Zenople AI Receptionist FAQs

Zenople AI Receptionist Agent — Frequently Asked Questions (FAQ)

1. General Questions

Q: What is the Zenople AI Receptionist Agent?

The Zenople AI Receptionist Agent is an AI-powered virtual receptionist that handles inbound phone calls for staffing companies. It conducts natural, multi-turn voice conversations to understand caller intent, authenticate identities, answer questions, and perform actions — all without human intervention. Unlike traditional IVR systems with rigid menus, its conversations are entirely natural language-based.

Q: Who is the AI Receptionist designed for?

It is built for staffing organizations that need to handle inbound calls from applicants, employees, new hires, client contacts, and unknown callers across multiple offices and tenants.

Q: Is the AI Receptionist available 24/7?

Yes. The agent handles calls around the clock, including after business hours. During quiet hours (outside configured office availability), the agent can still take calls, answer questions, and escalate when needed.

Q: How is this different from a traditional IVR system?

Traditional IVR systems force callers through rigid, menu-driven flows ("Press 1 for…"). The AI Receptionist conducts a natural, open-ended conversation. Callers simply state what they need, and the agent understands intent, asks follow-up questions, and takes action autonomously.

Q: What technology powers the AI Receptionist?

• LLM: Azure OpenAI (GPT-4.1-mini) for conversational AI
• Speech-to-Text / Text-to-Speech: Azure Cognitive Services Speech SDK
• Telephony: Plivo (inbound voice calls, SMS, WebSocket streaming)
• Architecture: Clean Architecture, Domain-Driven Design
• Database: PostgreSQL (conversation logs, multi-tenant configuration)

Q: What is the current release status?

The agent is currently at v2.1.0 Beta. Features continue to evolve based on feedback. Review the Roadmap section for upcoming capabilities.

2. Setup & Configuration

Q: What do I need before I can start using the AI Receptionist?

Setup is completed in two stages:

1. Initial Setup Request (Aqore Team)
   • Provision inbound phone number and SMS capability.
   • Configure required backend services.
   • Enable AI Receptionist Insights dashboard/ configuration navigation.
2. Client Configuration (Self-Service)
   • Configure receptionist behavior in ATM > AI Agent > AI Receptionist.

Q: What does Aqore set up before client self-configuration?

Aqore handles environment readiness:

• Active inbound number (voice + SMS)
• Backend integration setup
• Dashboard/configuration navigation enablement
• Readiness confirmation to the client

Q: What does the client configure after Aqore confirms readiness?

Using AI Agent navigation under ATM application clients needs to configure:

• Agent Identity
• Office availability hours and routing
• Escalation settings and availability
• Automation/write-back permissions
• Outbound delivery mode

Q: Can I customize the greeting message?

Yes. The greeting message is configurable per scope.
Default template: "Hi, this is Jennifer Adams with #Company Name. How may I help you today?"

Q: How do I configure office hours?

In the configuration form, set each day to open/closed and provide From/To times for enabled days.
Default is Monday-Friday, 9 AM-5 PM.

Q: Can I set different escalation modes for different caller types?

Yes. Escalation mode, destination, and availability are configurable per caller type (Unknown, Applicant, Employee, New Hire, Contact).

Q: Where do I access the Configuration UI?

1. Open ATM.
2. Go to AI Agent under Settings.
3. Open the AI Receptionist tab.

Q: How do scope and inheritance work in configuration?

• Select Tenant / Company / Office from top filters.
• The badge shows active scope.
• Inherited scopes display Inherits from ....
• Save automatically sets config_level:
  • Tenant -> Tenant
  • Company -> Organization
  • Office -> Office

Q: What do Manage Access, Edit, Copy Config, and Active Toggle do?

• Manage Access: bulk enable/disable receptionist access for organizations/offices (only enabled after tenant onboarding process completion).
• Edit: opens configuration drawer for selected scope entity.
• Copy Config: copy current scope entity configuration to selected organizations/offices and optionally adjust before apply.
• Active Toggle:
  • Off -> disables receptionist for selected scope entity.
  • On -> enables directly if onboarded; for first tenant onboarding launches two steps (Access, Configuration).

Q: How do I save configuration and copy it to other targets?

• Click Edit button after selecting correct scope entity, make necessary change and Save Configuration to persist changes for the current scope.
• Use Copy Config to clone current settings to one or more organizations/offices.
• In Copy Config, you can still adjust values before clicking Apply Configuration.

Q: What default configuration values are applied on first setup?

Q: Does the agent support multi-tenant and multi-branch setups?

Yes. Configuration follows a hierarchical model: Tenant → Company → Office. Each level can override settings from the level above, allowing different greetings, escalation targets, and hours per branch. However, for unknown callers settings like identification, greetings and escalation will be default from tenant level.

Q: Do configuration changes require a restart?

No. Configuration changes are applied through the Configuration UI and do not require service restart.

Q: Is there any important setup note for escalation?

Yes. Unknown -> Custom destination number must be explicitly configured because default placeholder values may be invalid.

3. Caller Identification & Personas

Q: How does the agent identify who is calling?

When a call is received, the agent:

1. Captures the inbound Caller ID (ANI).
2. Queries Zenople for matching records using phone number normalization (E.164 format).
3. Attempts to classify the caller as Employee, Applicant, New Hire, Contact, or Unknown.

If no exact phone match is found, the agent infers the persona from conversation context (e.g., "I'm not registered yet" → Applicant).

Q: What are the five caller personas?

Q: What information is needed to identify each persona by default?

Q: What happens if the agent cannot identify the caller?

If no profile match is found, the caller is treated as Unknown. They can still search for jobs, ask general questions, register as a new applicant, or be escalated to a human agent.

Q: What happens when a caller's phone number isn't in the database?

The caller is treated as Unknown. The agent then tries to infer the caller's persona from the conversation itself and tries to identify their existing profile or registered as an new applicant or contact. For example:

• "I'm looking for work" → AI detects Applicant intent
• "I'm calling about one of our workers" → AI detects Contact intent

Q: How accurate is persona detection?

Persona detection is approximately 95% accurate in the Beta release. Rare misclassifications are possible when the caller provides ambiguous intent. The AI includes a confidence level (high, medium, low) when detecting Unknown caller personas — low-confidence detections are handled conservatively.

4. Authentication & Security

Q: When is OTP authentication required?

OTP (One-Time Password) authentication is required for sensitive operations such as:

• Updating profile information
• Requesting login credentials
• Viewing payroll/paystub details
• Looking up assignments, schedules, and timesheets
• informing absences

Q: How does OTP work?

1. The agent sends a 6-digit OTP via SMS to the caller's registered phone number.
2. The code is valid for 5 minutes and is single-use.
3. The caller reads the code back to the agent.
4. Upon successful verification, the caller moves to the AUTHENTICATED state.

Rate limiting is enforced independently — see "What happens if the caller fails OTP verification?" below.

Q: What happens if the caller fails OTP verification?

The system enforces a maximum retry limit (default: 3 attempts). After exceeding the limit, the call enters a FAILED state and is escalated to a human agent.

Q: Is sensitive data spoken back to the caller?

No. Sensitive data (password etc.) is never spoken back verbatim. Information details like password can be sent via SMS rather than read aloud.

Q: How is session security maintained?

• Sessions are managed with asyncio.Lock for thread-safe concurrent access.
• Phone numbers are normalized to E.164 format to prevent spoofing mismatches.
• Partial identity matches are not accepted — exact match is required.
• Error messages are sanitized to prevent internal detail leakage.
• OTP is forced for sensitive infomation.
• Identification can be customized per persona.

5. Call Handling & Conversation Flow

Q: What is the state machine and how does it work?

The agent uses a state-based conversation flow with five primary states:

Q: Can the caller interrupt the AI mid-sentence?

Yes. The system supports barge-in detection — if the caller speaks while the AI is responding, the AI stops and processes the new input.

Q: What happens if an error occurs mid-call?

The agent has safe fallback handling. If a critical error occurs during processing, the call is gracefully escalated to a human agent rather than being dropped. Error states trigger the FAILED state with appropriate messaging.

Q: Can a caller switch language mid-call?

Yes. The agent supports real-time language switching during the call. If the caller requests or talk Spanish (or English), the agent:

1. Confirms the switch in current language.
2. Responds immediately in the new language for the rest of the call.

Currently supported: English and Spanish. Additional languages are on the roadmap.

6. Features & Capabilities

Q: What actions can each persona perform?

* Requires OTP verification

Q: What profile information can callers update over the phone?

Callers can update the following profile fields (all require OTP authentication):

The agent guides the caller through providing the required information and confirms before submitting.

Q: Can callers search for jobs over the phone?

Yes. The agent can search available jobs posted in zenople job portal by title, location, and distance. Results are presented verbally (summary of top matches) and the full list is sent via SMS with clickable links.

Q: Can callers apply for a job during the call?

Yes. Applicants, New Hires and Employees can apply for jobs directly during the call. If screening questions are required (Decided by agent itself), the agent conducts them automatically.

Q: Can the agent book appointments?

Yes. Applicants, New Hires, and Employees can request and schedule appointments. The agent log appointment request in identified callers profile comment. An SMS notification to office staff is sent through the escalation number.

Q: Can employees report absences?

Yes. Identified employees can report absences/no-shows. The agent logs the absence and sends an SMS notification to the configured escalation number with employee name, assignment, date, and reason.

Q: Can contacts (clients) report employee no-shows?

Yes. Client contacts can log employee call-offs (no-show, late, sick). The agent captures details (employee name, job site, reason) and sends an notification to the configured escalation number for contact caller type.

7. Escalation & Routing

Q: When does the agent escalate to a human?

The agent escalates in the following situations:

• The caller explicitly asks to speak with a human multiple time even if agent tries to help.
• OTP verification fails after maximum attempts
• An emergency or urgent issue is detected
• Complex billing or payroll disputes that exceed the agent's scope
• A critical error occurs during processing
• The caller's request falls outside the agent's available tools

Q: What escalation methods are supported?

Q: What happens if the escalation target is unavailable?

If escalation is configured as a live call and the human agent is outside availability hours, the system falls back to SMS escalation, depending on configuration.

Q: Can we use the same escalation number for all persona types?

Yes. The escalation number can be the same for all persona types or different for each, based on the client's preference. This is fully configurable during setup.

8. SMS & Notifications

Q: What types of SMS does the agent send?

Q: Are SMS messages sent during the call or after?

Every SMS messages are sent during the call in real time. For example, job search results and OTP codes are sent while the caller is on the line so they can verify or review immediately.

Q: Who receives escalation SMS notifications?

Escalation SMS notifications go to the configured escalation phone number for the relevant persona type. Different personas can have different escalation numbers.

9. Logging, Recording & Compliance

Q: What data is captured after each call?

After every call, the system automatically stores:

Q: What are the possible call outcome classifications?

Q: Are call recordings available for playback?

Yes, call recordings are stored in secure storage and available for playback in the Zenople UI comment navigation for 90 days however call transcript is always available.

Q: Does the agent comply with call recording consent laws?

The agent stores recordings, but it is the client's responsibility to ensure compliance with state and local call recording consent laws. Zenople recommends configuring a consent disclosure in the initial greeting message where required.

Q: How long is the call recording available?

In the Beta release, call recordings are available for 90 days from the call date. If needed beyond that period, recordings can be downloaded from the Zenople comment navigation for long-term retention.

Q: How long is call data retained?

Data retention is configurable per regulatory requirements. All call information except call recording like call transcript, summary, outcome, sentiment, AI actions, writeback information is stored in the client data.

10. Language & Voice

Q: What languages are supported?

Additional languages (French, Portuguese) are planned for post-GA releases.

Q: Can I choose the AI voice?

Yes. Voice settings are testable and configurable with speech rate and pitch fron the agent configuration UI.

Q: Does background noise affect call quality?

Yes. The following can impact speech recognition accuracy:

• Noisy environments (traffic, crowds)
• Poor phone connections
• Speakerphone mode (reduces barge-in sensitivity)

For best results, callers should use a clear audio environment.

11. Performance & Scalability

Q: How many concurrent calls can the system handle?

• 50+ concurrent calls are supported out of the box.
• 100+ concurrent calls are achievable with infrastructure scaling.

Q: How does the system maintain low latency?

• Prompt caching reduces AI response latency and cuts token costs by ~50%.
• Server-Sent Events (SSE) enable real-time streaming of AI responses.
• Async HTTP client (httpx) with connection pooling ensures non-blocking API calls.
• Realtime AI models reduce AI response latency.

12. Insights Dashboard

Q: Where can I access AI Receptionist insights?

Navigate to Dynamic Dashboard > Agent Insights and open AI Receptionist Insight tab.

Q: What business value can I track from the dashboard?

The Executive Summary tiles show measurable value and operations outcomes, including:

• Estimated Gross Value Created
• Total Calls and Hours Reclaimed
• After-Hours Calls Resolution and After-Hours Reclaimed Hours
• Identification, Resolution, Escalation, and Follow-up Required rates
• Zenople Action Taken and Multichannel Outbound

Q: Which charts are included in the dashboard?

• Growth Comparison: Volume & Productivity (Period vs Previous)
• Caller Distribution
• Multilingual Engagement
• Call Intent Distribution
• Sentiment Analysis
• Outcome Distribution
• Outcome by Caller Type
• Zenople Automation Write-Backs
• Escalation Reason Breakdown
• Identification Funnel

Q: How are the main chart values calculated?

• Most distribution charts use count / total * 100 for percentages in the selected date range.
• Caller Distribution groups calls by normalized caller type.
• Multilingual Engagement keeps top languages and groups the remainder into Other.
• Call Intent Distribution counts intent occurrences from intent arrays (a single call may contribute to multiple intents).
• Growth Comparison compares current-period totals against equal-length previous period and shows delta/delta%.

Q: How do drilldowns and View Call Insights work?

• Clicking a KPI tile opens a metric-specific drilldown table.
• Conversation Explorer provides call-level records (date/time, caller type, contact, category, duration, outcome, actions).
• Clicking View Call Insights opens full call detail (summary, transcript, recording when available).

Q: Does the dashboard use aggregate data or full call history?

Dashboard visuals/KPIs are driven by aggregated metrics for performance and consistency.
Detailed call-level investigation is provided through drilldown tables and View Call Insights.

Q: How are after-hours metrics determined?

After-hours metrics use the configured default tenant level office availability hours. Calls outside configured hours (including closed days) are counted as after-hours.

13. Known Limitations (Beta)

Q: What features are not available in the Beta?

The following are not included in the Beta release:

• Email convergence (outbound email handling)
• Multi-agent specialist routing deployment (infrastructure ready, not yet live)
• Proactive outbound calls
• Video call or group call support

Q: Will features change during Beta?

Yes. Features, conversation flows, and configurations will continue to evolve based on feedback and development priorities. Refer to the Roadmap section for planned improvements.

14. Troubleshooting

Q: The caller was not identified correctly. What should I do?

• Verify that the caller's phone number is registered correctly in Zenople (E.164 format).
• Check the identification method configuration for the relevant persona type.
• Review the call transcript in the conversation logs to see what information was provided.

Q: OTP is not being received by the caller. What could be wrong?

• Confirm the caller's registered phone number is correct and can receive SMS.
• Verify that the Plivo phone number is configured for SMS delivery.
• Check whether the send rate limit has been hit — max 3 OTP sends per 10 minutes per number. If so, the caller must wait before requesting a new code.
• Ensure the caller is not blocked by carrier-level SMS filtering.
• You can ask if agent to resend.

Q: The agent escalated when it shouldn't have. Why?

• Review the call transcript to understand the intent classification.
• The agent escalates when it encounters requests outside its available tools for the current state/persona.
• Check if the confidence threshold for escalation is set too low (default: 0.85).

Q: Call audio quality is poor. What can I check?

• Ensure stable network connectivity for the telephony infrastructure.
• Check provider call quality metrics for packet loss or jitter.
• Advise callers to avoid speakerphone mode and noisy environments.

Q: Who should I contact for Beta support?

• Email: aisupport@aqore.com

Monthly Beta review calls are also scheduled with your customer success manager.

15. Billing & Pricing

Q: Is there a cost for using the Beta?

Its free for beta participants.

Q: What drives cost?

Costs are driven by:

• Call Duration — proportional to conversation duration.
• Telephony usage — Plivo call and SMS charges.
• AI services — AI resources consumption.
• AI Infrastructure - Service scaling.

16. Roadmap & Future Features

Q: What is planned for the GA release?

Need more help? Contact your Zenople customer success manager or reach out to aisupport@aqore.com.