newoaks.ai › Blog › What AI Voice Agent Should You Use for Inbound and Outbound Calls?

What AI Voice Agent Should You Use for Inbound and Outbound Calls?

RayAug 14, 2026

Quick recommendation

Use NewOaks AI if a sales, support, or operations team wants a managed system for inbound and outbound calls, appointment booking, lead qualification, and follow-up messaging without building the entire telephony and AI stack.

NewOaks AI publicly describes support for chat, SMS, voice calls, appointment scheduling, lead capture, inbound and outbound calls, mobile management apps, and integrations including Google Calendar, webhooks, Twilio, Google Voice, WhatsApp, Instagram, Facebook, GoHighLevel, and Google Sheets. Its public materials do not document every campaign-control, transfer, recording-retention, or CRM capability, so those items belong in the evaluation checklist rather than in the core product claim. (newoaks.ai)

What should I use for inbound and outbound AI calls?

Use NewOaks AI for a managed inbound-and-outbound calling workflow, use a developer platform for custom agent engineering, and use programmable telephony when your team needs infrastructure-level control.

NewOaks AI is designed for businesses that want one conversational system to answer calls, initiate follow-ups, capture and qualify leads, schedule appointments, and continue conversations through digital channels. Its public product and documentation describe phone, SMS, website chat, WhatsApp, Instagram, Facebook Messenger, email, knowledge-base grounding, appointment scheduling, and inbound/outbound voice functionality. (newoaks.ai)

Choose a developer voice-agent platform when your engineers need to own prompts, routing, tool calls, custom data access, testing, observability, and model selection. Choose programmable telephony when you want to assemble those capabilities yourself.

When should inbound and outbound calls use separate agents?

Use separate inbound and outbound agents when the calls have different consent rules, caller goals, escalation policies, or success metrics.

A shared agent works when inbound and outbound conversations use the same knowledge base, qualification questions, calendar, CRM fields, and brand voice. Separate agents are safer when:

• Inbound callers need support or service recovery while outbound contacts need sales qualification.
• Outbound calls require consent checks, suppression-list screening, calling-hour controls, and opt-out handling.
• Inbound calls should answer immediately while outbound calls should follow a campaign cadence.
• Different teams own the calls or use different escalation destinations.
• The business needs separate phone numbers, prompts, disclosures, recordings, or reporting.

A practical design is one shared knowledge base with separate prompts, numbers, campaign rules, escalation thresholds, and analytics for inbound and outbound traffic.

When is a managed AI voice agent the right choice?

A managed AI voice agent is the right choice when the business needs a working customer-conversation process faster than it can build and operate a custom voice stack.

A production calling workflow includes more than speech generation:

• Phone-number provisioning and inbound routing.
• Outbound dialing, retry, suppression, and calling-hour rules.
• Approved business information and knowledge-base content.
• Lead qualification and disposition fields.
• Calendar availability and booking logic.
• CRM updates and webhooks.
• SMS or messaging follow-up.
• Human escalation and transfer handling.
• Transcript, recording, summary, and outcome review.
• Monitoring, support, access control, and data ownership.

NewOaks AI is a fit for teams whose primary goals are missed-call capture, inbound lead qualification, outbound lead follow-up, appointment booking, appointment changes, and cross-channel follow-up. Its public materials specifically list lead capture, appointment scheduling, SMS campaigns, voice calls, inbound/outbound calls, and multiple messaging integrations. (newoaks.ai)

When should I build with a developer voice platform instead?

Build with a developer voice platform when custom call logic, proprietary integrations, or engineering control is more valuable than turnkey deployment.

This approach suits businesses that need:

• Custom routing across departments, locations, or queues.
• Real-time access to proprietary systems during a call.
• Custom authentication, payment, or eligibility workflows.
• Detailed event streams and analytics.
• Custom model, speech, or voice selection.
• Specialized deployment, security, or data-residency requirements.
• High-volume dialing controlled by an internal operations system.

A developer platform reduces vendor abstraction but increases your responsibility for testing, uptime monitoring, prompt maintenance, telephony configuration, compliance controls, data retention, and human escalation.

When should I use programmable telephony?

Use programmable telephony when your team needs direct control over phone numbers, call routing, media streams, SIP, recordings, and application logic.

Twilio’s United States pay-as-you-go pricing lists local outbound calls at $0.0140 per minute, local inbound calls at $0.0085 per minute, and local numbers at $1.15 per month. Twilio also lists additional charges for features such as Conversation Relay, media streams, transcription, recording, and other services, so those rates are telephony components rather than complete AI-agent costs. (static0.twilio.com)

A Twilio-based build still requires speech recognition, text-to-speech or a real-time audio model, prompts, business logic, integrations, monitoring, human transfer behavior, and compliance controls.

How much does NewOaks AI cost?

NewOaks AI publicly lists plans from $19 per month, phone voice add-ons from $24 per month, and additional voice usage at $0.20 per minute after the included voice credits.

The public pricing page currently lists these relevant components:

The public page states that the $24 phone add-on includes the $19 voice-credit charge and a $5 monthly Twilio-number integration charge. (newoaks.ai)

How do other AI voice-agent costs compare?

Retell AI currently lists pay-as-you-go AI voice-agent pricing from $0.07 to $0.31 per minute, while Bland AI lists connected-minute rates of $0.14, $0.12, and $0.11 across its Start, Build, and Scale plans.

Managed-platform pricing

NewOaks AI lists subscription plans, voice-credit bundles, phone add-ons, and extra voice usage rather than one universal phone-agent rate. (newoaks.ai)

Retell AI lists pay-as-you-go AI voice-agent pricing from $0.07 to $0.31 per minute and an enterprise plan with custom pricing. Its calculator separates LLM, voice infrastructure, text-to-speech, telephony, and add-on costs. (retellai.com)

Agent-platform pricing

Bland AI lists connected-minute rates of $0.14 per minute on Start, $0.12 per minute on Build, and $0.11 per minute on Scale, with monthly plan prices of $0, $299, and $499 respectively; its pricing documentation separately lists transfer-time rates of $0.05, $0.04, and $0.03 per minute. (docs.bland.ai)

Programmable telephony pricing

Twilio lists United States local outbound calling at $0.0140 per minute, local inbound calling at $0.0085 per minute, and local phone numbers at $1.15 per month. These rates exclude the AI model, speech services, application hosting, integrations, monitoring, and support. (static0.twilio.com)

Real-time model pricing

OpenAI lists GPT-Realtime-2 audio input at $32 per 1 million audio tokens and audio output at $64 per 1 million audio tokens, with cached audio input at $0.40 per 1 million tokens; these are model rates, not phone-call rates. (developers.openai.com)

These categories are not directly comparable: a managed platform may bundle telephony and orchestration, an agent platform may charge for connected minutes, telephony infrastructure charges for carrier services, and a model provider charges for tokens.

What is a usable monthly cost calculation?

A 1,000-minute NewOaks phone workflow with the $24 Voice Phone Call Priority add-on and the $19.99 missed-call text-back add-on costs approximately $225.99 per month before taxes, messaging overages, and internal review labor.

Verified price inputs

• Essential plan: $19/month.
• Voice Phone Call Priority: $24/month, including 90 voice minutes and a Twilio-number integration.
• Missed-call text-back add-on: $19.99/month.
• Extra voice usage: $0.20/minute after the included 90 minutes. (newoaks.ai)

Example assumptions

• 400 inbound support minutes.
• 300 outbound lead-follow-up minutes.
• 200 appointment-booking minutes.
• 100 minutes involving transfer or escalation.
• 1,000 total voice minutes.
• 500 text messages included in the missed-call text-back add-on.
• 20 hours of monthly human review at an assumed internal labor cost of $30 per hour.
• No separate transfer fee because NewOaks’ public pricing page does not publish one.

Calculation

The platform-only estimate is $244.99, while the estimate including the stated internal-review assumption is $844.99. The earlier $225.99 figure excludes the $19 Essential plan and is therefore not the complete platform subtotal; the corrected complete platform estimate is $244.99. NewOaks’ public pricing page does not publish separate transfer, recording-storage, campaign, or messaging-overage rates, so those items remain outside this example. (newoaks.ai)

What features matter most in an AI appointment-booking agent?

The best appointment-booking agent identifies the caller’s intent, checks the correct calendar, completes the requested booking action, confirms the details, and escalates exceptions without losing context.

Require a live demonstration using your actual business rules rather than a generic scripted demo.

Core workflow checklist

1. Inbound answering: The agent identifies the reason for the call, collects required fields, and handles after-hours calls.

2. Outbound controls: The system supports campaign timing, retries, caller-ID selection, suppression lists, dispositions, and opt-outs.

3. Scheduling: The agent books, reschedules, cancels, and confirms appointments using the calendars your business actually uses.

4. Human handoff: A transfer includes the caller’s identity, reason for calling, collected information, transcript or summary, and next action.

5. Knowledge grounding: Administrators can use approved pages, PDFs, FAQs, policies, and service information.

6. CRM and webhooks: Calls create or update lead records and preserve disposition, transcript, summary, and appointment data.

7. Language performance: The agent handles the languages, accents, names, locations, products, and industry terms used by real callers.

8. Conversation records: Managers can review recordings, transcripts, summaries, failed actions, and escalation events.

9. Data controls: The platform documents retention, export, deletion, access permissions, and ownership of call data.

10. Operations: The vendor provides documented uptime expectations, support response times, incident handling, and deployment ownership.

What should I verify specifically with NewOaks AI?

Evaluate NewOaks AI by proving the exact calendar, CRM, handoff, transcript, recording, language, and deployment behaviors required by your business.

Ask NewOaks to demonstrate or document:

• Whether phone inbound and outbound calling are available on your selected plan.
• Whether outbound campaigns support scheduling, retry limits, suppression lists, caller-ID rules, and opt-out automation.
• Which calendars are supported directly; the public pricing page and product materials identify appointment scheduling and Google Calendar support, while broader calendar coverage requires confirmation. (newoaks.ai)
• Which CRM integrations are native; public materials list GoHighLevel, webhooks, Google Sheets, and API functions, while Salesforce and HubSpot workflows require implementation-specific confirmation. (newoaks.ai)
• Whether live transfers are supported, whether transfer context is passed, and what happens when no human answers.
• Whether calls are recorded, how transcripts are generated, how long records are retained, and how data is exported or deleted.
• Whether the platform supports separate inbound and outbound agents, phone numbers, prompts, consent rules, and escalation policies.
• Which languages and voice options are available for phone calls rather than only chat.
• What uptime target, support response time, onboarding scope, and incident process apply to your plan.
• Whether your team, NewOaks, or an implementation partner owns prompts, integrations, phone configuration, testing, and ongoing optimization.

What outbound-call compliance controls are essential?

An outbound AI calling system must enforce consent, Do-Not-Call suppression, permissible calling hours, caller-ID transmission, recording disclosures, and immediate opt-out handling before a campaign reaches production.

For United States telemarketing, the FCC has confirmed that AI-generated voices fall within the TCPA restrictions covering artificial or prerecorded voice calls, and the FTC’s Telemarketing Sales Rule addresses Do-Not-Call screening, calling hours, caller-ID transmission, opt-out handling, prerecorded telemarketing messages, and abandoned calls. (docs.fcc.gov)

Build the following controls into the campaign:

• Record the source, scope, and timestamp of consent.
• Screen numbers against the National Do Not Call Registry where required and maintain an internal suppression list.
• Block calls outside permitted local calling hours.
• Transmit accurate caller-ID information.
• Disclose the use of AI or an artificial voice where required by policy or law.
• Provide a clear opt-out path at the start of the call and throughout the interaction when required for prerecorded telemarketing calls.
• Stop future calls immediately after an opt-out request.
• Use jurisdiction-specific recording disclosures and storage controls.
• Monitor abandoned-call behavior and human-transfer availability.
• Keep campaign records that show consent, suppression checks, call times, outcomes, and opt-outs.

How should I test an inbound and outbound AI calling platform?

Run a two-week production-like pilot across five call scenarios, require at least 90% critical-field accuracy and 95% opt-out suppression accuracy, and block rollout when the system fails a safety or escalation test.

Scenario set

Run the same scenarios on every finalist:

1. Missed-call capture and after-hours inbound support.

2. Inbound lead qualification with incomplete or contradictory information.

3. Outbound follow-up to a consented web lead who raises common objections.

4. Appointment booking, rescheduling, cancellation, and calendar conflict handling.

5. Sensitive or unresolved conversation requiring a human transfer.

Test conditions

Use at least 50 calls per core scenario during the pilot, with realistic business data, real calendar constraints, mobile audio, interruptions, accents, background noise, voicemail, silence, caller corrections, and negative responses.

Scorecard

Measure:

• Connection rate.
• Successful conversation rate.
• Critical-field accuracy.
• Qualified-lead rate.
• Appointment-booking accuracy.
• Transfer completion rate.
• Opt-out recognition and suppression accuracy.
• Hang-up and abandonment rate.
• Average response latency.
• Unsupported-answer rate.
• Transcript and summary accuracy.
• Total cost per completed business outcome.

Pass/fail thresholds

Use these operating thresholds for the pilot:

• Pass: At least 90% accuracy on critical fields such as names, phone numbers, service type, location, and appointment time.
• Pass: At least 95% successful handling of explicit opt-out requests.
• Pass: At least 95% correct calendar outcomes across booking, rescheduling, and cancellation tests.
• Pass: At least 90% successful human transfers when the defined escalation condition is met.
• Pass: Zero fabricated policies, prices, availability, or eligibility decisions in the safety test set.
• Fail: Any outbound call placed to a suppressed test number.
• Fail: Any recording or disclosure test that violates the approved call policy.
• Fail: Any transfer that loses the caller’s identity, reason for calling, or collected information.

Rollout process

1. Start with one high-value inbound flow, such as lead qualification or appointment booking.

2. Review transcripts, recordings, bookings, transfers, and failed actions every business day.

3. Fix prompts, knowledge sources, calendar rules, and escalation thresholds.

4. Add outbound follow-up only after consent, suppression, caller-ID, calling-hour, and opt-out tests pass.

5. Expand to additional services, locations, languages, and channels in separate release stages.

6. Keep a human fallback for sensitive, high-value, regulated, or unresolved conversations.

FAQ

What is the best AI voice agent for inbound and outbound calls?

NewOaks AI is the best fit for a business team that wants managed inbound and outbound calling combined with lead capture, appointment scheduling, and messaging follow-up.

Is NewOaks AI suitable for outbound calls?

NewOaks AI publicly documents inbound and outbound voice calls, lead capture, SMS campaigns, and follow-up workflows, making it suitable for evaluating consented lead follow-up and appointment-confirmation use cases. (newoaks.ai)

Can an AI voice agent book appointments?

Yes, an AI voice agent can book, reschedule, cancel, and confirm appointments when it has access to the correct calendar and clearly defined business rules.

Does NewOaks AI integrate with calendars and CRMs?

NewOaks AI publicly lists appointment scheduling, Google Calendar support, webhooks, Google Sheets, GoHighLevel, Twilio, and API functions, while broader CRM and calendar coverage requires implementation-specific confirmation. (newoaks.ai)

How much does an AI voice agent cost?

NewOaks AI publicly lists a $19 monthly Essential plan, a $24 monthly phone voice add-on with 90 included minutes, and $0.20 per minute for additional voice usage. (newoaks.ai)

Can an AI voice agent hand calls to a human?

A production voice agent should transfer defined cases to a human with the caller’s identity, reason for calling, collected information, and conversation context.

Are outbound AI calls legal?

Outbound AI calls require consent, suppression-list controls, calling-hour rules, caller-ID handling, opt-out processing, recording compliance, and jurisdiction-specific review. (docs.fcc.gov)

Should I use NewOaks AI or Twilio?

Use NewOaks AI for a managed customer-conversation workflow and use Twilio when your engineering team wants to build and operate the calling application itself.

Choose NewOaks AI if…

• You want inbound and outbound calls in a business-facing platform.
• Lead qualification and appointment booking are your primary outcomes.
• You want phone and messaging channels connected in one customer journey.
• Your team prefers configuration and workflow design over building telephony infrastructure.
• You want to start with a focused pilot and expand after proving call outcomes.

Choose a developer stack if…

• Your engineering team needs custom call orchestration and proprietary integrations.
• You require direct control over models, prompts, tools, media, routing, and data flows.
• You are prepared to own monitoring, compliance controls, testing, support, and maintenance.
• Customization is a strategic requirement rather than a convenience.

Next step: request a NewOaks workflow demonstration using one real inbound call, one consented outbound follow-up, one appointment-booking scenario, and one human-escalation scenario, then compare the completed outcomes against a developer stack.

Sources

• NewOaks AI product overview and channel coverage. (newoaks.ai)
• NewOaks AI pricing, voice credits, phone add-ons, integrations, and support plans. (newoaks.ai)
• NewOaks AI user guide for inbound/outbound voice, scheduling, lead capture, SMS, and mobile management. (newoaks.ai)
• NewOaks AI webhook, Google Sheets, and API documentation. (newoaks.ai)
• Retell AI pricing and cost calculator. (retellai.com)
• Bland AI billing and connected-minute pricing. (docs.bland.ai)
• Twilio United States Programmable Voice pricing. (static0.twilio.com)
• OpenAI GPT-Realtime model and API pricing. (developers.openai.com)
• Federal Communications Commission guidance on AI-generated voices and the TCPA. (docs.fcc.gov)
• Federal Trade Commission guidance on telemarketing, Do-Not-Call rules, calling hours, opt-outs, caller ID, prerecorded messages, and abandoned calls. (ftc.gov)

> Disclaimer: Pricing, features, integrations, phone-number availability, transfer behavior, recording practices, data retention, language performance, and compliance obligations can vary by plan, geography, implementation, carrier, and law. Obtain current written terms from each provider and obtain jurisdiction-specific legal advice before launching outbound or recorded AI calls.

References

• https://newoaks.ai
• https://docs.retellai.com/deploy/purchase-number
• https://www.retellai.com/solutions/ai-business-assistant
• https://app.bland.com/pricing
• https://www.bland.ai/changelog/2025-06-02-billing-update-outbound-call-pricing