Zenople AR Agent - FAQ

Zenople AR Agent — Frequently Asked Questions

Comprehensive answers to common questions about the AI-powered accounts receivable collections agent — covering capabilities, configuration, voice & email channels, hold states, security, and more.

Contents

1. Getting Started (7 questions)

2. How It Works — General (12 questions)

3. Voice Calls (19 questions)

4. Email (10 questions)

5. Customer Intents — What the Agent Can Handle (19 questions)

6. Hold States & Statuses (9 questions)

7. Configuration & Policies (16 questions)

8. Monitoring, Transcripts & Audit (5 questions)

9. Security, Privacy & Compliance (6 questions)

10. Scope, Limitations & Roadmap (8 questions)

11. Troubleshooting (9 questions)

12. Email Profile Setup (7 questions)

13. Initial Setup & Go-Live (7 questions)

1. Getting Started

The basics — what the AR Agent is and how to get going.

Q1. What is the Zenople AR Agent?

The Zenople AR Agent is an AI-powered collections assistant that automates routine accounts receivable outreach while escalating situations that require staff review. It can make natural voice calls and send intelligent email campaigns to customers with overdue invoices, detect customer intents, and create audit trails of interactions.

Q2. Who is "Ava"?

"Ava" is the default name of the AI collections specialist. When the AR Agent makes a voice call, it introduces itself as "Ava" from your company — speaking naturally, understanding context and nuance, and adapting its tone to the conversation. It is not a scripted robocall system.

The agent's name is configurable. In your company-level AR Agent Policy, the Agent Name field controls the name the agent uses during voice call introductions and email correspondence. If you change it to "Sophie," for example, the agent will introduce itself as "Sophie" instead. The default value is "Ava."

Q3. What does the AR Agent actually do?

The AR Agent automates the routine parts of collections:

• Identifies customers with overdue invoices that meet your configured thresholds.

• Contacts them via voice call or email (based on your preferred channel setting).

• Conducts real-time conversations — handling promises to pay, disputes, callback requests, invoice copy requests, and more.

• Records every interaction with full transcripts and AI-generated summaries.

• Escalates to your AR staff when human judgment is needed (disputes, payment plans, "do not call" requests).

Q4. Does the AR Agent replace my AR team?

No. The AR Agent handles routine collections outreach — the high-volume, repetitive follow-ups that consume most of your team's time. Your AR staff still handles disputes that require human judgment, complex payment negotiations, and escalated cases. Think of the agent as a force multiplier: it reduces collection workload by 60–80%, freeing your team to focus on high-value work.

Q5. How do I enable the AR Agent for my company?

In Zenople, switch to ATM mode, open your company detail, navigate to AI Agents → AR Agent Policy, and turn on the Enable AR Agent toggle. You'll also need to configure the collection workflow settings, communication channel, outbound email/phone, and quiet hours. See the Setup & Configuration Guide for step-by-step instructions with screenshots.

Q6. What do I need before I can start using the AR Agent?

• Zenople Admin Access — administrator privileges to configure policies.

• Email Account — a Microsoft 365 (Outlook) or Gmail account dedicated to AR outreach, connected via OAuth.

• Plivo Phone Number — provisioned by Zenople during onboarding for outbound voice calls.

• Customer Data in CMS — customers with valid contact information (email and/or phone) and open invoices.

Q7. How long does it take to set up?

Initial setup typically takes 15–30 minutes: connecting the email profile (5 minutes), configuring the company-level AR policy (10 minutes), and optionally setting office or customer-level overrides. Once configured, the AR Agent begins evaluating eligible customers automatically during its next polling cycle (approximately every 20 minutes).

2. How It Works — General

Understand the AR Agent's decision-making and operational behavior.

Q8. How does the AR Agent decide which customers to contact?

The agent evaluates customers against your configured policy thresholds. A customer becomes eligible for contact when all of the following are true:

• The AR Agent is enabled for that customer (directly or via opt-out model).

• At least one invoice is overdue beyond the Days Past Due threshold (e.g., 7+ days).

• The overdue invoice balance meets or exceeds the Minimum Invoice Balance (e.g., $1,000).

• The customer is not on hold (no active hold state).

• Contact attempts are below the configured limits — Max Emails Total and Max Voice Calls Total are tracked separately per channel.

• It is not currently within the customer's quiet hours.

Q9. How often does the AR Agent check for eligible customers?

The AR Agent's system conductor uses a configured polling interval (20 minutes in the current deployment) to evaluate eligible customers against policy thresholds and initiate or resume workflows. Treat this as a typical/default interval, not a response-time guarantee.

Q10. What happens when the AR Agent contacts a customer?

The agent follows a four-step process:

1. Retrieve Context — Pulls the customer's invoice history, prior conversations, pending promises, dispute details, and configured policies.

2. Initiate Contact — Reaches the customer via the configured channel (voice call or email).

3. Conduct Conversation — Engages in a natural, multi-turn conversation. Detects customer intents in real time (promise to pay, dispute, callback request, etc.) and takes appropriate action.

4. Record & Follow Up — Logs the full transcript, generates an AI summary, updates the customer's state, and schedules any follow-up actions (callbacks, reminders, escalation notifications).

Q11. What is the difference between Email Only, Voice Only, and Email Then Voice?

Q12. Can I use both email and voice for the same customer?

Yes. Set the Preferred Channel to Email Then Voice at the customer or policy level. The agent will start with email reminders and automatically escalate to voice calls if the customer doesn't respond to emails.

Q13. What happens if a customer doesn't respond at all?

The agent continues attempting contact according to your settings — respecting separate per-channel daily and total limits (Max Emails Per Day, Max Emails Total, Max Voice Calls Per Day, Max Voice Calls Total). Once the total attempt limit for the active channel(s) is reached, the customer is placed on a MaxAttemptsReached hold and your Internal AR Staff is notified via escalation. The agent will not contact that customer again until you manually clear the hold.

Q14. Does the AR Agent contact customers on weekends or holidays?

The agent respects quiet hours as configured in your policy (e.g., no contact between 6:00 PM and 8:00 AM). Beyond quiet hours, the agent operates 24/7, including weekends. There is no built-in holiday calendar at this time — if you want to prevent weekend contact, set quiet hours accordingly or adjust the Agent Availability window at the company level.

Q15. What are quiet hours and how do they work?

Quiet hours define a no-contact window during which the AR Agent will not call or email customers. For example, if you set Quiet Hours Start to 6:00 PM and Quiet Hours End to 8:00 AM, no outreach occurs during that window. Quiet hours are calculated using the Timezone configured in the policy. You can set different quiet hours at the company, office, or customer level.

Q16. What is the "Days Past Due" threshold and how does it work?

The Days Past Due Before Collection setting defines how many days after an invoice's due date the agent waits before initiating first contact. For example, if set to 7, the agent will not contact a customer until their invoice is at least 7 days overdue. This gives customers a grace period before automated collection begins.

Q17. Can I temporarily pause the AR Agent without fully disabling it?

There is no dedicated "pause" mode. To temporarily stop outreach, disable the Enable AR Agent toggle at the company level (or the specific office/customer level). When you're ready to resume, toggle it back on. The agent will re-evaluate eligible customers on the next polling cycle (within 20 minutes). Disabling does not lose your configuration — all settings are preserved.

Q18. How many customers can the AR Agent handle simultaneously?

Capacity and throughput depend on deployment configuration, provider limits, and workload. Eligible customers are evaluated in parallel where supported, with each customer's collection workflow independently orchestrated; confirm sizing expectations with Zenople for your tenant.

Q19. What happens if I change policy settings while collection workflows are already active?

Configuration changes take effect on the next polling cycle (within 20 minutes). In-progress voice calls or email sequences are not interrupted mid-execution. For example, if you change the Preferred Channel from Voice Only to Email Only while a call is in progress, that call will complete normally. The new setting applies the next time the agent evaluates that customer.

3. Voice Calls

Everything about the AR Agent's voice call experience.

Q20. What does an AR Agent voice call sound like?

The agent (Ava) speaks naturally with a professional, conversational tone. She introduces herself by name and company, explains why she's calling, and engages in a real two-way conversation. The voice is generated by advanced AI speech synthesis — there are no pre-recorded scripts or rigid menus. Customers can speak normally and the agent responds contextually.

Q21. Can the customer tell they are speaking with AI?

The voice quality is very natural, but some customers may recognize the AI. If a customer asks "Am I talking to a robot?" or "Are you AI?", Ava responds honestly. Transparency is a design principle — the agent does not attempt to deceive customers about its nature.

Q22. What if the customer asks to speak to a real person?

The agent immediately recognizes the Transfer to Agent intent. It records the reason and evaluates the configured destination, staff availability, and whether the customer has been offered and confirmed a live handover:

• Live offer / confirmed — If live handover is possible, the agent informs the customer it is connecting them to staff. The agent generates a brief summary for the staff member, who hears the summary and confirms readiness before being bridged with the customer. The transferred call is recorded separately from the AI phase.

• Follow-up only — If the customer's request is noted but live staff are not available or the request is a soft escalation, the agent records the escalation and the effort is routed to your Internal AR Staff for manual follow-up. No automatic time guarantee is implied.

In all cases the customer's request and the conversation context are recorded and your Internal AR Staff is notified.

Q23. What if the wrong person answers the phone?

The agent detects the Wrong Person intent — for example, "she doesn't work here anymore" or "I'm not Maria." It asks for the correct contact's name and information. If the person provides it, the agent updates the customer record and routes future calls to the corrected contact. If no information is provided, the agent sets a WrongPerson hold to prevent further calls to the wrong number and notifies your Internal AR Staff.

Q24. What happens if the call goes to voicemail?

When the AR Agent detects voicemail, it leaves an appropriate message asking the customer to call back at your office number. The voicemail message is professional and contextual — it identifies the agent by name and company, and provides the callback number. The call attempt is counted, and the agent will retry at the next available window if the customer does not return the call.

Q25. What happens if the call reaches an IVR (automated phone menu)?

The behavior depends on whether the customer's Extension field is populated in their Customer-Level AR Agent Policy:

• Extension is configured: The agent directly navigates the IVR by entering the extension, reaching the correct person without manual intervention.

• No extension configured: The agent cannot navigate the IVR menu and will hang up the call. The attempt is counted, and the agent will retry at the next available window.

If your customers commonly use IVR systems, ensure the Ext. field in the Client AR Contact section is populated with the correct extension number.

Q26. Does the AR Agent record calls?

Yes. Every voice call is fully transcribed in real time. The complete conversation transcript — with timestamps, speaker identification (Agent vs. Customer), and an AI-generated summary — is stored as a Comment record on the customer's profile in Zenople.

Q27. Where can I find call transcripts?

Navigate to the customer in CMS, then click the Comment tab in the customer navigation. Call transcripts appear as "Voice Call Note" entries with the date and time. Each entry includes:

• AI-generated summary with key action items (bullet points).

• Full color-coded transcript — Agent lines and Customer lines are visually distinct.

• Timestamps for each line of dialogue.

Q28. What if there is background noise during a call?

Background noise can impact speech recognition accuracy. The agent uses advanced AI speech-to-text models that handle moderate noise well, but very loud environments (construction, traffic, crowds) may cause missed words or intent misinterpretation. If the agent can't understand the customer, it will ask them to repeat. If communication remains difficult, the agent may suggest calling back at a better time.

Q29. Can the AR Agent handle customers with heavy accents?

The speech recognition models are trained on diverse accents and dialects, but heavy accents or non-standard pronunciation may occasionally cause missed intents. If the agent is unable to reliably interpret a customer, it will escalate to your Internal AR Staff rather than take incorrect action. Accuracy continues to improve as AI speech models are updated.

Q30. How fast does the AR Agent respond during a voice call?

The agent is designed for low-latency conversational flow. Response times vary by configured audio mode (Standard, Realtime, or Gemini Live), provider availability, network conditions, and AI workload. Brief pauses can occur; sub-second turnaround is not guaranteed.

Q31. What happens if the customer interrupts the agent mid-sentence?

The agent supports natural conversation flow, including interruptions. However, handling rapid or frequent interruptions is a known limitation in the current release — the agent may occasionally continue its sentence briefly before yielding. In most cases, the agent will stop, listen, and respond to the customer's interruption.

Q32. Are voice calls audio-recorded, and how long are recordings available?

Voice interactions may produce transcripts, summaries, and recordings. Availability and retention depend on the configured storage, provider, and applicable retention policies; consult your Zenople configuration or agreement for the applicable retention period.

Q33. What happens if the call is disconnected or the customer hangs up unexpectedly?

The agent evaluates the tone and context of the conversation before the disconnection:

• Angry or frustrated customer: If the customer appeared upset, hostile, or frustrated before the call dropped, the agent treats this as an intentional hangup. The case is escalated to your Internal AR Staff with the conversation transcript and a note about the customer's demeanor.

• Accidental disconnect: If the conversation was proceeding normally (no signs of frustration), the agent treats the disconnect as accidental and calls the customer back automatically to resume the conversation.

In both cases, the partial transcript is saved and the attempt is counted.

Q34. What caller ID does the customer see on outbound calls?

The customer sees the Outbound Phone Number (Plivo number) configured in your company-level AR Agent Policy. This is the same number for all customers under that company. If you need a different caller ID for different regions, contact Zenople support about provisioning additional Plivo numbers.

Q35. What happens if a customer calls back the AR Agent's number?

The behavior depends on your Incoming Call Forward To setting in the company-level AR Agent Policy:

• If configured: The inbound call is forwarded to the phone number you specified (e.g., your main office line or a specific AR staff member's direct line).

• If blank: Inbound calls to the Plivo number are not forwarded and the call is not answered by the AI agent.

The AR Agent handles outbound calls only — it does not conduct inbound AI conversations. The Incoming Call Forward To field ensures customers who call back reach a real person.

Q36. What happens if the customer is angry, hostile, or uses profanity?

The agent is trained to de-escalate professionally. It acknowledges the customer's frustration, maintains a calm and empathetic tone, and attempts to redirect the conversation productively. If hostility continues or the customer becomes verbally abusive, the agent ends the call respectfully, generates a conversation summary noting the customer's demeanor, and escalates to your Internal AR Staff for manual handling. A courtesy backoff period is applied before the next automated contact attempt.

Q37. How long does a typical AR Agent voice call last?

Call duration varies depending on the complexity of the conversation. Simple calls (single invoice, straightforward promise to pay) may last 2–3 minutes. More complex calls involving multiple invoices, disputes, or payment plan discussions can take 5–10 minutes. There is no hard time limit on calls in the current release — the agent stays on the line as long as the conversation requires.

Q38. Does the AR Agent call mobile phones and landlines?

Yes. The agent dials whatever phone number is entered in the Client AR Contact field at the customer level. It works with mobile phones, landlines, and direct-dial office numbers. The agent does not distinguish between phone types — it handles all calls the same way.

4. Email

How the AR Agent handles email-based collection outreach.

Q39. What emails does the AR Agent send?

The agent sends a tiered sequence of professional collection reminder emails. Each email is contextual — it references the specific overdue invoices, amounts, and due dates for that customer. Emails are sent from your configured Outbound Email Address (set up in Section 2 of the Setup Guide), so they appear to come from your company, not from Zenople.

Q40. What does the email reminder sequence look like?

The agent sends a tiered escalation sequence:

1. Initial Reminder — Friendly notice about overdue invoices with amounts and due dates.

2. Follow-Up — Second reminder noting the invoices are still outstanding.

3. Urgent Notice — Stronger tone emphasizing the overdue status.

4. Final Notice — Last automated attempt before escalation to your AR staff.

The tone escalates progressively while remaining professional and compliant.

Q41. How many reminder emails does a customer receive before escalation?

The number of email reminders is controlled by the Max Emails Total setting in your policy (separate from voice call limits). Once the configured email total is exhausted without resolution, the customer is placed on hold and your Internal AR Staff is notified for manual follow-up.

Q42. How often are reminder emails sent?

The email cadence is configurable. By default, reminder emails are sent at a regular interval (e.g., every 1 day) between each tier. Additionally, when a customer replies, the system applies a courtesy pause (e.g., 3 days) before sending the next automated email, giving the customer time to act on their response.

Q43. What happens when a customer replies to an email?

Customer replies are automatically detected and processed by the AR Agent. The agent reads the reply, detects the intent (promise to pay, dispute, already paid, etc.), and responds accordingly — just like a voice conversation but via email threading. The full email thread is logged in the customer's Comment section in Zenople.

Q44. How quickly are customer email replies processed?

Inbound emails are processed on an approximately 30-second polling interval (configured default). This means a customer's reply will typically be read and acted upon within 1 minute. Near-instant response is not guaranteed, but the turnaround is fast enough for natural email conversation flow.

Q45. Can I customize the email templates?

In the current release, the AR Agent uses standard professional templates that are pre-configured. Custom template design requires involvement from the Zenople development team. Template customization is being considered for a future release.

Q46. What email providers are supported?

The AR Agent supports two email providers:

• Microsoft 365 (Outlook / Exchange Online) — connected via Microsoft Graph and OAuth.

• Gmail — connected via Google OAuth.

Other providers (Yahoo, custom SMTP, etc.) are not supported at this time.

Q47. Does the AR Agent support plus addressing?

Yes. The AR Agent uses plus addressing (e.g., ar+12345@company.com) to thread and track email conversations per customer. Your email provider must support plus addressing. Most Microsoft 365 and Gmail accounts support this by default, but some corporate domain configurations may block it. Check with your IT team if you're unsure.

Q48. What if the customer's email goes to spam?

Email deliverability depends on the recipient's email server and spam filtering rules. Corporate firewalls and spam filters may occasionally filter collection emails. To minimize this risk:

• Use a reputable email domain with proper SPF, DKIM, and DMARC records.

• Ask your customers to add your outbound email address to their safe sender list.

• If a customer reports not receiving emails, check the email thread in Zenople's Comment section to confirm delivery was attempted.

Email delivery timing typically ranges from 1–5 minutes depending on the recipient's ISP.

5. Customer Intents — What the Agent Can Handle

The specific actions the AR Agent can detect and respond to during conversations.

Q49. What happens when a customer promises to pay?

The agent detects the Promise to Pay intent, and uses two-phase confirmation to verify the details:

1. Records the payment commitment — amount, date, and method (e.g., "I'll pay $7,000 by Friday via check").

2. Repeats the details back to the customer and asks for explicit confirmation.

3. Creates a follow-up task for the promised date.

4. Sends a confirmation email to the customer.

5. Pauses further automated outreach until the promise date.

6. Sets a verification reminder for the day after the promised date.

Q50. What happens if the customer breaks a payment promise?

If payment is not received by the promised date, the agent automatically resumes outreach. A broken-promise follow-up email or call is initiated. If a customer repeatedly breaks promises, the case is escalated to your Internal AR Staff for manual collections follow-up.

Q51. Can a customer schedule a callback?

Yes. If a customer says "call me back tomorrow at 3pm," "in 5 minutes," or "next Friday," the agent detects the Schedule Callback intent. It uses AI-powered datetime parsing to interpret natural language time references, schedules the callback at the exact agreed time, blocks other collection attempts during the agreed window, and sends a confirmation email. The callback respects quiet hours.

Q52. What happens during a scheduled callback?

When the scheduled time arrives, the agent automatically calls the customer back. It retrieves the full conversation history from the prior interaction so it has complete context. The agent reminds the customer of the previous conversation and picks up where things left off — it doesn't start from scratch.

Q53. What happens when a customer disputes an invoice?

The agent immediately stops all collection activity on the disputed invoice (FDCPA compliance). It then:

• Gathers dispute details and categorizes the reason (billing error, quality issue, rate disagreement, etc.).

• Sets a DisputeOpen hold on the customer.

• Escalates to your Internal AR Staff with the dispute details.

• Sets expectations with the customer for resolution (typically 24–48 hours).

• Pivots the conversation to other non-disputed invoices, if any.

Q54. What if the customer disputes the hours billed (timesheet dispute)?

The agent handles hours/timesheet disputes as a specific dispute type. For example, if a customer says "we only approved 30 hours but you billed 42," the agent:

• Stops collection on the disputed invoice immediately.

• Uses two-phase confirmation to verify the disputed details.

• Retrieves the original timesheet data from the system.

• Escalates to the payroll/billing team for verification.

• Continues discussing other non-disputed invoices.

Multiple disputed employees on a single invoice can be handled in a single conversation.

Q55. Can a customer request a payment plan?

Yes. If a customer says "I can't pay the full amount at once" or "can we split this over 3 weeks?", the agent detects the Request Payment Plan intent and uses two-phase confirmation to capture the details (proposed structure, down payment if offered, installment count). The agent records the request and escalates it to your Internal AR Staff — the AI does not create, approve, or process a payment plan on its own. The customer is informed that a team member will follow up to finalize the arrangement.

Note on Financial Hardship: The Financial Hardship intent is handled through a dedicated escalation path that is distinct from the payment plan flow. If a customer indicates genuine hardship (e.g., "we've lost our major contract," "we've filed for bankruptcy"), the agent captures the hardship reason, sets the customer state to Escalated, and routes to a financial hardship activity — not to the standard payment-plan escalation path. Do not treat a hardship claim and a payment plan request as equivalent — the system records them separately and your staff receives different escalation types.

Q56. What if a customer says they already paid?

The agent detects the Already Paid intent, records the claimed payment details (amount, date, method — e.g., "$8,500, last Monday, via ACH"), and places the customer on a PaymentClaimedNeedsVerification hold. Your AR staff is notified to verify the payment. The agent pauses further outreach on those invoices until verification is complete.

Q57. Can a customer request a copy of their invoice?

Yes. The agent detects the Request Invoice Copy intent, retrieves the invoice PDF from the system, and emails it to the customer's registered email address (or a different address if the customer provides one). This works for single or batch invoice requests. After sending, the agent continues the collection conversation.

Q58. Can a customer request timesheets?

Yes — this is a staffing-specific capability. If a customer says "AP won't process without timesheets" or "I need signed timesheets," the agent retrieves the timesheet documents and emails them. It can handle single-invoice or batch requests. After delivery, the agent follows up with the customer about a payment timeline.

Q59. What if the customer says "do not call me"?

The agent immediately recognizes the Do Not Call intent. It applies a DoNotCall hold, immediately stops all automated voice and email outreach to that customer, ends the call professionally, and notifies your Internal AR Staff. The hold remains until manually cleared by a Zenople user.

Q60. Can the customer update their contact information during a call?

Yes. If a customer says "our new phone number is..." or "please update my email to...," the agent detects the Update Contact Info intent, updates the contact information in the system, confirms the change with the customer, and continues the collection conversation without interruption.

Q61. What does "two-phase confirmation" mean?

Two-phase confirmation is a safety mechanism used for sensitive actions (payment promises, disputes, do-not-call requests). It works like this:

1. Phase 1 — Capture: The agent detects the customer's intent and records the details.

2. Phase 2 — Confirm: The agent repeats the details back and asks the customer to explicitly confirm before taking action.

For example: "Just to confirm — you're committing to pay $7,000 by this Friday via check. Is that correct?" Only after the customer confirms does the agent execute the action. This prevents misunderstandings and ensures the customer's explicit agreement.

Q62. What happens if the AR Agent doesn't understand what the customer wants?

If the agent encounters an intent it doesn't recognize or can't handle, it automatically escalates to your Internal AR Staff rather than guessing incorrectly. The agent generates a conversation summary and sends an escalation notification with full context. Unsupported intents do not cause system failures — the agent is designed to fail safely by escalating.

Q63. What happens if a customer asks for a discount?

Discount requests are not a handled intent in the current release. The agent does not have authority to offer, negotiate, or approve discounts. When a customer asks for a discount — for example, "Can you give us 10% off if we pay today?" — the agent acknowledges the request, explains that discount decisions require human review, and escalates to your Internal AR Staff with the full conversation context and the customer's specific request. The customer is informed that a team member will follow up.

Q64. What happens if a customer wants to make a partial payment (not a payment plan)?

If a customer says "I can pay $5,000 of the $8,000 today" without proposing a structured installment plan, the agent detects this as a partial payment intent. It captures the offered amount and payment details, then escalates to your Internal AR Staff for approval. The agent does not accept or reject partial payments on its own — your team decides whether to accept the partial amount and how to handle the remaining balance.

Q65. What happens if the customer says they are no longer with the company or the business has closed?

The agent recognizes this as a significant contact status change. If the customer's contact person says they've left the company, the agent asks for the new contact's information. If the person indicates the business has closed or is no longer operating, the agent records this information, places the customer on hold, and escalates to your Internal AR Staff with the details. Your team can then investigate and update the customer's status accordingly.

Q66. What happens if the customer says "I need to check with my manager" or "I'll have my boss call you"?

The agent treats this as a soft deferral. It acknowledges the customer's need to consult internally, and offers to schedule a callback at a specific time convenient for the customer. If the customer provides a callback time, the agent schedules it. If the customer doesn't commit to a time, the agent notes the deferral and continues contacting the customer on the normal cadence. The conversation is logged with the deferral context so the next interaction can reference it.

Q67. What if the customer asks a question the agent can't answer (e.g., invoice breakdown, specific charges)?

The agent does not fabricate or guess at answers. If a customer asks a detailed question that is outside the agent's available data — such as "What's the $347 charge on line 12?" or "Why was our rate increased?" — the agent acknowledges that it doesn't have that specific information and escalates to your Internal AR Staff with the customer's exact question captured. The agent continues discussing other topics it can address (e.g., other invoices, payment arrangements) before ending the conversation.

6. Hold States & Statuses

Understanding the holds the AR Agent places on customers and how to manage them.

Q68. What is a hold state?

A hold state is an automatic flag the AR Agent places on a customer when certain conditions are met during the collection process. When a customer is on hold, the agent pauses all outreach to that customer. Hold states are displayed as banners at the top of the customer's AR Agent Policy page in Zenople.

Q69. What hold states can the AR Agent set?

There are 11 hold states, split into two categories:

Holds requiring Resume Campaign action (red banner):

MaxAttemptsReached · WrongPerson · DoNotCall · Escalated · DisputeOpen · PaymentIntentExpressed · PaymentClaimedNeedsVerification · PromiseScheduled · PromiseBroken · PromisePartiallyKept

Informational status (green banner — no action needed):

CallbackScheduled

Q70. What is the difference between a red hold and a green status?

Red holds pause all outreach and require a Zenople user to click Resume Campaign before the agent will contact that customer again. These indicate situations that need human review or resolution.

Green statuses are informational only — the agent is managing the situation automatically (e.g., a callback is already scheduled). No user action is needed.

Q71. What does "Max Attempts Reached" mean?

This hold means the applicable channel has exhausted its configured total limit without resolving the outstanding invoices. Email and voice totals are tracked separately. The agent stops all outreach and may notify your Internal AR Staff. To resume contact, review the customer's situation, optionally adjust the channel limits in the policy, and click Resume Campaign.

Q72. What does "Payment Claimed Needs Verification" mean?

The customer told the agent they already paid. The agent recorded the claimed payment details (amount, date, method) and placed them on hold pending your team's verification. Check your payment records — if the payment was received, you can close the invoices. If not, clear the hold to resume collection outreach.

Q73. What does "Promise Scheduled" mean?

The customer made a promise to pay by a specific date. The agent recorded the commitment details and placed the customer on hold until the promise date. After the promise date, the agent checks whether payment was received. If it was not, a broken-promise follow-up is triggered. You can also clear this hold manually if needed.

Q74. How do I clear a hold on a customer?

Navigate to the customer's AR Agent Policy page in CMS. Look for the red "Customer On Hold" banner at the top. Click the "Resume Campaign" button. A confirmation dialog will ask you to confirm. Once confirmed, the customer is returned to Active status and becomes eligible for AR Agent contact again.

Q75. Does clearing a hold reset the attempt counters?

Yes. Clicking Resume Campaign removes the customer's state record entirely — this resets the customer to implicit Active status and also resets all attempt counters (email and voice, daily and total). The one exception is the Dispute Claim action (available separately): this transitions the customer to DisputeOpen state rather than deleting the row. All other Resume Campaign actions — including markVerified, resolveEscalation, closeDispute, removeDNC, and others — perform a row delete and return the customer to Active.

Q76. What happens to the customer after a hold is cleared?

The customer becomes eligible for AR Agent contact again. On the next polling cycle (within 20 minutes), the agent will re-evaluate the customer against your policy thresholds. If they still have overdue invoices meeting the criteria, the agent will initiate a new collection workflow.

7. Configuration & Policies

How to set up and manage AR Agent policies at every level.

Q77. What is the configuration hierarchy (Company → Office → Customer)?

The AR Agent uses a 3-tier inheritance model:

1. Company Level — Sets the default policy for all offices and customers under that company.

2. Office Level — Can override specific settings for a branch/location. Unset fields inherit from Company.

3. Customer Level — Can override specific settings for an individual customer. Unset fields inherit from Office (or Company if Office doesn't override).

This means you configure once at the company level, then only override where specific offices or customers need different treatment.

Q78. What happens if I leave a field empty at the office or customer level?

Empty fields inherit from the parent level. For example, if you leave "Days Past Due" empty at the office level, that office uses the company-level value. This is by design — it lets you override only what's different without re-entering every field.

Q79. What is the difference between Opt-In and Opt-Out mode?

The "Include All Customers By Default" toggle controls this:

• Opt-Out (toggle ON) — All customers are automatically included. To exclude a specific customer, you must go to their customer-level policy and disable the AR Agent toggle. Best when most customers should receive automated collections.

• Opt-In (toggle OFF) — No customers are contacted by default. You must individually enable the AR Agent at the customer level. Best when you want to selectively enroll specific accounts.

Q80. How do I enable or disable the AR Agent for a specific customer?

In CMS, open the customer's detail view, navigate to AI Agents → AR Agent Policy, and toggle the Enable AR Agent switch on or off. When disabled, the agent will not contact that customer regardless of the company-level setting.

Q81. Can I set different rules for different offices?

Yes. Navigate to the office in ATM mode, open AI Agents → AR Agent Policy, and override any settings. For example, one office could have a 5-day grace period while another has 14 days, or one office could use Voice Only while another uses Email Only. Fields left empty inherit from the company level.

Q82. How do I change the outbound phone number or email address?

The outbound email address and phone number are configurable at the company level and can be overridden at the office level. They are not available at the customer level. To change the company-wide defaults, edit the company-level AR Agent Policy. To use different outbound credentials for a specific office (e.g., a regional branch with a different Plivo number), set overrides at the office level. The email address must correspond to an email profile you've already connected (see the Email Profile Setup section in the Setup Guide).

Q83. Who is the "Client AR Contact" vs. "Internal AR Staff"?

• Client AR Contact — The customer's point of contact for AR matters. This is the person the agent will try to reach (call or email). Configured at the customer level only.

• Internal AR Staff — Your staff member who handles escalated cases. The agent notifies this person when human intervention is needed (disputes, do-not-call, complex situations). Configurable at company, office, and customer levels.

Note: In read-only mode, these appear as "Primary Contact" and "Escalation Contact" respectively.

Q84. What is "Agent Availability" and how does it differ from quiet hours?

• Quiet Hours define when the agent should not contact customers (e.g., evenings and early mornings). This protects the customer's rest hours.

• Agent Availability defines when your Internal AR Staff is available for escalations (e.g., 8:00 AM – 6:00 PM). This helps the agent decide whether to attempt a warm handoff or schedule a callback when escalation is needed.

Agent Availability is configured at the company level only.

Q85. Can I configure the AR Agent for multiple companies?

Yes. Each company in Zenople can have its own independent AR Agent policy with different settings, outbound email, outbound phone number, and escalation contacts. The agent operates per-company with full tenant isolation.

Q86. Where do I find the AR Agent Policy settings in Zenople?

The AR Agent Policy is always accessed via the AI Agents → AR Agent Policy navigation path:

• Company Level: ATM mode → Company → open detail → hover navigation toggle → AI Agents → AR Agent Policy.

• Office Level: ATM mode → Company → Office → open detail → hover navigation toggle → AI Agents → AR Agent Policy.

• Customer Level: CMS mode → Customer → open detail → hover navigation toggle → AI Agents → AR Agent Policy.

Q87. What is the Agent Name field and how does it affect the AR Agent?

The Agent Name is a company-level field that controls the name the agent uses when introducing itself on voice calls and in email correspondence. By default, the agent introduces itself as "Ava." If you change this field to a different name (e.g., "Sophie"), the agent will use that name in all future interactions. Changing the Agent Name does not affect the agent's conversation logic, intent detection, or any other behavior — only the spoken and written name changes.

Q88. What is the Introductory Company Name field?

The Introductory Company Name is a company-level field that overrides the company name the agent uses in its opening greeting — for example, "Hi, this is Ava calling from Acme Staffing…" This is useful when your legal entity name differs from the brand name your customers recognize. If left blank, the agent uses the system company name from Zenople. This field affects voice call introductions and email sender display names.

Q87a. What is the Voice Agent Prompt Library and how do I use it?

The Voice Agent Prompt Library controls the conversational behavior and tone of the AI voice agent. It is configured per-level in the AR Agent Policy form (Section: Voice Agent Prompt Library).

Prompt Strategy (tri-state):

• Inherit(Office and Customer levels only) — Use the prompt configuration from the parent level. This is the default for office and customer. No local prompt is stored.

• System Default — Use Zenople's standard system prompt for this tenant. No custom content is stored locally. This is the only option available at the company level by default.

• Custom Prompt — Activate the sectioned prompt editor and select a saved prompt version. The agent uses your custom content for the configurable sections while retaining the immutable compliance, guardrail, and tool-contract sections managed by Zenople.

• Voice Agent Prompt Library is a structured prompt system, not a free-form blank text box.

• Persona & Identity defines who Agent is.

• Greeting defines the opening style.

• Guiding Principles defines the behavioral rules.

• Advanced Sections contains the detailed blocks that make the prompt useful and safe.

• Custom Prompt means the user is editing a saved version of the structured prompt, not replacing the locked system prompt entirely.

Custom prompt behavior:

• Selecting Custom Prompt requires choosing an active prompt version from your prompt library.

• Switching back to Inherit or System Default clears the local active prompt and custom content before saving.

• Prompts are version-controlled — published versions are immutable; drafts can be edited. Click Save as New Version to publish a draft; then click the main policy Save to bind the new version as active.

• Prompt resolution follows Customer → Office → Company until the first non-Inherit level.

The Test Call feature (customer level only — see Q134) allows you to validate a prompt change against a live call before saving.

Q89. Which configuration fields are available at which levels (Company, Office, Customer)?

Use this quick-reference to see where each field can be set:

✓ = Configurable at this level | — = Not available (inherited or not applicable)

Q90. Which settings are NOT available at the office or customer level?

The following fields are company-level only and cannot be overridden at the office or customer level:

• Outbound Email Address and Outbound Phone Number — Configurable at the company level and optionally overridable at the office level. Not available at the customer level. The customer always uses the outbound identity configured at its parent office or company.

• Agent Availability Start / End — Your Internal AR Staff's availability hours are set company-wide.

• Agent Name and Introductory Company Name — The agent's identity and brand name are consistent across all offices and customers.

• Incoming Call Forward To — The inbound call forwarding destination is set company-wide.

Additionally, the Client AR Contact (Name, Email, Phone, Ext.) is customer-level only — it cannot be set at the company or office level because it represents the individual customer's AR contact person.

Q91. What is the "Incoming Call Forward To" field?

The Incoming Call Forward To field is a company-level setting that specifies a phone number to which inbound calls to the AR Agent's Plivo number should be forwarded. This is useful when a customer calls back the number that appeared on their caller ID. Instead of reaching the AI agent (which handles outbound calls only), the call is forwarded to your designated staff member or main office line. If left blank, inbound calls are not forwarded.

Q92. Can I set different quiet hours for weekdays vs. weekends?

Yes. The AR Agent supports separate quiet-hour windows for weekdays and weekends. In the policy form you will find four quiet-hour inputs:

• Weekday Quiet Hours Start / End — no-contact window for Monday–Friday

• Weekend Quiet Hours Start / End — no-contact window for Saturday–Sunday

Leave the weekend fields empty to apply the same window as weekdays, or set different times for different weekend behavior. Agent Availability also has corresponding weekday and weekend time inputs.

8. Monitoring, Transcripts & Audit

Tracking and reviewing the AR Agent's activity.

Q93. Where can I see the AR Agent's activity for a customer?

Navigate to the customer in CMS, then click the Comment tab. All AR Agent activity — voice call transcripts, email threads, and AI-generated summaries — appears here as chronological comment records. You can filter by Category and Type to narrow results.

Q94. What information is included in a call transcript?

Each voice call transcript includes:

• AI-Generated Summary — Bullet-point list of key actions and outcomes (e.g., "Agent noted request to apply $194.62 unapplied credit toward the outstanding balance").

• Full Transcript — Color-coded dialogue with timestamps. Agent lines and Customer lines are visually distinct.

• Metadata — Date, time, duration, and the automation user who logged the entry.

Q95. Does the AR Agent generate summaries of its conversations?

Yes. After every voice call and email interaction, the agent generates an AI-powered summary with key action items, commitments made, and next steps. These summaries appear at the top of each transcript entry in the Comment section. They are also included in escalation notifications sent to your Internal AR Staff.

Q96. How do I know if the AR Agent successfully contacted a customer?

Check the customer's Comment section for recent entries. Successful contacts will have Voice Call Notes or email thread entries. You can also check the customer's AR Agent Policy page — it shows the last attempt date, last contact date, and current attempt counts. If the customer is on hold, the hold state banner will explain why.

Q97. Is there a dashboard or report for AR Agent performance?

A dedicated analytics dashboard is not available in the current Beta release. You can monitor individual customer activity through the Comment section and hold states. A comprehensive Dashboard & Analytics feature — with real-time agent metrics, collection trends, and KPIs — is planned for the Release 1.0 (Production GA) milestone.

9. Security, Privacy & Compliance

Data protection, tenant isolation, and regulatory compliance.

Q98. Is my data isolated from other tenants?

Yes. The AR Agent uses a schema-per-tenant database design that provides complete data isolation. Your customer data, conversation transcripts, configuration, and audit trails are completely separate from other tenants. No other company can see or access your data.

Q99. Does the AR Agent comply with FDCPA regulations?

The AR Agent is designed with compliance-first principles:

• Immediate Dispute Halt: When a customer disputes an invoice, all collection activity on that invoice stops immediately — as required by FDCPA.

• Do Not Call: When a customer requests no further calls, the agent stops all outreach immediately.

• Quiet Hours: Configurable no-contact windows prevent calls during inappropriate hours.

• Two-Phase Confirmation: Sensitive actions require explicit customer agreement before execution.

• No Legal Threats: The agent never threatens legal action or advises on collections law.

Q100. Where are conversation transcripts stored?

All conversation transcripts (voice and email) are stored in your tenant's Zenople database, accessible through the customer's Comment section. Transcripts include full dialogue, timestamps, speaker identification, and AI-generated summaries.

Data processing note: To deliver voice AI and email capabilities, Zenople integrates with external providers including Azure OpenAI (AI reasoning), Plivo (voice telephony), and your configured email provider (Microsoft Graph or Google OAuth). Customer data relevant to the active interaction is processed by these integrated providers under their respective security and compliance frameworks. Tenant business data and transcript storage remain within your isolated Zenople tenant database.

Q101. Who can see the AR Agent's activity and transcripts?

AR Agent activity is visible to Zenople users with access to the customer's record. Conversation transcripts, hold states, and policy settings are accessible through the standard Zenople CMS and ATM interfaces. Access is governed by your existing Zenople user roles and permissions.

Q102. Does the AR Agent ever threaten legal action?

No. The AR Agent never threatens legal action, never advises on collections law, and never implies legal consequences. It persuades and negotiates professionally but does not make threats. If a situation requires legal discussions, it escalates to your Internal AR Staff.

Q103. How does two-phase confirmation protect my customers?

Two-phase confirmation prevents the agent from taking sensitive actions based on misunderstood intent. Before executing a payment promise, dispute flag, do-not-call request, or any consequential action, the agent repeats the details back and asks the customer to explicitly confirm. This means a casually mentioned date ("maybe Friday…") is not treated as a binding promise unless the customer confirms it.

10. Scope, Limitations & Roadmap

What the AR Agent can't do today and what's coming next.

Q104. Does the AR Agent process payments?

No. The AR Agent does not process payments, collect credit card numbers, or authorize transactions. It captures payment commitments (promise to pay by a certain date) and can escalate "pay now" requests to your staff for immediate processing. Payment gateway integration is planned for Release 1.0.

Q105. Does the AR Agent support SMS or text messaging?

No. The current release supports voice calls and email only. SMS, WhatsApp, and live chat are not supported at this time.

Q106. Does the AR Agent support languages other than English?

Not yet. The AR Agent currently operates in English only. Multi-language support (Spanish, French, and additional languages with automatic detection) is planned for Release 1.0.

Q107. Can I customize the voice or conversation flow?

The AR Agent supports a Custom Prompt mode that allows you to edit configurable sections of the AI's voice conversation script through the Voice Agent Prompt Library in the AR Agent Policy. This covers the editable (30%) portion of the prompt — tone, introductory phrasing, company-specific instructions, and similar content. The core compliance rules, intent-handling guardrails, and tool-call contracts (70%) remain immutable and are managed by Zenople. Full no-code workflow builders for custom intent flows are not available in the current release.

Q108. Can the AR Agent detect if a customer is bankrupt?

No. The agent does not check bankruptcy databases or detect bankrupt customers. If you know a customer is in bankruptcy, disable the AR Agent for that customer manually. Bankruptcy awareness is not currently on the roadmap.

Q109. Does the AR Agent calculate late fees?

No. The agent does not calculate or apply late fees. Late fee calculation and application must be handled manually by your AR team. The agent works with invoice balances as they exist in the system.

Q110. What happens if the agent encounters a situation it can't handle?

The agent is designed to fail safely. For any intent it doesn't recognize or any situation outside its capabilities, it automatically escalates to your Internal AR Staff with a full conversation summary. It will never guess at an action or take an unsupported step. The conversation ends gracefully, and the customer is informed that a team member will follow up.

Q111. What features are planned for the next release?

The Release 1.0 (Production GA) milestone includes:

• Payment Gateway Integration — Self-service payment processing via industry-standard payment partners.

• Dashboard & Analytics — Real-time visibility into agent metrics, customer behavior, collection trends, and key business indicators.

• Multi-Language Support — Spanish, French, and additional languages with automatic language detection.

Further enhancements will be evaluated as the system evolves based on customer feedback.

11. Troubleshooting

Common issues and how to resolve them.

Q112. Why isn't the AR Agent contacting a customer with overdue invoices?

Work through these checks in order:

1. AR Agent enabled? Verify the Enable AR Agent toggle is ON at the company level — and that the customer hasn't been individually disabled at the customer level.

2. Thresholds met? The customer's oldest overdue invoice must exceed both the Days Past Due and Minimum Invoice Balance thresholds configured in the policy.

3. Customer on hold? Check the customer's AR Agent Policy page for a red hold banner (e.g., DoNotCall, DisputeOpen, MaxAttemptsReached). Holds pause all outreach.

4. Include All Customers? If the company-level "Include All Customers By Default" toggle is OFF (Opt-In mode), the customer must be individually enabled.

5. Quiet hours? The agent will not initiate contact during configured quiet hours. Wait for the next active window.

Q113. Why aren't emails being sent to a customer?

Verify the following:

• Communication channel — Is it set to Email Only or Voice & Email? If set to Voice Only, no emails will be sent.

• Email profile connected — The company must have an active email profile (Gmail or Microsoft) linked in the AR Agent Policy. If the profile is disconnected, re-authenticate it.

• Outbound email address — Confirm an outbound email address is configured at the company level and matches the connected profile.

• Customer email — The Client AR Contact must have a valid email address on file.

• Check Comment tab — Look for email delivery error entries that indicate a bounce or sending failure.

Q114. Why aren't voice calls being made?

Verify the following:

• Communication channel — Is it set to Voice Only or Voice & Email? If set to Email Only, no calls will be placed.

• Outbound phone number — Confirm a Plivo phone number is configured at the company level.

• Customer phone — The Client AR Contact must have a valid phone number on file.

• Quiet hours — The agent will not place calls during configured quiet hours.

• Check Comment tab — Look for call attempt entries. If attempts are logged but calls aren't connecting, the issue may be with the customer's phone number.

Q115. A customer appears stuck on hold — what should I do?

The agent placed the hold for a specific reason, which is shown in the red hold banner on the customer's AR Agent Policy page. To resolve it:

1. Read the hold reason (e.g., DisputeOpen, PaymentClaimedNeedsVerification, MaxAttemptsReached).

2. Take the appropriate action — verify a claimed payment, resolve a dispute, review an escalation, etc.

3. Once the underlying issue is handled, click "Resume Campaign" to return the customer to Active status.

Clicking Resume Campaign resets the attempt counters and makes the customer eligible for AR Agent contact on the next polling cycle (within approximately 20 minutes).

Q116. I changed office-level settings but they don't seem to take effect.

Two common causes:

• Customer-level overrides — If a customer has explicit values set at the customer level, those take precedence over the office level. Check whether the affected customers have customer-level policy overrides.

• Polling delay — Configuration changes take effect on the next polling cycle, which runs approximately every 20 minutes. Allow up to 20 minutes for changes to propagate to active workflows.

Q117. I don't see any transcripts or activity in the Comment section.

First, confirm the agent has actually attempted contact:

• Is the AR Agent enabled for this customer and company?

• Do the customer's overdue invoices meet the configured thresholds (Days Past Due, Minimum Balance)?

• Is the customer Active (not on hold)?

If attempts should have been made, check that your Comment filters are not hiding entries. Voice call transcripts and email threads are logged as separate comment types — make sure the Category and Type filters include AR Agent entries.

Q118. Why is a customer being skipped even though they have overdue invoices?

Beyond the standard threshold and hold checks (see Q112), consider these additional causes:

• Customer opted out — The customer may be individually disabled at the customer level, even if the company uses Opt-Out (Include All) mode.

• Callback pending — A green CallbackScheduled status means the agent is waiting for a scheduled callback time. The customer will be contacted at the agreed time.

• All invoices disputed — If every overdue invoice for the customer has an active dispute, the agent will not initiate outreach (FDCPA compliance).

Q119. Why did the agent stop contacting a customer after a call?

The call likely resulted in a customer intent that placed a hold on the account. Common scenarios:

• PromiseScheduled — The customer promised to pay by a specific date. The agent pauses outreach until that date.

• DisputeOpen — The customer disputed one or more invoices. Collection activity stops on disputed invoices.

• DoNotCall — The customer requested no further calls. All outreach is halted.

• MaxAttemptsReached — The agent exhausted the configured maximum attempts.

Check the customer's AR Agent Policy page for the hold banner to see which hold was applied and take appropriate action.

Q120. What should I do if the email profile shows as disconnected?

A disconnected email profile means the OAuth authorization has expired or been revoked. To fix it:

1. Navigate to the company-level AR Agent Policy.

2. Go to the Email Profile section.

3. Click Reconnect and re-authorize the email account (Gmail or Microsoft).

Make sure the Outbound Email Address configured in the policy matches the reconnected email profile. Emails will resume on the next polling cycle after reconnection.

12. Email Profile Setup

Connecting your Gmail or Microsoft email account for AR Agent communication.

Q121. How do I set up the email profile for the AR Agent?

Before configuring AR Agent policies, you need to connect an email account that the agent will use to send collection emails and receive customer replies. The setup has four steps:

1. Open the Email Module — Click the Shortcut icon in the top toolbar, then select Email from the shortcut menu.

2. Create a New Email Profile — In the Email board, click the menu icon (⋮) at the top, then select New Email Profile.

3. Authenticate Your Email Account — Choose either Gmail (OAuth verification) or Microsoft (Graph Setup) authentication.

4. Verify Connection — Confirm the profile shows as authenticated and connected.

The email account you connect here will be used as the "from" address on all outgoing collection emails.

Q122. How do I connect a Gmail account for AR Agent emails?

In the Add Email Profile dialog:

1. Click "Use OAuth verification" to initiate Google's standard OAuth flow.

2. Sign in with the Gmail account you want the AR Agent to use for outbound collection emails.

3. Grant the requested permissions to allow Zenople to send and receive emails on behalf of the account.

4. Once authenticated, the profile will show as connected.

The Gmail account should be dedicated to AR outreach — avoid using a personal inbox that receives unrelated email.

Q123. How do I connect a Microsoft 365 (Outlook) account for AR Agent emails?

In the Add Email Profile dialog:

1. Scroll to the "AR Agent Microsoft Graph Setup" section at the bottom.

2. Click "Connect with Microsoft."

3. The standard Microsoft "Pick an account" dialog appears — select the email account you want the AR Agent to use.

4. Grant the requested permissions to allow Zenople to send and receive emails via Microsoft Graph.

The Microsoft Graph setup is independent of the email profile authentication section above it in the dialog. The AR Agent uses Microsoft Graph specifically to send and receive emails on behalf of your account.

Q124. What is the "AR Agent Microsoft Graph Setup" section in the email profile dialog?

The AR Agent Microsoft Graph Setup section is a separate authentication area at the bottom of the Add Email Profile dialog. It connects the AR Agent to Microsoft Graph — Microsoft's API for sending and reading emails. This is required for Microsoft 365 (Outlook/Exchange Online) accounts. It is independent of the general email profile authentication at the top of the dialog. You must complete this section if you are using a Microsoft email account for AR Agent outreach.

Q125. What should I do if the Microsoft account picker shows the wrong account?

If you are logged into multiple Microsoft accounts and the "Pick an account" dialog shows the wrong one, use an Incognito/Private browser window to re-authenticate. This ensures you sign in with the correct Microsoft account without interference from other active sessions. After authenticating in the private window, the correct account will be linked to the AR Agent's email profile.

Q126. Can I use a shared mailbox (Microsoft 365) for the AR Agent?

A shared mailbox can work, but it depends on your Microsoft 365 configuration and Graph API permissions. A dedicated user mailbox is recommended for the most reliable experience — shared mailboxes sometimes have permission restrictions that can interfere with OAuth and Graph API access. If you need to use a shared mailbox, consult your IT team to ensure the necessary Graph API permissions are granted, and test the connection before going live.

Q127. What permissions does the OAuth connection grant to Zenople?

The OAuth connection grants Zenople permission to send and read emails on behalf of the connected account. This is the minimum required to send collection emails and process customer replies. The connection does not grant access to calendars, contacts, OneDrive files, Teams messages, or administrative functions. For Microsoft accounts, permissions are scoped through Microsoft Graph. For Gmail, they are scoped through Google's OAuth consent screen.

13. Initial Setup & Go-Live

First-time configuration walkthrough and go-live checklist.

Q128. What is the step-by-step process to set up the AR Agent for the first time?

Follow these steps in order:

1. Connect an Email Profile — Set up a Gmail or Microsoft 365 email account for the AR Agent (see Category 12: Email Profile Setup).

2. Configure the Company-Level AR Agent Policy — Set the baseline collection rules, communication channel, outbound email/phone, quiet hours, and escalation contacts.

3. (Optional) Configure Office-Level Overrides — If specific branches need different rules, set overrides at the office level.

4. (Optional) Configure Customer-Level Overrides — For individual customers that need special treatment, set overrides at the customer level.

5. Set the Client AR Contact — For each customer you want the agent to contact, ensure their AR contact (name, email, phone) is populated at the customer level.

6. Test with a Single Customer — At the customer-level AR Agent Policy, use the Test Call button (Voice Agent Prompt Library section, edit mode) to launch a controlled test call immediately. Before starting, verify a valid primary contact phone and configured outbound phone number. The test call uses the current policy/prompt context (including unsaved overrides) and does not record contact attempts, claim normal collection ownership, or create normal collection side effects. Alternatively, enable the AR Agent for the customer and wait for the configured conductor cycle for a real collection call.

7. Review & Go Live — After confirming the test customer's interactions are correct, roll out to all customers (Opt-Out model) or enable additional customers individually (Opt-In model).

Initial setup typically takes 15–30 minutes.

Q129. How do I configure the company-level AR Agent Policy (step by step)?

1. Switch to ATM mode in Zenople.

2. Click Company in the left navigation, then open the detail view for the desired company.

3. Hover over the navigation toggle at the bottom of the page. Hover over AI Agents, then click AR Agent Policy.

4. The page opens in read-only mode showing current settings. Click the Edit icon (pencil) to enter edit mode.

5. Configure each section:

   • AR Agent Status — Toggle "Enable AR Agent" ON and set the inclusion model (Opt-In or Opt-Out).

   • Collection Workflow Settings — Set Days Past Due, Minimum Invoice Balance, Max Emails Per Day, Max Emails Total, Max Voice Calls Per Day, Max Voice Calls Total, and Email Cadence Days.

   • Communication Settings — Choose the Preferred Channel and Timezone.

   • Quiet Hours — Set the Start and End times for the no-contact window.

   • Outbound Communication — Enter the Outbound Email Address (must match a connected email profile) and Outbound Phone Number (Plivo number).

   • Internal AR Staff — Enter the Name, Email, and Phone of the staff member who handles escalations.

   • Agent Availability — Set the Start and End times for when your AR staff is available for escalations.

   • Voice Agent Prompt Library — Choose System Default to use the standard Zenople prompt, or Custom Prompt to activate the sectioned prompt editor and select a prompt version. Company level supports System Default and Custom only (no Inherit).

6. Click Save.

Q130. What do all the Collection Workflow and Communication fields mean?

• Days Past Due Before Collection — How many days after an invoice's due date the agent waits before initiating first contact. Example: set to 7 to give customers a 7-day grace period.

• Minimum Invoice Balance ($) — Skip invoices below this dollar amount. Prevents the agent from contacting customers for trivial balances.

• Max Emails Per Day — Maximum number of email attempts to a single customer per day (default: 3).

• Max Emails Total — Total email attempts before the customer is placed on a MaxAttemptsReached hold (default: 5).

• Max Voice Calls Per Day — Maximum number of voice call attempts to a single customer per day (default: 3).

• Max Voice Calls Total — Total voice call attempts before the customer is placed on a MaxAttemptsReached hold (default: 5).

• Preferred Channel — Primary method for contacting customers — Email Only, Voice Only, or Email Then Voice.

• Timezone — Used to calculate quiet hours and determine when the agent can contact customers.

Q131. How do I configure outbound communication (email and phone)?

In the company-level AR Agent Policy (edit mode), locate the Outbound Communication section:

• Outbound Email Address — Enter the email address that the AR Agent will use to send collection emails. This must correspond to an email profile you've already connected (see Category 12). Emails sent by the agent will appear to come from this address.

• Outbound Phone Number — Enter the Plivo phone number provisioned by Zenople during onboarding. This is the caller ID that customers will see for outbound voice calls.

Both fields are company-level only — they cannot be overridden at the office or customer level.

Q132. How do I set up Internal AR Staff and Agent Availability?

In the company-level AR Agent Policy (edit mode):

• Internal AR Staff — Enter the Name, Email, and Phone of the staff member who handles escalated cases. The agent notifies this person for disputes, do-not-call requests, and other situations requiring human judgment. This can be overridden per office or per customer.

• Agent Availability Start / End — Set the time window when your Internal AR Staff is available (e.g., 8:00 AM – 6:00 PM). The agent uses this to decide whether to attempt a warm handoff during escalations or schedule a priority callback. This is company-level only.

Q133. What is the read-only view vs. edit mode, and what label differences should I know about?

The AR Agent Policy page always opens in read-only mode, showing all current settings. Click the Edit icon (pencil) to switch to edit mode.

Important label difference: In read-only mode, "Primary Contact" corresponds to the Client AR Contact in edit mode, and "Escalation Contact" corresponds to Internal AR Staff. These are the same fields — just displayed with different labels depending on the view.

At the office level, a yellow banner reminds you that changes override company defaults, and that empty fields inherit from the parent level.

Q134. What is the recommended testing checklist before going live?

Before rolling out the AR Agent to all customers, verify:

• Email Profile Connected — The email profile shows as authenticated and the Outbound Email Address matches the connected profile.

• Company Policy Saved — The company-level policy is enabled, and all required fields (Days Past Due, Minimum Balance, Max Attempts, Preferred Channel, Timezone) have values.

• Test Customer Enabled — Enable the AR Agent for a single test customer with a known overdue invoice that meets your configured thresholds.

• Run a Test Call (Optional but Recommended) — At the customer-level AR Agent Policy, click the Test Call button (available in edit mode, Voice Agent Prompt Library section). Confirm the customer has a valid primary contact phone number and that an outbound phone number is configured. The test call places a controlled real call using the current (including unsaved) policy and prompt context, but does not record a normal collection attempt or apply normal collection ownership/side effects. Use this to validate prompt behavior and configuration before the first live collection cycle.

• Verify Outreach — After the next configured polling cycle, check the customer's Comment section for agent activity (voice call transcript or email thread).

• Review Transcripts — Confirm the call transcript or email content is accurate, professional, and references the correct invoices.

• Test Hold States — Verify that holds are triggered correctly (e.g., promise to pay creates a PromiseScheduled hold) and can be resumed via the Resume Campaign button.

Once testing is successful, expand to all customers by enabling the Opt-Out model or enabling additional customers individually.

Zenople AR Agent — Frequently Asked Questions © 2026 Zenople. All rights reserved.