ZTABS builds AI voice agents for Barcelona enterprises in hospitality, healthcare, banking, telecommunications and public services. Barcelona's trilingual environment with Catalan, Spanish and English creates voice agent requirements that go beyond simple multilingual support. Catalan is the institutional language of the Generalitat de Catalunya and Ajuntament de Barcelona. Voice agents serving Barcelona enterprises evaluate Catalan with the same attention as Spanish and English. ZTABS is remote-first by design from our Houston, Texas office.
Barcelona voice agent demand spans hospitality reservation and concierge services, healthcare scheduling across CatSalut primary care centres (CAPs), banking customer service, telecommunications support and municipal citizen services. Our AI voice agent development services cover enterprise voice automation across markets.
AI voice agent development in Barcelona at a glance
| Need | What we build | Key requirement |
|---|
| Hospitality | Reservation, concierge, group booking | Multilingual guests, PMS integration |
| Healthcare | CAP scheduling, outpatient booking | CatSalut, GDPR Art 9 health data |
| Banking | Account servicing, card management | GDPR, PSD2 |
| Telecommunications | Service enquiries, plan changes | Catalan and Spanish |
| Public services | Citizen enquiries, appointment booking | Catalan institutional language |
How voice agents work
A voice agent receives or places calls through telephony infrastructure. Audio converted to text by STT. Text processed by language model. Response converted to speech by TTS.
STT selection. Provider and model evaluated for language, accent, noise and channel. Barcelona callers speak Catalan, Spanish and English. Tourism contexts may involve additional languages. STT accuracy varies by provider, model, accent and background noise conditions. Evaluation uses held-out test sets per agreed language with baseline, error categories and median, P95 and P99 metrics.
TTS selection. Provider and model for naturalness, latency and language. Catalan TTS evaluated for natural Catalan phrasing rather than machine-translated Spanish. Spanish and English voices selected for appropriate register per context.
Model selection. Language model for the specific task. Hospitality enquiries need different model behaviour than healthcare scheduling. Catalan-Spanish-English code-switching for brand names and local terms. Model size balanced against latency.
Building a voice agent for Barcelona hospitality or healthcare?
Tell us about your call volume, use case and compliance requirements. We start with assessment covering telephony, model evaluation and regulatory mapping.
Inbound versus outbound
Inbound voice agents answer calls. Risk-approved identity verification before revealing protected information. Handle within the owner-approved action envelope. Transfer to human when exceeding scope.
Outbound voice agents place calls for reminders, notifications or follow-ups. Automated outbound marketing calls require assessment under AEPD guidance and applicable Spanish marketing regulations. Service notifications are distinct from marketing. Recording consent disclosed at call start.
Conversational design
Identity and verification. Before revealing account details, health information or service status, verify caller identity through risk-approved methods. CLI, phone number, static date-of-birth, address or policy number alone are not adequate authentication for protected actions. Match verification level to action risk, enforce tenant-level and object-level authorisation.
Catalan, Spanish and English. The voice agent detects the caller's language within initial speech and routes to the appropriate conversational path. Catalan is a distinct language with its own grammar, vocabulary and pragmatics. The voice agent is evaluated for natural Catalan, not machine-translated Spanish. Language quality in each evaluated language is an agreed design goal with held-out test sets and acceptance criteria, not a native-level fluency guarantee across unlimited languages.
Speech endpointing. Configured for the syntax patterns of each language. Barge-in detection. Agent stops and processes.
Repair and confirmation. Low-confidence triggers confirmation. Read-back of numbers, dates, reference codes. Explicit confirmation before consequential actions.
AI identity and recording disclosure. The voice agent identifies itself as an automated system at call start and offers a human alternative, as an actual design choice separate from recording consent. Recording disclosure and consent follow immediately. EU AI Act Article 50(1) assigns the provider the obligation to ensure that AI systems intended to directly interact with natural persons are designed and developed so that the natural person is informed they are interacting with an AI system, unless obvious from context; paragraph 5 specifies clear, distinguishable, accessible notice at first interaction. This applies from 2 August 2026. The buyer may itself be the provider depending on actual role; buyers should identify the actual provider, agree how the notice obligation is implemented and operated, and document the disclosure design.
Latency. Response time varies by the STT, model and TTS provider chain. The project measures median, P95 and P99 end-to-end latency on agreed scenarios during evaluation, not as a fixed guaranteed response time.
Recording consent
GDPR applies directly in Spain as an EU member state. LOPDGDD (Ley Organica 3/2018 de Proteccion de Datos Personales y garantia de los derechos digitales) supplements GDPR with Spanish-specific provisions. The agreed consent-first architecture discloses recording at call start and obtains explicit consent as a design choice; this is distinct from determining the actual GDPR lawful basis for each processing activity. Callers who decline cannot have audio capture or live STT silently continue; architecture provides a lawful alternate path. Even without stored audio recording, live STT transmission and transcript storage remain privacy processing subject to GDPR lawful-basis assessment.
Safe bounded workflow
Low confidence. Transfer to human when intent unclear.
Unsupported intents. Acknowledge, explain limitation, offer transfer.
Refusal and opt-out. Human transfer through staffed queue during agreed service hours. No forced bot interaction. Outside staffed hours: voicemail, callback where supported.
Caller distress. Medical emergency or acute distress: tested escalation path. Urgent callers can use SEM (Servei d'Emergencies Mediques) and independent emergency services (112). Voice agent is not emergency service and does not delay emergency access.
Failed transfer. Callback where staffing supports it.
Unknown outcome. Reconcile current remote object state and authoritative local ledger before retry. Bounded retries with limits and manual exception path. Receiver-side deduplication and atomic local updates.
Hospitality voice agents
Barcelona's tourism economy generates significant voice agent demand across hotels, restaurants and activity providers.
Reservation handling. Booking, modification and cancellation across multi-property hotel groups. Real-time availability from PMS through operator-approved API integration with tenant and object-level authorisation. Caller language detected and conversation conducted in that language. Direct booking benefits communicated where applicable.
Concierge services. Restaurant recommendations, activity booking, transportation arrangements and local information. Integration with activity partners' booking systems where partner-approved APIs are available. Personalisation based on guest reservation data and, for returning guests, preference history accessed through operator-approved programme.
Group and event booking. Meetings, incentives, conferences and events. Group rate enquiries, room block management and event coordination. Complex multi-party bookings confirmed after authoritative system acknowledgment.
PMS and channel manager integration. Write permissions require the operator's approved programme. Booking confirmed after authoritative system acknowledgment. Integration with channel managers for rate and availability synchronisation scoped per operator's actual data flows.
Healthcare voice agents
Barcelona healthcare facilities and the Catalan health system handle significant appointment and enquiry volumes.
CAP scheduling. Centres d'Atencio Primaria (primary care centres) across Barcelona handle appointment booking for the public health system. Integration with CatSalut scheduling systems requires the provider's approved programme, tenant roles and permissions, not a generic public API promise.
Hospital outpatient scheduling. Managing appointment logistics for Barcelona hospitals. Write permissions require the provider's approved programme. Booking confirmed after authoritative system acknowledgment.
Patient intake. Demographics, insurance details (CatSalut or private), chief complaint. Health data encryption. Minimum necessary access. Audit logging. GDPR Article 9 special category data provisions apply.
La Meva Salut portal. La Meva Salut is the patient-facing portal for CatSalut. Integration with this portal is scoped to the provider's actual approved data flows and vendor programme, not equivalent to direct EHR API access.
Banking and financial services voice agents
Barcelona's banking sector includes major institutions with significant customer service volumes.
Account servicing. Balance enquiries, transactions, payment processing. Risk-approved identity verification. Integration with core banking through institution-approved APIs.
Card management. Card blocking, limits, replacement. Multi-factor verification for consequential actions. PSD2 strong customer authentication assessed per applicable requirements.
Integration requirements
Healthcare systems. CatSalut scheduling, hospital information systems through provider-approved APIs. Health data handling per GDPR Article 9.
Hospitality systems. PMS, channel manager, activity partner booking systems through operator-approved APIs.
Banking systems. Core banking, CRM through institution-approved APIs.
Telephony. SIP trunking, Spanish numbers, concurrent capacity. Provider selection for regional quality.
Voice-specific cost drivers
Voice agent development follows the custom software cost guide as general platform engineering context.
| Cost driver | Description |
|---|
| Telephony infrastructure | SIP trunking, Spanish numbers, concurrent capacity |
| STT provider usage | Per-minute by provider, model and language |
| TTS provider usage | Per-character or per-minute by voice and language |
| Model inference | Per-token pricing, scales with volume |
| Concurrency | Simultaneous capacity licensing |
| Recording and storage | Consent-gated configured coverage, transcript storage |
| Monitoring and evaluation | Quality monitoring, accuracy tracking |
Telephony, STT, TTS, LLM inference, minutes, concurrency and retention are separately scoped per engagement without invented operating quotes.
Illustrative Example
Hypothetical: Barcelona hotel group reservation voice agent
A Barcelona hotel group operating multiple properties receives thousands of weekly reservation calls in Catalan, Spanish and English, plus additional guest languages during peak tourism season.
A voice agent would answer inbound calls, detect language, verify identity through the operator's approved methods, query PMS through approved APIs with tenant and object-level authorisation, provide availability and rate information, process bookings and send confirmation in the guest's language. Recording consent disclosed at call start with explicit opt-in. Concierge enquiries routed to the knowledge base with curated local recommendations.
Assumptions: operator PMS API access confirmed, service scope documented, agreed consent-first recording architecture with lawful alternate path per GDPR, telephony supports SIP, and agreed language scope covers Catalan, Spanish and English with additional languages evaluated per actual caller demographics and available STT provider coverage. Residual risks include STT accuracy variation across accents and background noise, language detection edge cases during code-switching, PMS integration latency during peak booking periods and caller preference for human agents. No guarantee of specific automation rate, handle time reduction or satisfaction outcome.
This is a hypothetical scenario for illustration. Actual scope, timeline and cost depend on your requirements.
Cost and timeline
Voice agent development as general custom platform engineering uses four guide tiers.
| Scope | Cost range | Timeline |
|---|
| Internal Tool | $25K to $60K | 6 to 14 weeks |
| Business Application | $60K to $150K | 14 to 28 weeks |
| Enterprise Platform | $150K to $300K | 28 to 40 weeks |
| Large-Scale System | $300K to $500K+ | 40 to 60+ weeks |
2026 US vendor benchmark
Ranges are 2026 US-buyer guide-based planning benchmarks for a mid-complexity business application: multi-role access, 3 to 5 integrations, custom workflows and a modern web frontend.
| Vendor type | Typical cost | Timeline |
|---|
| Freelancer or dev collective | $20K to $75K | 4 to 10 months |
| Offshore agency | $40K to $120K | 5 to 11 months |
| Nearshore agency | $70K to $200K | 4 to 8 months |
| US agency (ZTABS tier) | $90K to $280K | 4 to 8 months |
| Big 4 or systems integrator | $250K to $800K+ | 8 to 16 months |
Ranges are 2026 US-buyer guide-based planning benchmarks for the scoped business-application comparison. Where applicable to the agreed engagement, legacy-system integration (SAP, Oracle or mainframe interfaces) and compliance engineering or readiness preparation (SOC 2, HIPAA safeguards or PCI DSS controls) carry an additional guide allowance of $20K to $80K across vendor paths. Independent CPA SOC 2 examination and opinion are scoped and priced separately from internal readiness engineering. The guide uses 15 to 25 percent of build cost annually as an ongoing-maintenance planning estimate; actual support scope and fees are agreed separately. These are scoped planning allowances, not an empirical market survey or voice-package rate card.
Need a scoped estimate for Barcelona voice agent development?
Share your use case, call volume and compliance requirements. We provide a scoped estimate after reviewing both four-tier scope and vendor benchmark context.
Proposal evaluation criteria
- Relevant voice workflow experience and customer-approved action authority with supporting evidence
- Engineering and integration reliability including unknown-outcome recovery, bounded retries and manual exception paths
- Assessed privacy, security and regulatory boundaries with clear independent assurance responsibilities
- Speech, language and accessibility evaluation with held-out test sets and agreed acceptance criteria
- Transparent scope, milestones, change control, operating fees and ongoing support terms
- Human handover, training, validation and support with actual customer and provider responsibility allocation, agreed staff and user languages, and documentation and translation responsibilities
Spanish and EU regulation
GDPR. Applies directly in Spain as an EU member state. Controllers or processors established in the EU, or offering goods or services to or monitoring individuals in the Union. Lawful basis under Article 6. Special categories under Article 9 for health data. Purpose limitation, data minimisation, storage limitation, data subject rights and cross-border transfer restrictions.
LOPDGDD. Ley Organica 3/2018 de Proteccion de Datos Personales y garantia de los derechos digitales supplements GDPR with Spanish-specific provisions. AEPD (Agencia Espanola de Proteccion de Datos) is the supervisory authority.
Recording consent. The agreed consent-first architecture discloses recording at call start and obtains explicit consent. Callers who decline cannot have audio capture or live STT silently continue. Lawful alternate path for refusal. Live STT and transcripts subject to GDPR lawful-basis assessment.
Outbound marketing. Automated outbound marketing calls assessed under current AEPD scope. Service calls distinguished from marketing. Actual current regulatory scope per engagement.
Catalan language. Catalan is the institutional language of the Generalitat de Catalunya and Ajuntament de Barcelona. Public services operate in Catalan. The voice agent's Catalan coverage is an agreed design scope evaluated with held-out tests, not a universal requirement imposed on all private businesses.
How delivery works
ZTABS is remote-first by design from our Houston, Texas office. Houston is Central time; Barcelona is Central European Time, usually 7 hours ahead and 6 hours during brief US-EU daylight-saving transition weeks. Houston morning overlaps with Barcelona afternoon for coordination.
When ZTABS is not the right fit
- Simple IVR tree with no AI for this engagement. Traditional IVR vendors handle press-1 menus.
- On-site embedded team for this engagement. ZTABS is remote-first by design.
- Speech model research for this engagement. Voice agent delivery uses evaluated STT providers; fundamental speech research is separately scoped.
- Clinical decision support for this engagement. Routing to clinicians, not autonomous decisions.
- AEPD regulatory consulting for this engagement. Compliance engineering is part of delivery; independent regulatory advice is separately agreed.
White-label AI products
ZTABS builds custom white-label AI products for clients to brand and resell. Barcelona enterprises can commission white-label voice AI platforms for distribution.
Evaluation methodology
Representative test calls. Expected uses, edge cases, accents, noise, channel quality.
Held-out language sets. Catalan, Spanish, English. STT accuracy on agreed test set per language with baseline, error categories and median, P95 and P99 metrics.
Task completion. Intent identified and action completed. Human baseline comparison.
False action and false refusal. Tracked separately.
Adversarial prompts. Injection, social engineering, authorisation bypass.
Privacy leakage. Tested to confirm tenant isolation prevents cross-caller data exposure, with residual-risk handling if leakage is detected.
Monitoring and operations
Real-time dashboards. Call volume, completion rates, transfer rates, STT accuracy, latency. Per-language metrics.
Model updates. Tested in staging. A/B testing. Rollback. Healthcare and financial model updates reviewed by designated authority.
Recording and transcription. Permitted coverage per consent architecture. Minimisation, access controls, retention, deletion and legal hold per engagement. No automatic cross-customer data reuse. No automatic reuse of recordings for training; agreed lawful purpose, rights, data-processing terms, model configuration, retention, deletion, legal holds and approved test data per engagement.
Cost monitoring. Per-call costs across providers. Monthly reporting.
Data governance
Scope authorised tenant and object data access and actual provider product, API, programme, version and licence. Retrieval and writes are separate. Permission changes propagate across indexes, embeddings, caches, logs and fine-tunes as applicable with disclosed residual lag and mitigations. No automatic reuse of recordings, PHI, customer research, employee chat, proprietary code or licensed publications for training. Agreed lawful purpose, rights, data-processing terms, model configuration, retention, deletion, legal holds and approved test data per engagement.
Handover
Voice agent handover includes runbooks, telephony configuration, model versions, provider accounts, monitoring dashboards, evaluation methodology, consent architecture and escalation procedures.
Ready to build a Barcelona voice agent?
Share your call volume, use case and compliance requirements. We start with assessment covering telephony, model evaluation and Spanish regulatory mapping.