An AI voice system from phone order to dispatch data.
This brief explains the industry problem, technical approach, high-level architecture, current Pilot status, public evidence, limitations, and next-stage R&D for AnyTaxi. It excludes customer names, recognition rates, SLAs, and revenue.
Version 2026-10-24 · Status: Pilot · Public level: summary
01
Industry problem
Information is lost between the call and dispatch.
Fleet phone orders span calls, address capture, order entry, and dispatch. Limited peak staffing, spoken-address variation, and fragmented tools create repeated checks, delays, and untraceable operations data.
Peak call volume exceeds staffing; missed calls can be lost.
Addresses, times, and notes are repeated and re-keyed.
Phone notes, messaging, and dispatch each hold fragments.
Without structured records, analysis and reconciliation are difficult.
02
Technical approach
Split the call into inspectable, reviewable, replaceable stages.
AnyTaxi is not a single model; it is a workflow from the phone entry point to operations data. Each stage has failure handling and human intervention so ambiguous information does not flow directly into dispatch.
Telephony and voice entry
SIP/IVR, human transfer, and call records with a traceable processing chain.
Speech-to-text
Handle call-audio quality, accents, and noise; retain timestamps.
Address and field extraction
Extract pickup, time, passenger count, and notes, including spoken variants.
Confidence and human review
Low-confidence or conflicting details are confirmed by a person before order creation.
Orders and dispatch
Create structured orders and connect them to dispatch and the fleet back office.
Operations data and learning
Event records, manual annotation, and observation; automated candidates are not activated.
03
High-level architecture
Layered design so voice, orders, and dispatch can be validated separately.
This is a concept architecture; deployment details and sensitive information are omitted. Actual deployment depends on the field and security requirements.
Voice entry
Inbound callVoice workflow
Voice processing
Speech-to-textIntent & field extraction
Operations core
Order serviceDispatch interfaceHuman review
Data & observability
Operations dataEvent logTest samples
Cross-cutting: access control · logging · monitoring · data minimisation
Concept architecture. Deployment details and sensitive information are intentionally omitted.
04
Current status
Completed, in validation, and in R&D are stated separately.
AnyTaxi remains at Pilot. Telephony and the fleet back office have operating records; recognition, address parsing, dispatch, and end-to-end reconciliation remain under validation.
Voice order intake and call recordsCompleted
Production phone lines, human transfer, and call records are in operational use. SLA and concurrent-call capacity are not published.
Completed
Address and destination parsingIn validation
Field use continues alongside low-confidence follow-up design. Per-language recognition and order-creation rates are not published.
In validation
Order management and dispatchIn validation
The dispatch engine and fleet back office are used in the field. End-to-end reconciliation from voice to order to dispatch remains under validation.
In validation
Fleet back office and operations dataCompleted
Call handling, extensions, and human voice review have release and operating records. Page-level permissions and latest releases still require periodic acceptance.
Completed
Driver and passenger appsIn validation
Device and field acceptance continue. Code existence is not treated as proof of full rollout.
In validation
Voice learning and automated candidatesIn development
Observation and manual annotation continue. Automated candidates are not yet activated in production.
In development
Multi-fleet and white-label expansionIn development
Multi-tenant isolation and onboarding a second fleet still require testing and field evidence.
In development
05
Testing and evidence
Only approved summaries are public; website tests are not product validation.
The evidence below comes from the October 2026 internal audit and field-record review. Raw reports, full test records, and customer data are not automatically published.
ID
Technology
Confirmed
Unverified / limits
EV-ATX-0012026-10-24
AnyTaxi telephony and fleet back office
Production phone lines, human transfer, call records, and the fleet back office have operating records.
SLA, concurrent-call capacity, every interface, and the complete permission matrix were not re-accepted in this round. Customer names, call content, internal URLs, and raw records are not published.
EV-ATX-0022026-10-24
AnyTaxi speech and address parsing
The speech and address workflow has field use and a human-review design.
Per-language recognition, address accuracy, order-creation rate, and latency are not published. Website tests or concept mockups are not speech-recognition results.
EV-ATX-0032026-10-24
AnyTaxi orders and dispatch
The dispatch engine, order management, and fleet back office are used in the field.
End-to-end reconciliation from voice to order to dispatch, full-automation success, and cross-fleet isolation remain under validation. No full-automation dispatch or commercial-outcome claim is made.
EV-ATX-0042026-10-24
AnyTaxi voice learning platform
Observation and manual annotation workflows exist and continue to be used.
Automated candidates are not activated; model versions, production deployment, and performance are not published. Offline annotation or planned automated learning is not a deployed feature.
Website build/static/browser tests prove website engineering only, not AnyTaxi recognition rates, SLAs, or commercial results.
06
Known limitations
Make limitations visible before scaling.
Call-audio quality, accents, and noise affect recognition.
Spoken address variants and local names require normalisation and review.
End-to-end reconciliation, cross-fleet isolation, and automated dispatch remain under validation.
Automated voice-learning candidates are not activated.
Recognition rates, SLAs, customer outcomes, and commercial data are not published.
AnyTaxi pricing, revenue, customer count, and signed contracts are not published. The following distinguishes current activity from planned work.
Taxi fleets and transport operators
Phone orders depend on staff, addresses are re-keyed, and dispatch and operations data are fragmented.
Small-bus, rehabilitation, and care transport
Scheduling, mileage, fares, and subsidy reporting require heavy manual reconciliation.
Airport transfer and corporate charter
Bookings, dispatch, notifications, and reconciliation need a consistent data source.
Enterprises adopting AI automation and integration
Voice, documents, and existing systems lack traceable automation.
Fleet field cooperation and pilots
Validate AnyTaxi in a defined field with agreed metrics; customer names and field data are not published.
Current
Custom software and system integration
Progress through discovery, milestones, and acceptance criteria; no fixed online quote.
Ongoing
Multi-fleet and white-label cooperation
Multi-tenant isolation, second-fleet onboarding, and white-label workflows still require testing and field evidence.
Planned / R&D
Technology and research partnership
Speech, address handling, dispatch integration, and AI governance are assessed with ownership and data boundaries.
By discussion
Planned / R&D items are not realised revenue or live services.
Disclosure boundaries
Public summaries and review-only material are managed separately.
Raw audit reports, full test records, field data, customer data, contracts, and internal system details are not automatically published. For SBIR/startup or formal partnership review, email us with the purpose and required items.
Customer names, personal data, contracts, and keys are not published.
Unconfirmed recognition rates, SLAs, revenue, or outcomes are not published.
Website and product tests are stated separately.
Work with us
Need AnyTaxi technical material or field cooperation?
Email [email protected] with the purpose, required material, and timeline.