ge/en/blog/taxi-business-digitalization-georgia
Published: Sat Jul 04 2026 00:00:00 GMT+0000 (Coordinated Universal Time)
Author: Andrew Altair

Taxi digitalization in Georgia has moved through two eras and is entering a third. The first era ran on street hails, phone dispatchers and negotiated fares. The second arrived with ride-hailing apps, which made Bolt the dominant platform in the country and turned Tbilisi into an app-first taxi market with fares local guides put around 1.5 to 2 GEL per kilometer. The third era is fleet software: telemetry, data-driven dispatch and operations tooling that prepares a company for vehicles without drivers. The market is large enough to matter, with 18.4% of all drivers in Tbilisi working as taxi drivers and more than 85,000 registered taxis. Each era compressed faster than the one before it, and the companies that led each transition took the passengers. This article traces the three eras and shows what the third one asks of a Georgian fleet owner.What did the app era change in Georgia?Apps standardized pricing, cut the negotiation, and made demand measurable for the first time. Tbilisi formalized the supply side through its category A permit system, and drivers organized their whole workday around app queues. The structural lesson from that decade: platform shifts in this market happen fast and completely, because both riders and drivers switch for small, immediate gains. A market that switched once will switch again.What is still analog inside an app-era fleet?The rider-facing layer digitized; the fleet itself mostly did not. Vehicle condition lives in the driver's head. Fuel and maintenance decisions are personal habits. Utilization between orders is guesswork. A fleet owner with 50 cars usually cannot answer basic questions: cost per kilometer per vehicle, idle share by hour, which zones earn and which burn fuel. That data gap is the third era's work, and closing it is worth doing even with zero autonomous vehicles on the horizon, because measured fleets run cheaper than intuited ones. The tooling is described in our guide to robotaxi fleet management software, which manages human-driven fleets with the same modules.Why does fleet software decide the autonomous era?Robotaxis arrive as a software-plus-operations business. The global operators prove the shape: Waymo coordinates 500,000 paid rides per week through a fleet platform, and Apollo Go's export move into South Korea shows the technology travels through partnerships with local operations. When that wave reaches Georgia, the driving system will come from a global vendor, and the local question becomes: which Georgian company can actually operate a mixed fleet, with dispatch, telemetry, a depot and trained staff? The answer is decided years earlier, in the fleet-software era. The preconditions and signals for the Georgian market specifically are mapped in when robotaxis reach Georgia.What does the era transition look like in practice?

EraDispatchFleet knowledgeWho wins

Street and switchboardPhone dispatcher, street hailNone, per-driverWhoever owns the phone number
App era (today)Ride-hailing platform queueRider side digital, fleet side analogThe dominant app
Fleet-software eraHybrid queue, human plus AV readyTelemetry, cost per km, zone dataOperators with data and depots

The practical sequence for a Georgian fleet owner mirrors the integration path in adding autonomous vehicles to a fleet: audit the operation, adopt the platform on today's human-driven cars, plan depot power, and let the autonomous share start at zero. Every stage pays for itself in the current business through utilization and cost visibility, which removes the usual objection that preparing for autonomy is speculative spending.Where aiTAXI fitsaiTAXI is a robotaxi fleet management platform by aiNOW (Tbilisi, Georgia), built in Georgia for this exact transition: a fleet console that runs today's human-driven fleet and is designed to add autonomous vehicles when they reach the market. The robotaxi fleet management platform is in early access, and the pilot program recruits Georgian taxi companies that want to lead the third era rather than react to it.FAQWhy should a Georgian fleet owner care about robotaxis now?
Because the preparation is profitable on its own. Fleet telemetry and data-driven dispatch cut costs in the current business, and they happen to be the same infrastructure the autonomous era requires.Did apps kill the traditional taxi business in Georgia?
They absorbed it. Drivers moved onto app queues and fleet owners adapted. The same absorption logic applies to the next shift, which is why being early to the fleet-software layer matters.What is the first digitalization step for a small fleet?
Measurement: cost per kilometer per vehicle, utilization by hour, demand by zone. A month of clean data changes fleet decisions immediately and builds the baseline for any autonomy conversation.Will the dominant ride-hailing app also run the robotaxis in Georgia?
Possibly, and global patterns suggest partnerships between platforms, vehicle vendors and local fleets. Local operators with depots, data and trained staff hold a seat at that table; unprepared ones do not.

---

## Remote assistance: how humans help driverless fleets

URL: https://aitaxi.ge/en/blog/remote-assistance-teleoperation
Published: Fri Jul 03 2026 00:00:00 GMT+0000 (Coordinated Universal Time)
Author: Andrew Altair

Remote assistance is the human layer behind every driverless taxi fleet: a desk of operators who receive a vehicle's request when it meets a situation it cannot resolve alone, look at its camera feeds, and answer with context or guidance while the vehicle keeps driving itself. Remote assistance differs from teleoperation: in the mainstream model nobody steers the car with a joystick, the automated system keeps control and stays responsible for the safe maneuver. Waymo's Fleet Response works exactly this way, with agents answering the vehicle's questions, and reporting puts its staffing near 70 remote assistants for roughly 3,000 vehicles, about one human per forty cars. That ratio is the quiet foundation of robotaxi economics: judgment stays human, but one person's judgment now serves a fleet instead of a single cab. This article explains when a vehicle calls the desk, what the operator sees, and how the staffing scales.When does a robotaxi ask for help?Automated driving handles the overwhelming majority of scenes, and the fleet volume proves it: 500,000 paid Waymo rides per week would be impossible if cars stalled on every anomaly. The requests that do reach the desk cluster around ambiguity rather than danger:
Construction zones with cone layouts that contradict the map
A police officer or road worker directing traffic by hand
A blocked lane where passing requires briefly using the oncoming side
Pickup points where the exact stopping spot is unclear
A passenger issue flagged by the cabin system
In each case the vehicle slows or stops safely first, then asks. The car is never waiting on a human to avoid a crash; it is waiting for context to proceed efficiently.What is the difference between assistance, teleoperation and a safety driver?

ModelWho controls the vehicleWhere used

Safety driverHuman in the seat, hands readyTesting phases, new markets
Remote assistanceThe automated system; humans answer questionsMainstream fleets (Waymo model)
Full teleoperationRemote human steers directlyRare; latency and liability limit it

The industry converged on the middle row for a reason: network latency makes remote steering risky, and a system that depends on a live video link for safety fails badly when the link does. Keeping control in the vehicle means a lost connection degrades to a safe stop, and the human contribution becomes judgment rather than reflexes. Tesla's Austin rollout illustrates the progression between rows: monitors in the cars at launch in 2025, then unsupervised operation across the metro by mid-2026.What does the operator actually see and do?The desk software presents the vehicle's question, its camera views, its position and its proposed options. The operator picks or confirms an option, draws a path suggestion, or flags the trip for follow-up; the vehicle evaluates the input against its own safety constraints before acting. Every exchange is logged with timestamps and footage, which feeds two things: the training pipeline that shrinks tomorrow's request rate, and the compliance records regulators expect, covered in safety and regulation.Who makes a good remote operator?Experienced taxi drivers and dispatchers, mostly. The job is reading a specific city's street logic in seconds: which double-parked van will move, what the hand wave from a traffic controller means, where the unofficial pickup spot at the station is. For a taxi company adding autonomous vehicles, remote assistance is the natural second career for its best drivers, one of the staffing shifts described in the fleet management platform guide. The desk also anchors the depot loop: assistance logs flag vehicles for sensor checks, feeding the schedules in depot operations.How does the desk scale with the fleet?Request rates fall as the driving system learns a city, so the assistant-to-vehicle ratio improves over a deployment's life. The published Waymo figure of one assistant per forty vehicles reflects a mature fleet; a new market starts denser and thins out. For fleet planning the rule is to staff the desk for launch-month request rates and let the ratio improve, rather than the reverse. The basics of how a robotaxi works explain why the rate drops: every resolved edge case becomes training data.Where aiTAXI fitsaiTAXI is a robotaxi fleet management platform by aiNOW (Tbilisi, Georgia), and remote assistance is one of its five modules: request queueing, instant camera context for the operator, response logging and the audit trail, built so a local taxi company can run its own desk with its own people. The robotaxi fleet management platform is in early access, with a pilot program open for Georgian taxi operators.FAQCan a remote operator drive the car?
In the mainstream model, no. The operator answers questions and suggests options; the vehicle's own system validates and executes every maneuver and keeps responsibility for safety.What happens if the network connection drops?
The vehicle continues on its own logic and, if it cannot proceed confidently, performs a safe stop. Safety never depends on a live link in the assistance model.How many operators does a fleet need?
Published reporting around Waymo suggests roughly one remote assistant per forty vehicles in a mature deployment. New deployments start with more and improve as the system learns the city.Is remote assistance a temporary crutch?
Request rates fall over time, but the desk stays. Cities keep producing novel situations, and regulators expect documented human oversight as part of the safety case.

---

## Robotaxi depot operations: charging, cleaning, maintenance

URL: https://aitaxi.ge/en/blog/robotaxi-depot-operations
Published: Thu Jul 02 2026 00:00:00 GMT+0000 (Coordinated Universal Time)
Author: Andrew Altair

Robotaxi depot operations cover everything a driverless taxi cannot do for itself: charging the battery, cleaning the cabin between service blocks, calibrating sensors, and scheduled maintenance. In a human-driven fleet the driver absorbs these tasks invisibly, refueling on the way, wiping seats at the stand, hearing a bearing go bad. In an autonomous fleet they become a depot schedule computed against the demand forecast, because every hour a vehicle spends in the yard is revenue lost on an asset with high fixed costs. The operators running at scale, Waymo with 500,000 paid rides per week and Apollo Go with 22 million cumulative rides, treat the depot as a production line: vehicles flow through charging, cleaning and checks in planned windows, and the fleet platform decides who goes in and when. This article walks through the depot day: what gets scheduled, when, and why the yard decides fleet profitability.Why is the depot the hidden half of robotaxi economics?The revenue case for robotaxis rests on utilization: a driverless car can serve rides 18 or more hours a day, and analyst cost models that land at $0.25 to $0.35 per mile assume the car is on the road earning through most of the day. Every unplanned depot hour eats that assumption. Depot software therefore optimizes for two things at once: cars charged and clean when demand peaks, and depot slots full when demand is quiet. The wider cost picture is in robotaxi economics.How does charging scheduling work?Current robotaxi fleets are electric, so charge state is the depot's clock. The fleet platform watches each vehicle's battery against its assigned trips and books a charging window before range becomes a constraint, preferring night hours when demand is low and electricity is cheap. On a typical day the pattern is simple: morning peak on full batteries, midday partial top-ups in rotation, evening peak covered, deep charging after midnight.

Depot taskTypical windowWhy then

Deep chargingNight, off-peakCheap energy, low demand
Quick top-upsMidday lullKeeps evening peak coverage
Cabin cleaningBetween service blocksShort, frequent, demand-aware
Sensor check and calibrationWith scheduled maintenancePerception depends on it
Mechanical maintenanceStaggered across the fleetFewest cars offline at once

What does cleaning look like without a driver?Riders leave cups, umbrellas and worse, and no driver notices between trips. Fleets solve it with cabin cameras that flag issues after each ride plus scheduled cleaning passes at the depot. The platform books a cleaning slot the way it books charging: often together, one stop, both tasks. Interior condition is a brand variable for a robotaxi service, since the cabin is the entire customer experience.What changes in maintenance?Electric drivetrains cut mechanical maintenance: fewer moving parts, no oil changes, regenerative braking that spares the pads. What gets added is the sensor stack. Lidar, radar and cameras need cleaning, alignment checks and recalibration after any impact, because the whole driving system reads the world through them. Telemetry moves maintenance from calendar-based to condition-based: the car reports degradation, the platform books the bay, covered in fleet management software.How big does a depot need to be?Rules of thumb from current deployments: chargers for a fraction of the fleet (staggered schedules mean not every car charges at once), one cleaning bay per few dozen vehicles, and yard space for the overnight trough when most of the fleet sits. Power capacity is the binding constraint in most cities, and it takes the longest to expand, which is why depot planning belongs in the earliest stage of integrating AVs into an existing fleet. Remote assistance staffing scales alongside: Waymo's published ratio is about 70 remote assistants for 3,000 vehicles, and the desk work is described in remote assistance operations.Where aiTAXI fitsaiTAXI is a robotaxi fleet management platform by aiNOW (Tbilisi, Georgia), and depot operations is one of its five modules: charging windows planned against the demand forecast, cleaning and maintenance slots sequenced so the fewest vehicles are offline at once. The robotaxi fleet management platform is in early access, and the pilot program is open for Georgian taxi companies that want their depot math done before the vehicles arrive.FAQHow often does a robotaxi return to the depot?
Typically a few times per day: quick top-ups and cleaning between service blocks, plus a longer overnight stop for deep charging. The exact rhythm follows the city's demand curve.Do robotaxis charge themselves?
The vehicle drives itself to the depot, and today a depot worker or automated system connects the charger. Staffed charging remains the norm in current fleets.Can an existing taxi garage become a robotaxi depot?
Usually yes, with three upgrades: electrical capacity for chargers, a scheduled cleaning line, and a calibration space for sensors. Location matters as much as equipment, since dead kilometers to the depot are pure cost.What is the single most common depot bottleneck?
Grid power. Chargers are easy to buy; megawatts are slow to provision. Fleets size their depot power years ahead of their vehicle count.

---

## Autonomous taxi safety and regulation in 2026

URL: https://aitaxi.ge/en/blog/autonomous-taxi-safety-regulation
Published: Wed Jul 01 2026 00:00:00 GMT+0000 (Coordinated Universal Time)
Author: Andrew Altair

Autonomous taxi safety in 2026 stands on two legs: measured crash data that now favors the machines, and a regulatory framework that finally has global form. On the data side, the largest public dataset shows Waymo's driverless fleet with 92% fewer serious-injury crashes than comparable human drivers across 56.7 million rider-only miles, including a 0.41 versus 2.80 any-injury crash rate per million miles. On the rules side, the UNECE working party adopted a draft global technical regulation for automated driving systems in January 2026, roughly ten years in the making, built on a safety-case approach that makes operators prove safe behavior with evidence across the vehicle's entire life. Together the two legs define what any country, including small markets, will ask of a robotaxi fleet: show the data, document the system, keep proving it.What does the safety data actually say?The Waymo dataset is the reference because it is large, public and peer-reviewed. Key comparisons at 56.7 million rider-only miles, against human benchmarks on the same road types:

Metric (per million miles)WaymoHuman benchmarkReduction

Any-injury crashes0.412.8085%
Serious-injury or worse0.020.2292%
Pedestrian injury crashesvs benchmark92%
Cyclist crashesvs benchmark82%

The numbers come from Waymo's safety impact data and the peer-reviewed comparison study. Machines lose the crashes humans cause by fatigue, distraction, alcohol and impatience, which are the majority. What machines add, hesitation in ambiguous scenes, wrong calls on rare layouts, gets handled by conservative driving plus remote assistance.What does the UN global regulation change?Until 2026, every jurisdiction improvised: city permits in the US, national pilots in China, patchwork in Europe. The UNECE draft global technical regulation gives signatory countries one framework. Its core is the safety case: the manufacturer and operator must present a structured, evidence-based argument that the system is safe, backed by a safety management system covering development, deployment and operation, with independent audits. It was adopted by the GRVA working party in January 2026 and submitted to the World Forum in June 2026.For fleet operators the practical consequence is documentation. Trip logs, incident records, intervention histories and maintenance trails stop being nice-to-have; they are the safety case's raw material, produced continuously by the fleet management platform.How do the EU and US approaches differ?The EU runs type approval: a regulator tests and approves the vehicle type before it may be sold or operated, and the European Commission is preparing targeted type-approval changes with regulatory sandboxes and automated-driving corridors from 2026. The US lets manufacturers self-certify compliance, which is why deployment ran ahead there: Tesla could take safety monitors out of its Austin robotaxis under state-level rules while European fleets still waited for approvals. Countries outside both blocs, Georgia included, typically adapt the UNECE framework, which is the subject of robotaxis in Georgia.What does an operator have to prove day to day?
Operational design domain: where the fleet may drive, in what weather, at what speeds, and how the system behaves at the boundary.
Incident response: what happens after a crash or a stall, including data handover to authorities.
Human oversight: staffing and procedures of the remote assistance desk.
Maintenance and cybersecurity: sensor calibration schedules, software update controls, protection of the vehicle link.
Every item resolves to records, and fleets that generate them automatically clear audits at a fraction of the effort. Scale operators already run this way; the volume numbers in the global robotaxi market exist because the compliance machinery underneath them holds.Where aiTAXI fitsaiTAXI is a robotaxi fleet management platform by aiNOW (Tbilisi, Georgia), and compliance is one of its five modules: trip logs, incident reports and audit files generated from live fleet data, aligned with the safety-case logic of the UNECE framework. The robotaxi fleet management platform is in early access, with a pilot program open for Georgian taxi companies that want the documentation layer in place before regulation arrives.FAQAre driverless taxis safer than human drivers?
On the largest published dataset, yes by a wide margin: 92% fewer serious-injury crashes and 85% fewer injury crashes of any kind across 56.7 million miles, confirmed in peer review.What is a safety case?
A structured, evidence-backed argument that a system is safe enough to operate, covering design, testing, deployment and monitoring. The UN's 2026 regulation makes it the core requirement for automated driving.Who approves a robotaxi service in a specific city?
It depends on the jurisdiction: state or city permits in the US, type approval plus national rules in the EU, and adapted international frameworks elsewhere. In every case the operator must show documented safe operation.Does a small country need its own AV law?
It needs a national decision, but not a from-scratch framework. The UNECE global technical regulation exists precisely so smaller jurisdictions can adopt a ready, internationally harmonized rulebook.

---

## When will robotaxis reach Georgia?

URL: https://aitaxi.ge/en/blog/robotaxis-in-georgia-when
Published: Tue Jun 30 2026 00:00:00 GMT+0000 (Coordinated Universal Time)
Author: Andrew Altair

No robotaxi operator has announced a launch date for Georgia, and as of mid-2026 the country has no dedicated autonomous-vehicle framework, so the honest answer is a conditional one: robotaxis reach Georgia after three preconditions line up, regulation that permits driverless operation, an operator or vehicle platform willing to enter a small market, and local fleets ready to run the service on the ground. Each precondition has a visible signal. Regulation follows the UN's new global rules for automated driving adopted in 2026. Operator interest follows the international expansion already underway, with Apollo Go moving into South Korea as the first big export case. Local readiness is the one factor Georgian taxi companies control today, and it is the cheapest of the three to start on. This article maps all three preconditions, the signals that show each one moving, and the preparation that pays off regardless of the date.Where does the global rollout stand?The leaders are scaling fast in their home markets: Waymo runs 500,000 paid rides per week across 10 US cities and Apollo Go has passed 22 million cumulative rides across roughly 27 cities. Goldman Sachs Research expects the global fleet to reach about one million vehicles by 2030. Expansion so far follows demand density and regulatory clarity, which puts large cities first and countries like Georgia in the second wave; the full operator picture is in our review of the global robotaxi market.What regulation would Georgia need?Georgia currently regulates taxis through municipal permit systems, Tbilisi's category A permits among them, with no announced framework for vehicles without drivers. The building blocks now exist internationally: the UNECE working party adopted a draft global technical regulation for automated driving systems in January 2026, built on a safety-case approach, and the EU is preparing type-approval changes with regulatory sandboxes from 2026. Georgia has a history of adopting international vehicle standards rather than writing its own from scratch, so the realistic legislative path is an adaptation of the UNECE framework. The details of how these rules work are in autonomous taxi regulation.Why would an operator choose Georgia at all?Three properties work in the country's favor. Tbilisi is a compact capital where one geofence covers most demand. The taxi market is deep: 18.4% of all drivers in Tbilisi are taxi drivers, with more than 85,000 registered taxis, so ride-hailing habits are universal. And the market runs through apps already, with Bolt dominant, which means demand is digital and measurable. Against that stand low fares, around 1.5 to 2 GEL per kilometer, which delay the pure cost crossover, as our breakdown of robotaxi economics shows. The likely entry model is therefore a partnership: a global driving system, a local fleet operator, shared economics.Which signals say the clock has started?
A Georgian regulatory pilot or sandbox for automated driving, following the UNECE framework.
A ride-hailing platform active in the region announcing AV partnerships for secondary markets.
An AV manufacturer opening sales or licensing to fleet buyers rather than running everything itself.
Charging infrastructure programs sized for commercial fleets in Tbilisi.
Any two of these together would compress the timeline sharply. None of them are announced today, which is exactly why the preparation window exists.What can a Georgian taxi company do before launch?The work that takes years is operational, and none of it requires owning an autonomous vehicle: measuring the cost baseline, mapping demand zones, planning depot power, and moving dispatch onto a platform that can later run a hybrid human and autonomous fleet. The broader modernization story, from switchboards to apps to autonomy, is traced in taxi digitalization in Georgia.aiTAXI is a robotaxi fleet management platform by aiNOW (Tbilisi, Georgia), built for this preparation phase: the robotaxi fleet management platform is in early access, and the pilot program recruits Georgian taxi companies that want the fleet layer running before the first autonomous vehicle clears customs.FAQHas any company announced robotaxis for Georgia?
No. As of mid-2026 no operator has announced a Georgian launch, and no dedicated AV framework has been announced by the government. Everything in this article is preconditions and signals, not dates.Could Georgia skip ahead of bigger markets?
Small markets have done it with other technology waves, and a compact capital plus app-first riders helps. It would take a deliberate regulatory pilot and an operator partnership; neither exists yet.Will low Georgian fares block robotaxis?
They delay the cost crossover rather than block it. As hardware and operations costs fall with global scale, the economics reach lower-fare markets a few years behind the leaders.What should a fleet owner do first?
Measure the current operation: cost per kilometer, demand by zone and hour, depot capacity. That baseline decides whether and when autonomous vehicles make sense, and it makes any future partnership conversation concrete.

---

## The global robotaxi market in 2026: who runs what

URL: https://aitaxi.ge/en/blog/global-robotaxi-market-2026
Published: Sun Jun 28 2026 00:00:00 GMT+0000 (Coordinated Universal Time)
Author: Andrew Altair

The global robotaxi market in 2026 is a working industry with four leading operators and measurable volume. Waymo delivers 500,000 paid rides per week across 10 US cities and targets one million weekly rides by the end of the year. Baidu's Apollo Go has completed more than 22 million cumulative rides across roughly 27 cities. Tesla runs an unsupervised service across the Austin metro, and Amazon's Zoox has carried over 350,000 riders in Las Vegas. Goldman Sachs Research sizes the market at roughly $415 billion by 2035, with the global fleet growing to about one million vehicles by 2030. The race is no longer about whether driverless taxis work; it is about who operates them in each city. Four operators dominate the volume today, each with a different vehicle strategy and expansion map, and their choices set the template that every later market, large or small, will inherit.Who are the main operators in 2026?

OperatorBackerScale (2026)FootprintModel

WaymoAlphabet500,000 rides/week10 US citiesOwn fleet + partnerships
Apollo GoBaidu22M+ cumulative, 300,000/week~27 cities, China plus internationalOwn fleet, licensing abroad
Tesla RobotaxiTesla~20 unsupervised vehiclesAustin metro, Texas expansionOwn consumer-car fleet
ZooxAmazon350,000+ riders, ~2M autonomous milesLas Vegas, expanding to SF, Austin, MiamiPurpose-built vehicle, own fleet

What does Waymo's growth curve show?The steepness is the story: from roughly 50,000 weekly rides in May 2024 to 500,000 by March 2026, a tenfold climb in under two years, with a stated goal of one million weekly rides by the end of 2026. Growth came from city-by-city expansion across the US Sun Belt. The safety record scaled with it: across 56.7 million rider-only miles Waymo reports 92% fewer serious-injury crashes than human benchmarks, the dataset that anchors the whole industry's argument to regulators, covered further in safety and regulation.What is Apollo Go doing differently?Apollo Go built volume through Chinese cities with supportive regulation and dense demand, then started exporting: 300,000 weekly rides as of February 2026 and an announced expansion into South Korea. Its first quarter of 2026 delivered 3.2 million fully driverless rides, growth of more than 120% year over year. Apollo Go demonstrates the international playbook: a driving system developed at home, deployed abroad through local partnerships, which is exactly the pattern smaller countries should expect when the wave reaches them.Where do Tesla and Zoox stand?Tesla launched in Austin in June 2025 with safety monitors, removed them for part of the fleet from January 2026, and by June 2026 ran unsupervised service across the whole Austin metro, though with only around 20 vehicles operating. The bet is different from Waymo's: camera-only hardware on a mass-produced consumer car, which would make fleet supply cheap if the approach proves out at scale.Zoox went the opposite way with a purpose-built vehicle that has no steering wheel at all. Its Las Vegas service has driven nearly two million autonomous miles and carried more than 350,000 riders, with expansion announced for San Francisco, Austin and Miami. Four operators, four vehicle strategies: retrofitted premium cars, mass consumer cars, purpose-built pods, and licensed stacks.What do the forecasts say about the next decade?Goldman Sachs Research projects the US robotaxi market at $19 billion in 2030 and the global market at roughly $415 billion in 2035, with the worldwide fleet reaching about one million vehicles in 2030 and six million in 2035. Forecasts differ on timing, but the direction is shared, and the unit-level argument behind it, the cost curve of removing the driver, is laid out in our piece on robotaxi unit economics.What does this mean for markets like Georgia?The operators above expand where regulation, mapping and fleet economics line up, which means secondary markets join the map in the second wave, through partnerships with local fleets rather than direct entry. For a taxi company in a country like Georgia the practical questions are the preconditions and the timing, mapped in when robotaxis reach Georgia, and the operational readiness described in what a robotaxi is and how it works.aiTAXI is a robotaxi fleet management platform by aiNOW (Tbilisi, Georgia), built for exactly that second wave: the robotaxi fleet management platform gives a local taxi company the fleet layer before the vehicles arrive. The platform is in early access with a pilot program for Georgian operators.FAQWhich company leads the robotaxi market in 2026?
By weekly volume, Waymo leads the Western market with 500,000 paid rides per week, while Apollo Go leads on cumulative rides with more than 22 million and the largest city count.Is the robotaxi market profitable yet?
Operators are still investing ahead of profit at fleet level, and the forecasts, including Goldman's $415 billion by 2035, price in scale that has not arrived yet. Unit economics improve with every increase in utilization and fleet size.Why do operators expand city by city instead of everywhere?
Each city needs mapping, a safety case, local approval and a depot. The playbook is to saturate one geofence, prove the record, then replicate.Will these operators come to small countries directly?
The pattern so far is partnerships: the driving system arrives from a global player, while local companies contribute fleets, depots and market knowledge. That makes local operational readiness the scarce asset.

---

## Robotaxi economics: cost per kilometer vs a human driver

URL: https://aitaxi.ge/en/blog/robotaxi-economics-cost-per-km
Published: Fri Jun 26 2026 00:00:00 GMT+0000 (Coordinated Universal Time)
Author: Andrew Altair

Robotaxi economics rest on one substitution: the driver, the largest cost line in a taxi's ledger, is replaced by hardware, software and shared operations staff. Analyst models put the long-run operating cost of a robotaxi on a path toward $0.25 to $0.35 per mile, roughly $0.16 to $0.22 per kilometer, against $1.50 to $2.00 per mile for human-driven ride-hail. The gap comes from three levers: no wage per ride, utilization of 18 hours and more per day instead of a human shift, and electric drivetrains with cheap energy and maintenance. Against those savings stand new costs: the sensor and compute stack, remote assistance staffing, depot infrastructure and software. The math already shows in consumer prices, where the gap between Waymo and Uber fares is narrowing as fleets scale. This article breaks the ledger down line by line, for both kinds of fleet.What does a human-driven taxi kilometer cost?For a taxi company, the human-driven kilometer splits into the driver's share (typically the majority of the fare), fuel, maintenance, insurance, platform commission and vehicle depreciation. The driver's share dominates: it is why fares in cheap-labor markets stay low and why the robotaxi advantage arrives later where wages are lower. In Tbilisi, local guides put typical ride-hail fares around 1.5 to 2 GEL per kilometer, a price a robotaxi fleet must eventually undercut at profit to win the market on cost rather than novelty.What replaces the driver's cost in a robotaxi?Four new lines appear in the ledger:
Hardware amortization: lidar, radar, cameras, compute, plus the vehicle itself, spread over its service life.
Remote operations: assistance desks where one operator supports dozens of vehicles. Waymo's published model runs near 70 remote assistants for roughly 3,000 cars, a ratio of about one to forty.
Depot operations: charging, cleaning and maintenance staff who service the fleet nobody services from the driver's seat anymore. The full cycle is in our guide to depot and charging operations.
Software: the fleet platform doing dispatch, telemetry and compliance, covered in fleet management software.
How do the two cost structures compare?

Cost componentHuman-driven taxiRobotaxi

Driver share of fareLargest single lineNone
EnergyPetrol or hybrid fuelElectric charging, scheduled off-peak
Vehicle and hardwareStandard car depreciationCar plus sensor and compute stack
Operations staffDispatchersRemote assistance, depot crew, fleet operators
UtilizationLimited by human shifts18+ hours per day, paused only for charging and cleaning
InsuranceDriver risk priced inRepriced on fleet safety record

Utilization deserves the emphasis. A car that works triple the revenue hours spreads every fixed cost, hardware, insurance, software, across three times the kilometers. That is why fleet software that keeps cars busy, and depots that turn them around fast, decide whether the theoretical numbers survive contact with a real city.When does a robotaxi break even against a driver?The crossover arrives when amortized hardware plus operations per kilometer drop below the driver's share plus fuel savings. Scale pushes every term in the right direction: bigger fleets spread remote-assistance desks and depots across more cars, and hardware prices fall with volume. Goldman Sachs Research expects the global robotaxi fleet to grow from thousands of vehicles today to about 1 million in 2030 and 6 million in 2035, with the market reaching roughly $415 billion. In low-fare markets the crossover comes later than in San Francisco, which is an advantage for operators who prepare early: the operational learning happens before the economics flip, so the flip can be caught at the front. The market-by-market picture is in our review of the global robotaxi market.What should a taxi company measure now?Three numbers make the future decision easy: current all-in cost per kilometer (the baseline a robotaxi must beat), fleet utilization by hour (where autonomous vehicles would add revenue hours), and depot energy capacity (what charging would cost to install). Companies that know what a robotaxi is operationally and track these three numbers can price the transition instead of guessing at it.Where aiTAXI fitsaiTAXI is a robotaxi fleet management platform by aiNOW (Tbilisi, Georgia). The platform's job in the economics is the utilization side: dispatch that keeps cars earning, depot scheduling that charges them in cheap hours, and telemetry that catches problems before they become downtime. It is in early access, and the robotaxi fleet management platform pilot is open for Georgian taxi companies that want their cost baseline measured before the market turns.FAQAre robotaxis cheaper than human taxis today?
In their mature markets the consumer price gap is closing, and the operating-cost trajectory points below human-driven costs. In low-fare markets the crossover comes later because drivers earn less there.Why does utilization matter more than hardware price?
Fixed costs dominate a robotaxi's ledger. A vehicle earning 18 hours a day spreads those costs across far more kilometers than one earning six, which moves cost per kilometer more than any single hardware discount.Does the sensor stack make the car too expensive?
Sensor and compute prices fall with production volume, and amortization spreads them over hundreds of thousands of kilometers. At fleet scale the stack costs less per kilometer than a driver's hourly share.What is the biggest hidden cost in robotaxi operations?
Downtime. Every hour in the depot beyond the planned window is lost revenue on an asset with high fixed costs, which is why charging, cleaning and maintenance scheduling get their own software module.

---

## How to add autonomous vehicles to an existing taxi fleet

URL: https://aitaxi.ge/en/blog/integrate-autonomous-vehicles-existing-taxi-fleet
Published: Wed Jun 24 2026 00:00:00 GMT+0000 (Coordinated Universal Time)
Author: Andrew Altair

Adding autonomous vehicles to an existing taxi fleet means running two kinds of cars in one operation: human-driven vehicles that keep earning today, and driverless vehicles that join the same dispatch queue as the technology and regulation allow. The integration path has four stages: audit the current operation, deploy a fleet platform that handles hybrid dispatch, upgrade the depot for electric charging and sensor care, and retrain part of the staff from driving to fleet operations. Taxi companies that follow this path keep their two strongest assets, local demand and local infrastructure, while the driving itself gradually shifts to machines. The prize is the cost curve: analyst models put robotaxi operating costs on a path toward $0.25 to $0.35 per mile, far below the $1.50 to $2.00 per mile of human-driven ride-hail, and the operators who integrate early capture that margin first.Why integrate instead of waiting it out?Robotaxi services grew from novelty to volume in about two years: Waymo went from roughly 50,000 weekly rides in May 2024 to 500,000 by March 2026, and Apollo Go carried 3.2 million fully driverless rides in the first quarter of 2026 alone. When this wave reaches a new country, the local companies that already run dispatch, depots and customer relationships are either partners for it or bypassed by it. Integration is how an incumbent stays the operator instead of becoming the displaced party.Stage 1: what does a fleet audit cover?The audit maps the operation you already have, because the hybrid fleet inherits it. Four questions matter:
Zones and demand: where do orders concentrate by hour, and which zones have the simple road patterns autonomous vehicles handle best?
Dispatch flow: how do orders reach cars today, and through which app or switchboard?
Depot reality: parking, power capacity, space for chargers, cleaning line.
Economics: cost per kilometer of the current fleet, as the baseline the autonomous cars must beat. Our breakdown of robotaxi economics shows which numbers to collect.
Stage 2: how does hybrid dispatch work?The core move is putting human-driven and autonomous vehicles into one queue under one fleet management software layer. The platform routes simple, well-mapped trips to autonomous cars and keeps complex ones, airport runs with luggage help, unmapped villages, unusual pickups, with human drivers. Riders see one service; the platform decides which kind of vehicle serves each order.Hybrid dispatch also de-risks the rollout. The autonomous share starts at a few percent of orders inside a small geofence and grows only as reliability data accumulates. No customer-facing switch gets flipped; the mix shifts gradually.Stage 3: what changes at the depot?Robotaxis in current deployments are electric, so the depot gains chargers, a scheduled cleaning line (nobody wipes the seats between riders anymore) and a calibration bay for sensors. Charging windows get planned against the demand forecast so vehicles top up during quiet hours. The full daily cycle, charging, cleaning, maintenance, sensor checks, is covered in charging and depot planning.Stage 4: what happens to the people?Roles shift rather than disappear at fleet level. Dispatchers become fleet operators watching telemetry. Experienced drivers make strong remote assistance operators, the people who answer a car's question when it meets a strange intersection, because they know the city's edge cases. Mechanics add high-voltage and sensor skills. In Waymo's published model, remote support runs near 70 assistants for roughly 3,000 vehicles, which shows the shape of the new staffing pyramid: fewer people per car, higher skill per person.What does the sequence look like end to end?

StageWhat runsExit criterion

AuditCurrent fleet, measuredZone map, cost baseline, depot plan
PlatformHuman fleet on the new dispatchDispatchers fluent, data flowing
Pilot AVsSmall geofence, few vehiclesStable utilization, clean incident log
ScaleGrowing AV share in the mixUnit economics beat the human baseline

The platform stage deserves emphasis: adopting the fleet software before any autonomous vehicle arrives means the organization learns the tools on familiar cars, and the eventual AV pilot lands in a running operation. Timing depends on the market; our analysis of when robotaxis reach Georgia maps the preconditions for one specific country, and the broader digitalization story is in taxi digitalization in Georgia.Where aiTAXI fitsaiTAXI is a robotaxi fleet management platform by aiNOW (Tbilisi, Georgia), built around exactly this integration path: hybrid dispatch, telemetry, remote assistance and depot planning in one console that an existing taxi company can adopt before its first autonomous vehicle arrives. The platform is in early access, and the robotaxi fleet management platform pilot program is open for Georgian taxi operators.FAQDo I need to buy autonomous vehicles to start?
No. The audit and platform stages run entirely on the existing human-driven fleet. The point is to be operationally ready before the vehicles are purchasable in your market.Will autonomous cars replace all my drivers?
Not in any realistic planning horizon. Hybrid fleets keep human drivers for complex trips and peak coverage for years, while some staff move into fleet operations, remote assistance and depot roles.Which vehicles will the platform support?
A vehicle-agnostic platform integrates with whichever AV systems enter your market. Committing to a fleet layer does not lock you into one car manufacturer.What is the first concrete step for a taxi company?
A fleet audit: zones, demand curves, dispatch flow, depot capacity and the cost baseline. It costs planning time, produces the integration roadmap, and every later stage builds on it.

---

## Robotaxi fleet management software: the complete guide

URL: https://aitaxi.ge/en/blog/robotaxi-fleet-management-software
Published: Sun Jun 21 2026 00:00:00 GMT+0000 (Coordinated Universal Time)
Author: Andrew Altair

Robotaxi fleet management software is the control layer that runs a fleet of driverless taxis as a business: it assigns rides to vehicles, monitors every car's state in real time, escalates unusual situations to remote human operators, schedules charging and cleaning at the depot, and produces the safety and compliance records regulators expect. The autonomous driving system inside each car handles steering and braking; the fleet platform handles everything a taxi company's dispatchers, mechanics and managers used to coordinate by phone and habit. The two layers are separate products, and the fleet layer is where a taxi operator keeps its leverage: local demand knowledge, depot infrastructure and operations staff. At scale the pattern is visible in operators like Waymo, which coordinates 500,000 paid rides per week across 10 cities through exactly this kind of platform sitting above the vehicles.Why does a driverless fleet need its own software layer?A human driver is a distributed operations department. One person accepts orders, chooses waiting spots, notices a strange noise, decides when to refuel, cleans the seats and calls the office when something breaks. Remove the driver and none of those decisions disappear; they concentrate in software. Understanding what a robotaxi is makes the split clear: the car drives itself, and the fleet platform runs the business around it.The consequence for cost is direct. Analyst models project robotaxi operating costs on a path toward $0.25 to $0.35 per mile against $1.50 to $2.00 for human-driven ride-hail, and most of that gap survives only if the software layer keeps utilization high and downtime short.What does the dispatch module do?Dispatch for driverless cars looks like classic taxi dispatch with the human negotiation removed. The platform matches each ride request to a vehicle using battery state, position, and predicted demand, with no acceptance step and no shift schedule. It rebalances idle cars toward zones where orders will appear, sequences pickups to cut empty kilometers, and routes around closures using the fleet's own recent traces.Key dispatch functions:

Ride matching with battery and position constraints
Predictive rebalancing between city zones
Routing that learns from the fleet's own trips
Hybrid mode: human-driven and autonomous cars in one queue
The hybrid queue is the part that matters for an existing operator. A company does not switch to autonomy in one day; it runs both kinds of vehicles side by side for years, and the dispatch layer has to treat them as one fleet. That transition path is the subject of our guide to adding autonomous vehicles to an existing fleet.What is vehicle telemetry for?Telemetry is the fleet's nervous system. Every vehicle streams battery level, sensor health, door state, cabin camera flags, tire pressure and software version to one screen. The platform turns the stream into alerts with thresholds: a battery trending low before a long trip, a sensor reporting degraded range, a door that failed to latch. Operations staff see problems before passengers do, and maintenance moves from a fixed calendar to condition-based scheduling.How does remote assistance fit in?No automated driving system resolves every scene alone. When a car meets an ambiguous construction layout or a police officer directing traffic by hand, it asks for help, and a remote operator answers with context and guidance. Waymo's Fleet Response works this way: the agent responds to the vehicle's questions while the vehicle keeps control, and reported staffing sits near 70 remote assistants for roughly 3,000 cars. The fleet platform's job is queueing those escalations, giving the operator the right camera views instantly, and logging every intervention. The full picture of how this desk works is in remote assistance for driverless fleets.What happens at the depot?Robotaxi fleets today are electric, so the depot is a charging operation with a cleaning and maintenance line attached. The software plans charging windows against demand forecasts so cars top up when the city is quiet, books cleaning slots between peaks, and sequences maintenance so the fewest vehicles are offline at once. Depot planning is a scheduling problem the platform solves continuously, and we walk through a full day of it in depot operations.What records does the platform keep?Every trip produces a log: route, events, interventions, incidents. The UN's draft global regulation for automated driving, adopted by the UNECE working party in January 2026, is built on a safety-case approach with a safety management system across the vehicle's life, which in practice means operators must show documented, auditable fleet records. A fleet platform generates that documentation from live data instead of reconstructing it by hand after the fact.Module summary

ModuleReplacesCore output

Dispatch and ride matchingDispatcher and driver decisionsAssigned, routed, balanced rides
TelemetryDriver's eyes and earsLive vehicle state, alerts
Remote assistanceDriver judgment in edge casesResolved escalations, logs
Depot and chargingRefueling habits, garage planningCharging, cleaning, maintenance schedule
Safety and compliancePaper trip sheetsTrip logs, incident reports, audit files

Where aiTAXI fitsaiTAXI is a robotaxi fleet management platform by aiNOW (Tbilisi, Georgia): the five modules above in one console, designed for taxi companies that plan to run hybrid fleets in smaller markets rather than for operators building everything in-house. The platform is in early access, and the robotaxi fleet management platform pilot is open for Georgian taxi companies that want to shape the first integration.FAQIs fleet management software the same as the self-driving system?
No. The automated driving system lives in the vehicle and does the driving. Fleet management software lives in the cloud and runs the business: rides, monitoring, depot, records. Operators need both, from the same or different vendors.Can one platform manage vehicles from different manufacturers?
That is the goal of vehicle-agnostic platforms. Integration happens per AV platform through APIs for ride assignment, state streaming and assistance escalation, so a fleet can mix vehicle sources as the market matures.Does the software drive the car remotely?
No. In the standard model the vehicle always keeps control. Remote operators answer questions and give guidance; direct joystick driving is not part of mainstream robotaxi operations.What does a taxi company gain by adopting the fleet layer early?
Operational readiness. Dispatchers learn the console on the existing human-driven fleet, data starts accumulating, and when autonomous vehicles arrive in the market the company plugs them into a running operation instead of starting from zero.

---

## What is a robotaxi and how does it work?

URL: https://aitaxi.ge/en/blog/what-is-a-robotaxi
Published: Thu Jun 18 2026 00:00:00 GMT+0000 (Coordinated Universal Time)
Author: Andrew Altair

A robotaxi is a taxi that drives itself: a passenger car fitted with sensors and an automated driving system that picks up paying riders and completes trips with no human behind the wheel. The vehicle perceives the road with cameras, radar and usually lidar, plans its path with onboard software, and coordinates with a fleet platform that assigns rides, monitors vehicle health and calls a remote human when the car meets a situation it cannot resolve alone. Robotaxis already operate as public, paid services: Waymo reported 500,000 paid rides per week across 10 US cities in March 2026, and Baidu's Apollo Go has crossed 22 million cumulative rides. The technology is commercial reality in a growing list of cities, and the operating playbook behind it is what this article explains. The same stack scales from a single test car to a citywide fleet with no driver in any seat.How does a robotaxi see the road?The perception stack combines three sensor families. Cameras read color and texture: traffic lights, lane markings, signage, brake lights. Radar measures speed and distance directly and works through rain and fog. Lidar fires laser pulses and builds a precise three-dimensional map of everything within a couple hundred meters. The onboard computer fuses these streams dozens of times per second into one model of the scene: where the road is, what every object is, how fast it moves, and where it is likely to go next.On top of perception sits prediction and planning. The software assigns each pedestrian, cyclist and vehicle a set of probable trajectories, then plans a path that keeps safe margins around all of them. The result is a driving style that feels conservative: early braking, wide berths around cyclists, patient behavior at complex intersections.Who assigns the rides if there is no driver?A driverless car removes the person who used to make small operational decisions: which ride to accept, where to wait between orders, when to refuel, when to call the office because something looks wrong. Those decisions move into robotaxi fleet management software. The platform matches ride requests to vehicles, balances the fleet across zones so cars wait where demand will appear, routes around closures, and schedules every car's charging and cleaning windows.This split matters for anyone running taxis today. The autonomous driving system is a product you license or buy with the vehicle. The fleet layer is the part that touches your business: your city, your demand curve, your depot, your dispatchers.What happens when the car gets confused?Every serious operator runs a remote assistance desk. When a robotaxi meets an ambiguous scene, an unusual cone pattern, a hand-signaling police officer, a blocked lane, it slows or stops and asks for help. A remote human looks at the camera feeds and answers the car's question, for example confirming which lane to take. In Waymo's Fleet Response model the agent gives context and guidance while the vehicle keeps control and remains responsible for the safe maneuver, and reporting puts the ratio at roughly 70 remote assistants for about 3,000 vehicles. The economics follow from that ratio: one operations person supports dozens of cars instead of one driver per car.Are robotaxis safe?The largest public dataset says the safety record beats human benchmarks by a wide margin. Across 56.7 million rider-only miles, Waymo reports 92% fewer serious-injury crashes than comparable human drivers, with an any-injury crash rate of 0.41 per million miles against a 2.80 human benchmark, an 85% reduction confirmed in a peer-reviewed study. Regulators are catching up with the evidence: the UN adopted a draft global technical regulation for automated driving systems in January 2026. The full picture, including what it means for smaller markets, is in our guide to autonomous taxi safety and regulation.What does a robotaxi cost to run?Removing the driver removes the largest cost line in a taxi's economics, and replaces it with capital costs (sensors, compute, the vehicle itself) plus fleet operations. Analyst models put the long-run operating cost of a robotaxi on a path toward $0.25 to $0.35 per mile, against $1.50 to $2.00 per mile for human-driven ride-hail. We break the full cost stack down in robotaxi cost per kilometer.Where do robotaxis run today?

OperatorScale (2026)Where

Waymo (Alphabet)500,000 paid rides per week10 US cities
Apollo Go (Baidu)22M+ cumulative rides, 300,000 weekly~27 cities, China and beyond
Tesla Robotaxi~20 unsupervised vehiclesAustin metro, expanding in Texas
Zoox (Amazon)350,000+ riders carriedLas Vegas, expanding to SF, Austin, Miami

The pattern behind the table: scale came first to cities with dense demand, mapped streets and cooperative regulators, and each operator pairs its driving system with its own fleet operation. A fuller operator-by-operator breakdown is in our review of the global robotaxi market.What does this mean for taxi companies?Taxi companies own the two assets robotaxi operators need in every new market: local demand and local operations. The missing piece is the software layer that runs driverless vehicles day to day. aiTAXI is a robotaxi fleet management platform by aiNOW (Tbilisi, Georgia): dispatch, telemetry, remote assistance and depot operations in one console, built so an existing taxi company can run a hybrid fleet. The platform is in early access, and the robotaxi fleet management platform pilot program is open for Georgian taxi operators who want to be first in line.FAQIs a robotaxi the same as a self-driving car?
A robotaxi is a self-driving car used as a commercial taxi. The distinction is the service: robotaxis carry paying passengers in a defined service area, under an operator responsible for the fleet.Does anyone sit in the driver's seat?
In mature deployments, no. Waymo, Apollo Go and Zoox run rider-only trips, and Tesla removed safety monitors from part of its Austin fleet in 2026. New markets usually start with a safety attendant and remove them as the safety case matures.Can a robotaxi drive anywhere?
No. Each fleet operates inside a mapped, approved service area, and expands it zone by zone. Weather, road works and local rules shape the boundary.Who is responsible if a robotaxi crashes?
The operator and manufacturer carry responsibility under the emerging rules, backed by trip logs and sensor records. The UN's 2026 draft regulation formalizes a safety-case approach that makes the operator document and prove safe behavior across the vehicle's life.

---