Call centers do not buy VoIP for cost savings anymore. The savings are real, but they are not the reason. The reason is that a hosted platform gives you queue logic, recording, real-time dashboards, and AI handling that an on-prem PBX cannot match without a six-figure project.
The features that move metrics
Queues with skill-based routing
A queue is not just a hold line. Our queues route by agent skill, priority tier, time in queue, and last agent contacted. If a caller has been waiting more than 90 seconds, they jump priority. If they are a known high-value customer (matched via CRM), they go to a dedicated queue. This is configuration, not custom development.
The routing logic supports fallback chains: try the primary agent group, then the secondary, then a supervisor queue, then voicemail. Each step has its own ring timeout and treatment. Build the chain once, adjust it as your staffing model changes.
Whisper, barge, and monitor
Supervisors listen in silently (monitor), coach the agent without the caller hearing (whisper), or join the call as a third party (barge). For training new agents, whisper alone shaves weeks off ramp time. The supervisor view lists every live call in the center with one-click access to each of those modes.
Call recording with searchable transcription
Every call recorded, retained per your policy, searchable by keyword via transcription. Useful for QA, compliance, and the occasional dispute where the customer is sure they were promised something they were not. Transcription is a paid add-on, billed per minute of recording.
Recording works at the trunk level (every call) or at the agent level (per-agent rules). Most centers record everything and then sample for QA, which is the cheapest path to defensible call documentation.
Wallboards and real-time dashboards
Calls in queue, average wait, service level, agents available, calls handled today. On a TV on the wall or on the manager's laptop. Drives the right behavior because nobody wants to be the red number.
IVR and self-service
Routing menus that pull from your CRM (account number lookups, order status, balance) so callers who do not need an agent never wait for one. We build these on our Phone Service platform with no custom dev work for most flows.
SMS and two-way text
Inbound SMS to your DIDs routes into the agent client. Outbound SMS from the same client. Templates for common responses. Auto-reply on missed calls ("sorry we missed you, reply YES to schedule a callback"). A2P 10DLC registration handled by us; carriers require brand and campaign registration on all business SMS now, and we walk you through it as part of onboarding.
Where the AI Receptionist fits
For small and mid-sized call centers, the math has shifted. A traditional agent answering basic questions ("are you open Saturday," "can I book an appointment," "what is your address") costs $20-$30 an hour fully loaded. The AI Receptionist at $199/mo Pro or $299/mo Enterprise handles thousands of those calls without the queue.
It does not replace skilled agents. It deflects the calls that should never have hit a queue in the first place, which is usually 30-50% of inbound volume in a service-business call center. The agents you keep talk to the customers who actually need them.
What the AI Receptionist does well
- Books appointments to your calendar in real time.
- Captures lead info into your CRM with field validation.
- Answers FAQ-style questions in your business voice.
- Handles after-hours and overflow without a queue.
- Routes complex calls to a human queue with context already gathered.
- Sends a follow-up SMS to the caller after the call.
What it does not do well
- Complex troubleshooting with multiple decision points.
- Empathy-heavy calls (escalations, complaints, sensitive matters).
- Multi-turn negotiation (pricing, scope, contract terms).
- Anything that requires the caller to feel heard by a human.
For those, route the call into the queue with full context attached. The AI is a triage layer, not a replacement.
What it costs per seat
- Per-Minute plan: $15/user/mo plus $0.025/min outbound. Good for inbound-heavy queues with low outbound volume.
- All-Inclusive plan: $32/user/mo with unlimited US and Canada. Standard for outbound call centers and any seat that makes more than ~700 minutes a month.
- SIP Trunking for centers running their own dialer: $15/channel/mo, $0.015/min outbound US-CA, $0.005/min inbound US-CA. Details on SIP Trunking.
- AI Receptionist: $99/$199/$299 per office. HIPAA add-on $49/mo where required.
- Per-number porting: $15 per number, both directions.
- E911 misdial: $150 non-refundable per misdial from an unauthorized device.
Picking between SIP trunks and seats
If you already own a contact center platform (Five9, Talkdesk, Genesys, whatever) and you want lower-cost trunks, SIP trunking is the right buy. You pay per channel and per minute. If you do not have a contact center platform and you want one, our hosted seats include the queue logic and the agent client. The decision is rarely about cost; it is about whether you have an existing investment in a platform that works for you.
A real 30-agent build
A typical mid-sized inbound center handling 8,000 calls a month with a 70/30 inbound/outbound split:
- 30 All-Inclusive seats at $32 = $960/mo.
- 3 supervisor seats with monitor/whisper/barge enabled (no upcharge).
- One wallboard TV in the room, driven from a portal URL.
- Call recording on every seat, retained 90 days, transcription enabled on QA-flagged calls only (paid add-on, controlled to keep costs reasonable).
- AI Receptionist Enterprise at $299/mo handling tier-zero deflection and after-hours, routing the calls that need a human into the queue with caller context already gathered.
- CRM integration to the customer's existing platform (no upcharge for the common ones).
Total monthly: roughly $1,259 plus transcription usage. Compare to a comparable on-prem build with a five-figure capital outlay and recurring maintenance, and the case writes itself.
What that build looks like for vertical-specific centers
An HVAC dispatch center running ServiceTitan: same seat count, but the IVR routes by zip code into geographic teams, the screen-pop shows equipment history, and dispatch sees presence for techs in the field. A real estate buyer lead center running Follow Up Boss: same seat count, but the queue routes by agent round-robin with presence-aware skipping, and the AI Receptionist qualifies leads before they reach an agent. A law firm intake center running Clio: same seat count, but recording is enforced, time entries pre-populate, and conflict checks happen at intake.
Metrics this stack actually moves
- Service level (% answered within target). Skills routing and AI deflection push this up by routing calls to the right agent faster and removing calls that never needed an agent.
- Abandon rate. Call-back option ("press 1 and we will call you back, keeping your place") is the single biggest abandon-rate fix.
- Average handle time. CRM screen-pops at the start of the call save 20-30 seconds per call by removing the "can I have your account number" exchange.
- First-call resolution. Skills routing puts the right person on the call.
- Agent ramp time. Whisper coaching reduces the weeks of side-by-side training new hires need.
- Cost per call. AI deflection of routine calls drops the human-handled call mix to higher-value interactions.
The honest limitations
VoIP needs internet that does not blink. For call center seats, we strongly recommend a wired connection per seat, QoS configured at the switch, and a backup circuit. A 50-seat call center on a single residential-grade fiber line will sound bad some afternoons. This is solvable for a few hundred dollars a month in circuit redundancy. It is not solvable by hoping.
Cybersecurity is the other one. SIP credentials get scanned for and brute-forced constantly. We rate-limit, geo-fence, and lock down by default. If your IT team wants to manage the trunks themselves, talk to us about the security baseline before you do.
What we do not pretend to be
- We are not a workforce management platform. If you need shift scheduling and forecasting at scale, integrate with a WFM tool like NICE or Verint.
- We are not an outbound dialer. We provide the trunks and seats; predictive and progressive dialers come from specialist platforms.
- We are not a full omnichannel ticketing platform. Voice and SMS are our strength; email and ticket handling integrate to the helpdesk you already use.
- We do not pretend AI replaces senior agents. It deflects the calls that should never have reached one.
- We are not a CCaaS platform. If you need Salesforce-grade contact-center orchestration with bespoke workflow logic, we will tell you so up front.
Common mistakes when standing up a call center on VoIP
- Tagging every agent with every skill. Defeats the purpose. Be honest about who handles what.
- Not testing the call-back feature. It is the highest-leverage queue feature and most centers leave it off.
- Recording without a written policy. States vary on consent. Get the policy in writing before you turn recording on.
- Single circuit, no failover. The day the fiber gets cut is the day you wish you had a backup LTE circuit.
- No wallboard. If the team cannot see the queue depth, they treat slow afternoons the same as busy ones.
- Hiring more agents instead of fixing the IVR. If 40% of calls do not need a human, hiring is the wrong fix.
- Ignoring after-call work time. Agents who have to retype call notes are slower. CRM logging cuts ACW.
What to ask any call center VoIP provider
- Are you the operator or a reseller of someone else's platform?
- What is your published per-channel and per-minute SIP trunking price? (Ours: $15/channel, $0.015 outbound, $0.005 inbound US-CA.)
- Is call recording included? Is transcription? (Recording yes, transcription paid add-on.)
- Do you charge separately for whisper, barge, and monitor? (No.)
- What is your E911 misdial fee? (Ours: $150, non-refundable.)
- What happens to my recordings and call data if I leave?
- How do you handle A2P 10DLC registration for SMS? (We do it.)
- What is your sustained QPS for screen-pops against a customer-provided CRM API?
What we tell customers about ramp-up
The first 30 days after go-live are not the steady state. Service levels dip in week one as agents learn the new screen-pop pattern. Queue tuning is the work of weeks two and three; the IVR drop-off report tells us where to adjust. By week four, the metrics are usually better than the pre-migration baseline. We schedule weekly tuning calls during ramp-up so the system gets fixed at the speed real operating data comes in, not at the speed of quarterly business reviews. Customers who skip the tuning phase get a working call center; customers who do the tuning get a measurably better one.
Reporting that drives decisions
The reports we look at most often are the ones that tell a supervisor what to do tomorrow. Service level by hour of day shows when staffing is short. Abandon rate by queue shows which queue is bleeding. Average handle time by agent shows who is fast and who is slow. First-call resolution by skill shows where the skills matrix needs work. Reports that do not lead to a specific action are reports that do not get pulled twice. The portal exports CSV for the customers who feed Power BI or Looker; for most customers, the built-in reports are enough.
How call centers fit with the rest of your operation
Most call centers are not standalone businesses. They are the inbound layer of a larger operation: dispatch for field service, intake for law firms, appointment booking for healthcare practices, tenant support for property management, or lead handling for sales teams. The features above only deliver value when they are tied into the operation they serve. We spend more time on the integration and the workflow than on the queue logic, because the queue logic is the easy part.
Where to start
If you are running 10 or more agents, the gap between what a hosted platform does and what an old PBX does is wide enough that an evaluation makes sense even mid-contract. Send us your current ACD setup and we will tell you what would change. For the AI-augmented front line, start at AI Receptionist. Full pricing detail at pricing. CRM-specific integration patterns at integrations. Comparisons with the providers you have probably already evaluated at vs.