Voice FAQ

Voice is an AI phone receptionist. When someone calls your business phone number, Voice answers — just like a human receptionist would. It greets the caller, understands what they need, answers questions from your knowledge base, books appointments, and transfers to a human when needed.

The caller hears a natural, human-sounding voice — not a robotic IVR ("Press 1 for sales, Press 2 for support"). It's a real conversation.

And it works 24/7/365. No sick days, no lunch breaks, no missed calls.

Missed calls cost businesses real money. Industry estimates suggest 20-40% of inbound calls to small and medium businesses go unanswered. Every missed call is a lost appointment, a lost sale, or a lost referral.

Voice solves three problems:

1. Revenue leakage: Every call is answered. No more voicemail tag. No more "sorry, we missed your call" callbacks.
2. Staffing cost: A full-time receptionist costs $30K-$50K+/year. Voice handles the same workload for a fraction of the cost — and never takes a day off.
3. After-hours gaps: Most businesses can't staff phones after 6pm or on weekends. Voice can — 24/7.

Here's what Voice handles on a typical call:

• Answer & greet: Picks up within 3 rings with a natural greeting ("Hi, thanks for calling Westside Dermatology. This is Ava. How can I help?")
• Appointment booking: Checks your calendar, proposes available slots, confirms the booking, sends a confirmation SMS/email
• Q&A from your knowledge base: Answers questions about your services, hours, pricing, policies, location, insurance accepted — anything in your documents
• Call routing: Transfers to the right person or department based on the caller's need
• Urgent call handling: Detects urgency and pages the on-call person immediately
• Voicemail capture: If the issue can't be resolved and no human is available, takes a detailed message
• After-hours handling: Different behavior for business hours vs. after-hours vs. holidays

The voice sounds natural — most callers won't realize it's AI unless they ask. However, we recommend transparency:

• You can configure the greeting to say "I'm the AI assistant" if you prefer honesty upfront
• If a caller asks "Are you a robot?", Voice will answer honestly
• If a caller requests a human, Voice transfers immediately — no deception

The goal isn't to trick callers — it's to give them a great experience. If they want a human, they get one. If they're happy with the AI, they get their issue resolved faster.

Completely different experience:

Voice works with your existing phone number. You have two options:

1. Port your existing number: Transfer your current business number to Voice. Callers dial the same number they always have.
2. Get a new number: Purchase a new number through our carrier partners (Twilio, Telnyx, Vonage) and use it alongside your existing number.

You can also use your own SIP trunk if you have one. Voice is telephony-agnostic — it works with any SIP-based phone system.

Yes. You can configure Voice to handle:

• Multiple phone numbers: Different numbers for different locations or business lines
• Department routing: "Sales," "Support," "Billing" — Voice routes callers to the right team
• Multiple agents: Different AI personas for different departments (e.g., a sales-focused agent vs. a support-focused agent)

All managed from a single admin console.

Three things can happen:

1. Transfer to human: Voice warm-transfers the call to a human agent (with full context — the human knows what the caller said). You configure which numbers/departments to transfer to.
2. Voicemail capture: If no human is available, Voice takes a detailed message (name, number, reason for calling) and notifies your team via SMS/email/Slack.
3. Callback scheduling: Voice can schedule a callback for the caller at a specific time.

Voice never leaves a caller hanging. If it can't help, it makes sure a human follows up.

Voice handles the routine, repetitive calls — appointment booking, FAQs, after-hours coverage, call routing. This frees your human staff to focus on:

• In-person customers (front desk)
• Complex calls that need judgment or empathy
• High-value interactions that benefit from a human touch

Most customers find that Voice handles 80-85% of calls without human intervention. The remaining 15-20% are transferred to humans — but now your staff has time for those calls because they're not buried in routine bookings.

Think of Voice as a force multiplier, not a replacement.

Voice is designed for low-latency conversation:

• Call answer time: Under 800ms (half a second) from when the phone starts ringing
• Response latency: Under 900ms (typical) from when the caller stops speaking to when Voice starts responding
• 95th percentile: Under 1.5 seconds even in worst-case scenarios

This is fast enough to feel natural. For comparison, human receptionists typically take 1-3 seconds to respond. Voice is actually faster than most humans.

Yes. Voice supports barge-in (interruption handling). If the caller starts speaking while Voice is talking, Voice stops immediately and listens to the caller.

This is critical for natural conversation. Humans interrupt each other all the time — Voice handles it gracefully.

Voice uses high-quality neural text-to-speech (TTS) — the same technology used by premium voice assistants. The voices sound natural, with proper intonation, pacing, and emotion.

You can choose from a library of preset voices:

• Male and female voices
• Different accents (US, UK, Australian)
• Different tones (friendly, professional, warm, authoritative)

Voice uses state-of-the-art speech recognition that handles a wide range of accents, dialects, and speech patterns. It's trained on diverse datasets covering:

• Regional accents (Southern US, British regional, Australian, etc.)
• Non-native English speakers
• Background noise (call centers, streets, cars)
• Fast or slow speech

If Voice can't understand the caller after multiple attempts, it will:

1. Politely ask the caller to repeat or rephrase
2. Offer to transfer to a human
3. Take a detailed voicemail

Voice never pretends to understand when it doesn't.

Voice handles silence and ambiguity gracefully:

• Short pause (2-3 seconds): Voice waits — the caller may be thinking or looking something up
• Longer pause (5+ seconds): Voice gently prompts: "Are you still there?" or "Is there anything else I can help with?"
• No response: After multiple prompts, Voice offers to take a message or end the call politely

This prevents awkward dead air and ensures the caller doesn't feel abandoned.

Yes. Voice includes noise suppression and echo cancellation to handle real-world call conditions:

• Background noise (traffic, crowds, machinery)
• Echo (speakerphone, car Bluetooth)
• Low-quality phone connections

The speech recognition model is trained on noisy data and can extract speech even in challenging conditions. If the audio quality is too poor, Voice will ask the caller to call back from a quieter location or offer to continue via text (SMS/chat).

Yes. You configure Voice through the admin console:

• Greeting: Custom greeting text ("Thanks for calling [Business Name]. This is [Agent Name]. How can I help?")
• Business hours: Different greetings for business hours vs. after-hours vs. holidays
• Tone: Friendly, professional, formal, casual — matches your brand voice
• Behavior: Configure what Voice should and shouldn't do (e.g., "Always offer to book an appointment," "Never discuss pricing — transfer to sales")

You can preview and test the agent before going live.

Voice uses neural TTS (text-to-speech) — the latest generation of voice synthesis. It sounds natural, with:

• Proper intonation and emotion
• Natural pacing and pauses
• Conversational fillers ("um," "let me check") when appropriate
• Ability to express empathy ("I'm sorry to hear that")

Most callers can't tell the difference between Voice and a human receptionist — especially on a phone call where audio quality is limited.

Voice competes on privacy + integration, not just voice quality. Raw voice realism is table-stakes — every competitor has it. What sets Voice apart is sovereignty and suite integration.

These are all excellent voice AI platforms, but they have different design priorities:

Choose Retell/Vapi/Bland if you're a developer building a custom voice app. Choose Voice if you're a regulated business that needs sovereignty, compliance, and out-of-the-box integration with your calendar/CRM.

Voice integrates with your calendar via Loop workflows (ODW.ai's automation engine):

1. Caller says "I'd like to book an appointment"
2. Voice checks your calendar availability (Google Calendar, Outlook, Cal.com, etc.)
3. Voice proposes available slots to the caller
4. Caller picks a slot
5. Voice books the appointment, sends a confirmation SMS/email, and creates a record in your CRM

All of this happens in real-time during the call. The caller doesn't need to hang up and wait — the appointment is confirmed before the call ends.

Voice integrates with Vault (ODW.ai's knowledge base). When a caller asks a question:

1. Voice queries Vault for relevant documents (your service menu, pricing, policies, FAQs, etc.)
2. Vault returns the most relevant chunks
3. Voice generates an answer grounded in those documents

This means Voice answers questions from your own documents — not from generic AI training data. If your documents say "We're closed on Sundays," Voice will say "We're closed on Sundays." If your documents don't mention something, Voice won't guess — it will say "I'm not sure, let me transfer you to someone who can help."

"Sovereign deployment" means Voice runs on your own infrastructure — your servers, your cloud, your data center. Call recordings, transcripts, and all data stay on your infrastructure.

This matters for:

• Healthcare (HIPAA): Call recordings may contain protected health information (PHI). You can't send that to a third-party cloud without violating HIPAA.
• Legal (attorney-client privilege): Call transcripts may contain privileged information. You need to control where that data lives.
• Financial services: Call recordings may contain financial advice or sensitive client information.
• General privacy: Some businesses simply don't want call data on someone else's servers.

Voice offers three deployment modes:

1. SaaS: Hosted by ODW.ai (convenient, but data on our servers)
2. Private cloud: Deployed in your AWS/GCP/Azure VPC (data on your cloud account)
3. On-premises: Deployed in your data center, even air-gapped (maximum sovereignty)

Yes. Voice is model-agnostic — you choose which AI model powers the conversation:

• Local models: Run on your own hardware (maximum privacy, no per-call cost)
• Private cloud models: Deployed in your VPC (good balance of privacy and quality)
• Frontier models: OpenAI, Anthropic, Google (highest quality, but data leaves your infrastructure)

You can even configure hybrid routing: use local models for routine calls (FAQs, booking) and frontier models for complex calls (complaints, edge cases).

Voice is business-hours aware. You configure:

• Business hours: Monday-Friday 9am-6pm, Saturday 10am-2pm, etc.
• Holidays: List of holidays when the office is closed
• After-hours behavior: What Voice should do outside business hours (e.g., "Offer to book an appointment for tomorrow," "Take a message," "Transfer to on-call")

Voice automatically adjusts its behavior based on the time of day and day of week. Callers during business hours get the full experience. Callers after hours get a tailored message and options.

Not in v1.0. Voice v1.0 is inbound-only — it answers incoming calls but doesn't initiate outbound calls.

Outbound calling is planned for v1.1 and will include:

• Callbacks: "We'll call you back at 3pm" → Voice calls the customer at 3pm
• Appointment reminders: Voice calls customers 24 hours before their appointment to confirm
• Follow-ups: Voice calls leads who requested information

Outbound calling requires careful compliance controls (TCPA in the US, GDPR in Europe) to avoid spam calls. We're building the guardrails first.

Voice uses state-of-the-art speech recognition with typical accuracy of 95%+ for clear speech in English. Accuracy depends on:

• Audio quality: Clear phone connections → higher accuracy. Noisy environments → lower accuracy.
• Accent: Common accents (US, UK, Australian) → 95%+. Rare or heavy accents → 85-90%.
• Speech clarity: Clear, moderate-paced speech → higher accuracy. Mumbling or very fast speech → lower accuracy.

If Voice can't understand the caller, it asks for clarification or offers to transfer to a human. It never guesses when it's unsure.

Voice uses retrieval-augmented generation (RAG) — it answers from your documents, not from memory:

1. Caller asks a question
2. Voice searches your knowledge base (Vault) for relevant documents
3. Voice generates an answer grounded in those documents
4. If the documents don't contain the answer, Voice says "I'm not sure" and offers to transfer

Voice never makes up information. If it doesn't know, it says so.

Voice includes multiple safeguards:

1. Confirmation: Before booking, Voice reads back the details: "So that's Thursday at 2pm with Dr. Patel. Is that correct?" The caller confirms or corrects.
2. Post-call summary: After every call, Voice generates a structured summary (intent, outcome, key details). Your team can review these summaries and catch errors.
3. Human override: If a caller requests a human, Voice transfers immediately. The human can correct any mistakes.
4. Audit trail: Every call is logged (transcript, recording, structured data). If something goes wrong, you can review exactly what happened.

Mistakes are rare, but when they happen, they're caught and corrected quickly.

Yes. Every call generates:

• Transcript: Full text of the conversation (caller + Voice)
• Recording: Audio recording of the call (optional, consent-aware)
• Structured summary: Intent, outcome, key entities (name, phone, appointment time, etc.)

All of this is available in the admin console. You can:

• Search calls by date, caller, intent, outcome
• Listen to recordings
• Read transcripts
• Export data for analysis

Voice includes PII detection and redaction (via ODW.ai Shield):

• PII detection: Voice scans the conversation for sensitive information (names, phone numbers, health info, credit card numbers)
• Redaction: Sensitive information is redacted from transcripts and logs (e.g., "John Smith" → [NAME])
• Routing: If sensitive information is detected, Voice can route the call to a local model only (no cloud API)

For healthcare (HIPAA), legal (privilege), and financial services, Voice ensures sensitive data never leaves your infrastructure unless you explicitly configure it to.

Yes. Voice includes a preview and testing mode:

• Sandbox mode: Test Voice with simulated calls in a sandbox environment. No real calls are made.
• Demo number: Call a test phone number to hear Voice in action. Try different scenarios (booking, Q&A, transfer).
• Shadow mode: Voice processes real calls but doesn't respond — instead, it generates draft responses for human review. This lets you evaluate Voice performance before enabling auto-response.

We recommend testing thoroughly before going live. Most customers spend 1-2 days testing before enabling Voice for production calls.

Voice targets a task completion rate of 85%+ — meaning 85% of calls achieve the caller's stated intent without human transfer.

Typical outcomes:

• Appointment booking: 70%+ of booking requests result in a confirmed booking
• Q&A: 80%+ of questions answered correctly from the knowledge base
• Call routing: 90%+ of calls routed to the right department/person
• Overall: 85%+ of calls resolved without human intervention

The remaining 15% are transferred to humans — either because the caller requested a human, or because Voice couldn't handle the request (low confidence, complex issue, etc.).

Voice is compatible with HIPAA when deployed correctly. HIPAA compliance depends on how you deploy Voice:

• Private cloud or on-premises: ✅ Fully HIPAA-compliant. Call recordings and transcripts stay on your infrastructure. You control the BAA (Business Associate Agreement).
• SaaS (ODW.ai-hosted): ⚠️ Possible, but requires a BAA with ODW.ai. Contact us for details.

Voice includes HIPAA-specific features:

• PHI redaction: Automatically redacts protected health information from transcripts
• Consent prompts: Configurable consent prompts before recording calls
• Data retention: Configure how long call data is kept (e.g., "delete after 7 years")
• Audit logs: Every call, every access, every configuration change is logged

Yes. Call recordings are stored with enterprise-grade security:

• Encryption at rest: AES-256 encryption for all recordings
• Encryption in transit: TLS 1.3 for all network communication
• Access controls: Only authorized users can access recordings (role-based access control)
• Audit trail: Every access to a recording is logged

For sovereign deployments (private cloud, on-premises), recordings stay on your infrastructure. ODW.ai has zero access.

Yes. Voice includes consent-aware recording:

• Consent prompts: You can configure Voice to announce "This call may be recorded for quality and training purposes" at the start of the call
• Opt-out: If the caller says "I don't consent to recording," Voice can stop recording and continue the call (or transfer to a human)
• Regional compliance: Different regions have different consent laws (one-party vs. two-party consent). Voice can be configured to match your jurisdiction's requirements

Recording is optional — you can disable it entirely if you prefer.

Yes. Voice supports data deletion:

• Individual deletion: Delete specific calls (e.g., if a caller requests deletion under GDPR)
• Bulk deletion: Delete all calls for a specific customer or date range
• Automatic retention: Configure retention policies (e.g., "delete calls after 2 years") — Voice automatically deletes old calls

All deletions are logged in the audit trail for compliance.

Your data is yours. If you stop using Voice:

• Export: You can export all call data (transcripts, recordings, structured summaries) in standard formats (JSON, CSV, WAV/MP3)
• Deletion: After export, you can request deletion of all data from ODW.ai servers (if using SaaS)
• Sovereign deployments: If you're using private cloud or on-premises, the data is already on your infrastructure — just keep it or delete it as you wish

No vendor lock-in. Your data is portable.

It depends on your deployment:

• SaaS (ODW.ai-hosted): Yes, ODW.ai has access to call data (but we don't look at it unless you ask us to for support). We're bound by our privacy policy and any BAA.
• Private cloud or on-premises: No. ODW.ai has zero access. The data is on your infrastructure, and we can't reach it.

For maximum privacy, choose private cloud or on-premises deployment.

Voice is compatible with GDPR and other privacy laws when deployed correctly:

• Data sovereignty: You control where call data lives (your infrastructure, your jurisdiction)
• Consent: Configurable consent prompts for call recording
• Data export: Export call data for subject access requests
• Data deletion: Delete call data on request (right to erasure)
• Retention policies: Configure how long data is kept
• Audit logs: Every access and action is logged

Voice gives you the tools for compliance, but you're responsible for configuring them correctly. Consult your legal team for specific compliance advice.

Voice follows industry best practices for security:

• Encryption: AES-256 at rest, TLS 1.3 in transit
• Access controls: Role-based access control (RBAC), multi-factor authentication (MFA)
• Audit logs: Tamper-evident audit trail
• Penetration testing: Regular third-party security audits (planned for v1.1)

We're pursuing SOC 2 Type II certification (expected late 2026). For now, we provide a security whitepaper and can answer specific security questions from your IT team.

Voice v1.0 supports:

• English: US, UK, Australian accents
• Spanish: Latin American and Castilian
• French: European and Canadian
• German: Standard German

Additional languages can be added by configuring the AI model and TTS engine. Most frontier models (GPT-4, Claude) support 50+ languages.

Not in v1.0. In v1.0, Voice uses a single language per call. The language is set at the start of the call (either auto-detected or configured).

Mid-call language switching is planned for v1.2. This will allow Voice to handle callers who switch between languages (e.g., Spanglish, Franglais).

Voice is designed to be inclusive, but has limitations:

• Speech impairments: The speech recognition model is trained on diverse speech patterns, including some speech impairments. However, severe speech impairments may reduce accuracy.
• Hearing impairments: Voice is a phone-based system, so it's not suitable for callers who are deaf or hard of hearing. We recommend offering a text-based alternative (SMS, chat, email) for these callers.

If Voice can't understand a caller after multiple attempts, it offers to transfer to a human or take a message via text (SMS).

Yes. Voice understands natural language commands:

• "Go back" / "Start over" → Voice restarts the current flow
• "Talk to a human" / "Transfer me" → Voice transfers to a human agent
• "Cancel" → Voice cancels the current action (e.g., booking)
• "Repeat that" → Voice repeats the last message

These aren't rigid voice commands — Voice understands the intent, even if the caller phrases it differently.

Not in v1.0. TTY/TDD (teletypewriter) support is not available in v1.0.

This is a known limitation. We recommend offering a text-based alternative (SMS, chat, email) for deaf or hard-of-hearing callers. TTY/TDD support is being evaluated for a future release.

Yes. You can configure Voice to:

• Speak more slowly: Adjust the TTS speaking rate
• Pause between items: When reading a list (e.g., available appointment times), pause longer between options
• Repeat on request: If the caller says "Can you repeat that?" or "What was the first option?", Voice repeats the information

This is especially useful for elderly callers or callers who need extra time to process information.

Voice offers three deployment options:

1. SaaS (ODW.ai-hosted): We host everything. You just configure via the admin console. Fastest to get started (minutes).
2. Private cloud: Deploy in your AWS/GCP/Azure VPC. We provide Helm charts and deployment guides. You manage the infrastructure.
3. On-premises: Deploy in your data center. Full sovereignty, even air-gapped. Requires DevOps support.

For most customers, SaaS is the fastest path. For regulated industries (healthcare, legal), we recommend private cloud or on-premises.

Setup time depends on your deployment:

• SaaS: 30 minutes to 2 hours. Connect your phone number, configure your greeting, upload your knowledge base, test, and go live.
• Private cloud: 1-2 days. Deploy the infrastructure, configure networking, test, and go live.
• On-premises: 2-5 days. More complex due to air-gapping, custom networking, etc.

Most customers are live within a week. Complex deployments (multiple phone numbers, custom integrations) may take 2-3 weeks.

For SaaS: No. The admin console is designed for non-technical users (office managers, practice managers). You can configure greetings, business hours, routing rules, and more without writing code.

For private cloud or on-premises: Yes. You need someone who can manage Docker/Kubernetes, configure networking, and monitor infrastructure. This is typically a DevOps engineer or IT consultant.

Day-to-day operation (reviewing calls, updating knowledge base, adjusting settings) is non-technical for all deployment modes.

Yes. You can port your existing phone number to Voice. The process takes 1-2 weeks (carrier-dependent).

Alternatively, you can get a new number through our carrier partners (Twilio, Telnyx, Vonage) and use it alongside your existing number.

You can also use your own SIP trunk if you have one. Voice is telephony-agnostic.

Voice works with any SIP-based telephony provider:

• Partner carriers: Twilio, Telnyx, Vonage (we have pre-built integrations)
• Other carriers: Any carrier that provides SIP trunks (e.g., Bandwidth, Flowroute, local carriers)
• On-premises PBX: If you have an on-premises phone system (Cisco, Avaya, etc.), Voice can integrate via SIP trunking

Voice doesn't require you to switch carriers. It works with your existing telephony infrastructure.

Voice is configured via the admin console (web-based UI):

• Greeting: Custom greeting text, business hours, after-hours behavior
• Knowledge base: Connect to Vault, upload documents, configure which documents Voice can access
• Routing rules: Configure when to transfer to a human, which numbers/departments to transfer to
• Model selection: Choose local, private cloud, or frontier AI models
• Voice selection: Choose the AI voice (male/female, accent, tone)
• Compliance: Configure consent prompts, recording toggles, data retention policies

No code required. All configuration is done through the UI.

Yes. Voice includes a preview mode:

• Test calls: Call a test number to hear Voice in action. Try different scenarios (booking, Q&A, transfer).
• Shadow mode: Voice processes real calls but doesn't respond — instead, it generates draft responses for human review. This lets you evaluate Voice performance before enabling auto-response.
• Sandbox: Simulate calls in a sandbox environment. No real calls are made.

We recommend testing thoroughly before going live. Most customers spend 1-2 days testing.

For SaaS: Updates are automatic. We deploy new versions without downtime (rolling updates).

For private cloud or on-premises: Updates are performed via Docker/Kubernetes:

1. Pull the new Docker image
2. Restart the containers (rolling update for zero downtime)
3. Database migrations run automatically

Total update time: <5 minutes (plus downtime during restart for non-rolling updates).

Voice pricing is based on call volume:

Additional costs:

• Phone number: $1-$5/month per number (carrier-dependent)
• Call minutes: $0.01-$0.02 per minute (carrier-dependent, passed through at cost)
• Frontier AI models: If you use OpenAI/Anthropic, you pay per token (~$0.01-$0.03 per call)

Yes, significantly. A full-time receptionist costs $30K-$50K+/year (salary + benefits + overhead). Voice costs $1,200-$3,600/year (Professional tier) — 10-40x cheaper.

And Voice works 24/7/365 — no sick days, no lunch breaks, no holidays off.

Of course, Voice doesn't replace a receptionist entirely. It handles routine calls (80-85%), freeing your human staff for complex, high-value interactions. But the cost savings are substantial.

Yes. Traditional call centers and answering services charge $0.50-$2.00 per minute. For a business receiving 1,000 calls/month averaging 3 minutes each, that's $1,500-$6,000/month.

Voice charges a flat monthly fee ($99-$299) plus carrier costs (~$0.01-$0.02 per minute). For the same 1,000 calls, Voice costs ~$300-$600/month — 3-10x cheaper.

And Voice provides a better caller experience (natural conversation vs. "Press 1 for sales") and deeper integration with your calendar/CRM.

No hidden costs. The total cost of ownership includes:

• Voice subscription: Monthly fee based on call volume
• Phone number: $1-$5/month per number
• Call minutes: $0.01-$0.02 per minute (carrier costs, passed through at cost)
• Frontier AI models: Optional, if you use OpenAI/Anthropic

That's it. No setup fees, no per-user fees, no surprise charges.

Yes. We offer a 14-day free trial of the Professional tier. No credit card required.

During the trial, you can:

• Set up Voice with a test phone number
• Configure your greeting, knowledge base, and routing rules
• Test with simulated calls
• Go live with real calls (up to 2,000 calls/month)

At the end of the trial, you can subscribe or downgrade to the free tier (limited features).

We accept:

• Credit/debit cards: Visa, Mastercard, American Express (via Stripe)
• Bank transfer: For annual subscriptions or enterprise customers
• Cryptocurrency: Bitcoin, Ethereum, USDC (for privacy-conscious customers)

Invoicing is available for annual subscriptions. Contact us for details.

Yes. We offer a 30-day money-back guarantee. If you're not satisfied within the first 30 days, contact us for a full refund — no questions asked.

After 30 days, you can cancel anytime. You'll continue to have access until the end of your billing period, but no refund is provided for partial months.

Voice scales horizontally:

• SaaS: Unlimited concurrent calls (we handle scaling)
• Private cloud: Scales with your infrastructure. A single node can handle ~50 concurrent calls. Add more nodes for higher capacity.
• On-premises: Scales with your infrastructure. Typical deployments handle 50-200 concurrent calls.

For very high-volume deployments (500+ concurrent calls), we recommend Kubernetes with auto-scaling.

If Voice goes down, calls are handled by your telephony provider's failover:

• Voicemail: Calls go to voicemail (your existing voicemail system)
• Forwarding: Calls can be forwarded to a backup number (e.g., your mobile, a colleague's number)
• Queue: Some carriers offer call queuing — callers wait until Voice comes back online

For high-availability deployments, we recommend:

• SaaS: We provide 99.9% uptime SLA (Enterprise tier)
• Private cloud: Use multiple replicas across availability zones
• On-premises: Use failover clustering

Voice includes an analytics dashboard:

• Call volume: Total calls, answered calls, missed calls, transferred calls
• Resolution rate: % of calls resolved without human intervention
• Latency: Answer time, response latency (p50, p95)
• Top intents: What callers are asking for (booking, Q&A, transfer, etc.)
• Transfer reasons: Why calls are transferred to humans

You can also export data to external analytics tools (Grafana, Datadog) via Prometheus metrics or API.

Voice uses Vault for knowledge. To update the knowledge base:

1. Upload new documents to Vault (or update existing ones)
2. Vault processes the documents automatically (indexing, chunking, embedding)
3. Voice immediately uses the updated knowledge — no restart required

Updates are incremental — only changed documents are re-processed. This is fast and efficient.

Yes. Voice supports multi-tenant deployment:

• Multiple locations: Different phone numbers, greetings, and knowledge bases for each location
• Multiple businesses: If you're an agency or MSP, you can run Voice for multiple clients from a single deployment

Each tenant is isolated — separate data, separate configuration, separate billing.

Voice is designed to be intuitive, but we provide training resources:

• Documentation: Comprehensive user guides for admins
• Video tutorials: Step-by-step walkthroughs of key features
• Onboarding session: For paid customers, we offer a 1-hour onboarding session (remote)
• Support: Priority support for Professional and Enterprise customers

Most teams are productive within 1-2 hours of training. The admin console is particularly intuitive — no technical skills required.

Yes. Voice is highly configurable:

• Greeting: Custom greeting text, business hours, after-hours behavior
• Routing: Configure when to transfer to a human, which numbers/departments to transfer to
• Behavior: Configure what Voice should and shouldn't do (e.g., "Always offer to book," "Never discuss pricing")
• Voice: Choose the AI voice (male/female, accent, tone)
• Compliance: Configure consent prompts, recording toggles, data retention

All configuration is done via the admin console — no code required.

Voice integrates with your calendar via Loop workflows (ODW.ai's automation engine):

• Google Calendar: Native integration
• Microsoft Outlook: Native integration
• Cal.com: Native integration
• Other calendars: Any calendar with CalDAV or API support (custom integration via Loop)

Voice checks availability, proposes slots, books appointments, and sends confirmations — all in real-time during the call.

Voice integrates with your CRM via Loop workflows:

• Salesforce: Native integration (planned for v1.1)
• HubSpot: Native integration (planned for v1.1)
• Other CRMs: Any CRM with an API (custom integration via Loop)

After every call, Voice can:

• Create or update a contact/lead
• Log the call transcript and summary
• Trigger follow-up tasks (e.g., "Call back in 3 days")

Yes. Voice can send notifications via Loop workflows:

• SMS: Send appointment confirmations, reminders, or follow-ups
• Email: Send detailed summaries, receipts, or documents
• Slack: Notify your team when a call is received, booked, or transferred
• Webhook: Trigger any external system (e.g., create a ticket in your helpdesk)

All notifications are configurable via the admin console.

Yes, via custom Loop workflows. Voice can:

• Create patient records in your EHR
• Update appointment schedules
• Log call transcripts (redacted for PHI)

EHR integrations require custom configuration (every EHR system is different). We provide integration guides and support for common EHR systems (Epic, Cerner, Athenahealth). Contact us for details.

Yes, via custom Loop workflows. Voice can:

• Create new matter/intake records
• Run conflict checks (if your system has an API)
• Log call transcripts (redacted for privileged information)
• Route urgent calls to the on-call attorney

Legal practice management integrations require custom configuration. We provide integration guides and support for common systems (Clio, MyCase, PracticePanther). Contact us for details.

Yes. Voice can trigger custom Loop workflows for any action:

• Create a ticket in your helpdesk (Zendesk, Freshdesk, etc.)
• Update a database record
• Send a webhook to any external system
• Trigger a Zapier/Make automation

Loop workflows are configured via the admin console (no code required for common actions). For complex workflows, you can write custom logic in Python or JavaScript.

Yes. Voice is part of the ODW.ai suite and integrates natively with:

• Vault: Knowledge base (answers questions from your documents)
• Loop: Workflows (book appointments, send notifications, trigger actions)
• Shield: PII detection and redaction (for compliance)
• Desk: Chat support (Voice can transfer callers to Desk for text-based follow-up)
• Auth: Single sign-on (shared authentication across all ODW.ai products)

If you already use other ODW.ai products, Voice works seamlessly with them — no duplicate configuration.

Voice is powerful, but it has limitations in v1.0:

• No outbound calling: Voice answers incoming calls but doesn't initiate outbound calls (callbacks, reminders). Planned for v1.1.
• No video calls: Voice is audio-only. No video component.
• No contact-center features: Voice is a receptionist, not a full contact-center platform (no ACD queues, agent desktop, workforce management).
• No custom voice cloning: Use preset voices only. Custom cloning planned for v1.2.
• No real-time translation: Language is set per-call. No live translation mid-call.
• No payment processing: Voice can't take credit card numbers over the phone (PCI scope). Planned for v2.0.
• No conferencing: Single caller + agent only. No multi-party calls.

Outbound calling is planned for v1.1 (Q4 2026). It will include:

• Callbacks: "We'll call you back at 3pm" → Voice calls the customer at 3pm
• Appointment reminders: Voice calls customers 24 hours before their appointment
• Follow-ups: Voice calls leads who requested information

Outbound calling requires careful compliance controls (TCPA in the US, GDPR in Europe) to avoid spam calls. We're building the guardrails first.

Planned for v1.2 (Q1 2027). Custom voice cloning will allow you to create a voice that sounds like a specific person (e.g., your company's spokesperson, a celebrity endorser).

This requires advanced AI models and careful ethical controls (to prevent misuse). We're working with leading TTS providers to offer this feature safely.

Planned for v1.2. Real-time translation will allow Voice to handle callers who switch between languages mid-call (e.g., Spanglish, Franglais).

This is technically challenging (requires real-time language detection and translation) but is a high-priority feature for multilingual markets.

Planned for v2.0 (2027). Payment processing over the phone requires PCI DSS compliance (Payment Card Industry Data Security Standard), which is a significant undertaking.

When implemented, Voice will be able to:

• Collect credit card numbers securely
• Process payments via integrated payment gateways (Stripe, PayPal, etc.)
• Send receipts via SMS/email

For now, if a caller needs to make a payment, Voice can transfer to a human or send a secure payment link via SMS.

Key milestones:

• v1.0 (Q3 2026): MVP launch. Inbound calls, appointment booking, Q&A, call routing, Vault integration, sovereign deployment.
• v1.1 (Q4 2026): Outbound calling (callbacks, reminders), multi-party warm transfers, native CRM integrations (Salesforce, HubSpot).
• v1.2 (Q1 2027): Custom voice cloning, real-time translation, sentiment detection and escalation triggers.
• v2.0 (Q2 2027): Payment processing (PCI-compliant), voice biometrics for caller authentication, proactive outbound campaigns.

Roadmap is subject to change based on customer feedback. Contact us for the latest roadmap or to request features.

Voice is not the right fit if:

• You need a full contact-center platform: Voice is a receptionist, not a CCaaS replacement. If you need ACD queues, agent desktop, workforce management for 100+ agents, use Genesys, Five9, or NICE.
• You need outbound marketing campaigns: Voice doesn't do high-volume outbound calling (telemarketing, robocalls). This is ethically and legally problematic.
• You need video calls: Voice is audio-only. No video component.
• You need payment processing: Voice can't take credit card numbers over the phone (yet). Planned for v2.0.
• You need TTY/TDD support: Voice doesn't support TTY/TDD for deaf callers. Offer a text-based alternative.

Three paths:

1. Free trial: Sign up for a 14-day free trial. No credit card required. Set up Voice with a test phone number and try it out.
2. Demo: Book a demo with our team. We'll walk you through Voice, answer your questions, and help you evaluate if it's right for your business.
3. Contact sales: For enterprise customers or custom deployments (private cloud, on-premises), contact our sales team for a tailored quote.