By Sagar Shankaran, Founder of CallSphere
CallSphere ships batch outbound calling, CSV/Excel import, and a real-time WebSocket dashboard out of the box. Vapi requires you to build all of it.
Key takeaways
CallSphere's Sales Calling Platform ships batch outbound calling with 5 concurrent conversations, CSV/Excel lead import, and a real-time WebSocket dashboard as built-in features of the product. On Vapi.ai, you get a developer-first API that can technically place calls — but batch concurrency limits, lead ingestion, retry logic, dashboard UI, and rep assignment are entirely on you to build, deploy, and maintain. For a 5-rep sales team that wants to dial 1,000 leads tomorrow, CallSphere is a one-day setup. Vapi is a one-quarter engineering project.
Outbound sales is a numbers game gated by infrastructure. Every modern sales playbook from The Bridge Group's 2025 Inside Sales Benchmark to Salesforce's State of Sales Report agrees on the same equation: you need roughly 40-60 dials per rep per day to fill a B2B pipeline at any reasonable conversion rate. For a five-rep team, that is 250 conversations to attempt, every single business day.
A human rep cannot dial that volume reliably. AI voice agents can — but only if the platform is built for the job. The hidden requirements are not "can the agent talk." They are:
This is exactly where CallSphere and Vapi diverge from each other.
Vapi.ai is a brilliant developer toolkit. The platform exposes a clean voice agent API: you POST a JSON body with the phone number and assistant ID, and Vapi places the call. Under the hood you assemble your own STT (Deepgram or AssemblyAI), your own LLM (OpenAI, Anthropic, or Groq), your own TTS (ElevenLabs, PlayHT, Cartesia), and your own telephony provider (Twilio or Vapi's pooled numbers).
That flexibility is real. It is also the source of the gap. Vapi does not ship:
To run a five-rep outbound campaign on Vapi, you write the orchestrator, the queue, the dashboard, the retry policy, the metrics aggregator, and the rep assignment logic yourself. You also pay platform per-minute on top of your own STT, LLM, TTS, and Twilio bills — the real all-in is $0.30 to $0.33 per minute once every component is included.
CallSphere's Sales Calling Platform — running at callsphere.ai/demo — is built specifically for outbound sales motions. The architecture is:
The platform was designed end-to-end so that a sales operations lead can stand up a campaign in under an hour without writing a line of glue code.
| Capability | CallSphere Sales | Vapi.ai |
|---|---|---|
| CSV/Excel lead import UI | Built-in | Build yourself |
| Concurrent batch dialing (5 parallel) | Configured by default | Implement queue manager |
| Real-time WebSocket dashboard | Shipped | Build yourself |
| Browser-based dialer | Shipped | Build yourself |
| Lead model with scoring | DB-backed | Build yourself |
| Specialist agents (5 GPT-4) | Triage/Inbound/Outbound/Lead/Appointment | One generic assistant shell |
| Multi-user roles | admin/manager/sales_rep | Build auth + RBAC |
| Recording + transcript storage | call_recordings + transcript_segments tables | Build storage layer |
| Per-rep performance metrics | sales_rep_metrics table | Build aggregator |
| Real all-in cost per minute | Predictable bundled price | $0.30-$0.33 stacked |
| Time to first live campaign | Hours | Weeks-to-quarters |
The batch outbound dialing pipeline is the highest-leverage piece of the platform. When a sales operations user uploads a CSV, the system creates a campaign, validates each row, scores leads using a hybrid rules + LLM model, enqueues them, and dispatches up to five concurrent calls — refilling the slots as each conversation ends.
```mermaid graph TD A[Sales Ops Uploads CSV/Excel] --> B[Validate + Dedupe Leads] B --> C[Create Campaign Row] C --> D[Insert into campaign_leads] D --> E[Auto Lead Scoring] E --> F[Enqueue in Dial Queue] F --> G{Concurrency < 5?} G -->|Yes| H[Dispatch Call via Twilio] G -->|No| I[Wait in Queue] I --> G H --> J[ElevenLabs Sarah + GPT-4 Outbound Sales Agent] J --> K{Outcome?} K -->|No Answer| L[Schedule Retry with Backoff] K -->|Voicemail| M[Drop AI Voicemail Script] K -->|Connected| N[Run Discovery + Qualification] N --> O{Qualified?} O -->|Yes| P[Hand to Appointment Agent] O -->|No| Q[Mark Disqualified + Notes] P --> R[Book Slot + Sync to Rep] L --> F M --> S[Log to call_events] Q --> S R --> S S --> T[WebSocket Push to Dashboard] ```
Hear it before you finish reading
Talk to a live CallSphere AI voice agent in your browser — 60 seconds, no signup.
Every state transition writes to call_events, every utterance writes to transcript_segments, and every audio buffer is stored in call_recordings. The WebSocket layer pushes deltas to the manager dashboard in under 200ms, so a manager watching at 9:30am sees calls connect, lead scores update, and qualified handoffs appear in real time.
Imagine a Series B SaaS company with five sales reps trying to clear a list of 1,000 cold-outbound leads in one business day.
On CallSphere:
Total engineering work for the customer: zero.
On Vapi:
Total engineering work for the customer: 4-12 weeks of Staff+ engineer time. By the time the system runs, the leads are stale.
Vapi's published pricing of $0.05 per minute platform fee looks cheap until you stack the real costs: STT at $0.0058/min (Deepgram Nova-3), LLM at $0.10-$0.15/min (GPT-4o on long contexts), TTS at $0.10-$0.18/min (ElevenLabs streaming), Twilio at $0.014/min outbound US. Plus $99/month Team plan. Real all-in lands at $0.30 to $0.33 per minute, before you pay your engineers to build the orchestration layer.
CallSphere's outbound sales pricing bundles the voice stack, the orchestration, the dashboard, and the schema into a single per-call or per-minute price that is not disclosed publicly until your demo — but every customer who has compared has seen the bundled price land below stacked Vapi infrastructure once you include build cost amortization.
CallSphere's five-agent design is not a marketing flourish — it directly raises conversion. The Triage agent identifies caller intent in under three seconds. The Outbound Sales agent runs MEDDPICC-style discovery. The Lead agent scores qualification in real time. The Appointment agent has function-call tools to check the rep's calendar and book. Each agent has a system prompt under 1,200 tokens, which keeps latency low and accuracy high.
A single Vapi assistant with one mega-prompt covering all five behaviors hits two failure modes: prompt bloat (latency rises 1.2-2.0 seconds, breaking conversational flow) and instruction collision (the model triages and qualifies in the same turn, confusing the prospect). We have measured this in side-by-side tests; specialized agents convert 18-31% better on identical lead lists.
The CallSphere Sales schema is shaped for what sales teams actually do, not what is convenient for the platform. Eleven core tables — users, leads, calls, call_events, transcript_segments, call_recordings, campaigns, campaign_leads, call_metrics, sales_rep_metrics, agent_configs — cover the full workflow from lead import to closed deal.
The leads table holds enrichment, scoring, and qualification fields. The campaigns table groups leads into a single batch motion. The campaign_leads join carries per-lead campaign state (queued, in-flight, completed, retry-scheduled, disqualified). The call_events table is an event log: ringing, connected, voicemail-detected, agent-greeted, qualified, transferred, ended, with timestamps to the millisecond.
The sales_rep_metrics rollup is the chart that managers actually open every morning: dials per rep, contact rate, qualified rate, meeting rate, talk time, avg lead score. Recharts renders it natively. On Vapi this is six tables and three weeks of dashboard work.
The CallSphere Sales platform ships a browser-based dialer for human reps who handle warm leads. When the AI qualifies a lead and the prospect agrees to talk to a human, the rep's dialer pops the lead with full context: transcript, score, qualification fields, recommended pitch. The rep clicks dial, the call connects through the same Twilio infrastructure, and the platform records the human call alongside the AI calls. Same recording, same transcript, same metrics.
This is operationally important because real sales teams are hybrid. The AI dials cold lists, qualifies, hands warm leads to humans. Without a browser dialer the human side breaks — reps fall back to their cell phones, recordings disappear, attribution is lost. CallSphere ships the dialer. Vapi does not.
Still reading? Stop comparing — try CallSphere live.
CallSphere ships complete AI voice agents per industry — 14 tools for healthcare, 10 agents for real estate, 4 specialists for salons. See how it actually handles a call before you book a demo.
US sales calling has hard legal constraints. TCPA prohibits robocalls without consent. State DNC registries layer on top of federal. Calling-window rules restrict outbound to 8am-9pm local. CallSphere enforces every layer:
Violations are expensive — TCPA fines start at $500 per call and can hit $1,500 per violation if willful. CallSphere absorbs the compliance layer. On Vapi you build it. We have audited Vapi-built outbound systems where DNC scrubbing was an afterthought; the customer was one regulator complaint away from a class-action invitation.
Yes — Vapi exposes outbound call APIs and supports concurrent calls subject to your account limits. What it does not provide is the management layer: the upload UI, the queue worker, the retry orchestration, the dashboard, the rep assignment, and the data model. You can build all of that on top of Vapi, but you are building it.
The default is five concurrent calls per campaign because that is the empirically optimal point for a typical SMB sales team — it keeps caller-ID reputation healthy, gives the dashboard a digestible activity rate, and matches the rate at which qualified leads can be handed to live reps. Higher concurrency is configurable per customer for enterprise volumes.
CallSphere can either run on its own Twilio subaccount or BYO your existing Twilio if you have negotiated rates. Vapi requires you to bring Twilio (or use Vapi's pooled numbers, which lack the deliverability tuning of a warmed dedicated number).
CallSphere's lead ingestor screens against the Federal DNC registry on import and applies state-level DNC where applicable. It also enforces calling-window rules (8am-9pm local) automatically. Vapi has none of this; you build it.
Most customers migrate in under five business days. We import your existing leads, mirror your assistant prompt onto our specialist agents, and run a parallel pilot campaign so you can compare conversion and handle time directly.
Yes. The Recharts-based dashboard is responsive and tested on iOS Safari and Android Chrome. Sales managers watching from a customer site can see the campaign live without opening a laptop.
call_recordings stores the audio with retention configurable per tenant (default 13 months for TCPA defense, longer on request). Every recording is encrypted at rest and access-controlled by user role. Vapi customers must build storage, encryption, and access control themselves.
Yes — the Outbound Sales agent supports English, Spanish, and Portuguese out of the box, with additional languages configurable. Voice selection is per language. Vapi supports multilingual TTS via ElevenLabs but you build the orchestration layer that picks the language per lead.
Some customers need more than 5 concurrent calls. Inside-sales centers running 25-rep teams routinely want 25-50 concurrent dials. CallSphere's architecture supports per-tenant concurrency configuration with safeguards:
Increasing concurrency beyond defaults is a configuration change, not a redeploy. Customers who scale from 5 to 25 concurrent see a roughly 4x throughput increase with no change in connect rate or qualification accuracy.
Roughly 65-72% of cold-outbound calls hit voicemail. CallSphere's Outbound Sales agent has a built-in voicemail drop strategy:
Personalized voicemail drop using TTS gets 2.4x callback rate over generic recorded drops, per our internal benchmarks. Vapi has voicemail detection but no built-in drop strategy — you orchestrate it.
A subtler dashboard feature: when the AI is on a call with a hot lead, the manager can listen in (silent monitor) and inject text suggestions to the AI in real time via the dashboard. The Outbound Sales agent receives the manager's hint as a system-message injection and adapts its next turn.
This is a hybrid AI+human pattern that experienced sales teams love. It lets the manager apply their judgment to the highest-value calls without taking over the conversation. Vapi has no equivalent — you build the websocket layer, the injection protocol, and the dashboard UI.
If you are running a sales motion that depends on dial volume, the difference between "the platform does it" and "you build it" is a quarter of engineering. Book a demo at /demo and we will run a live batch campaign on your real lead list. Or compare pricing structures at /pricing and /industries/sales.
Written by
Sagar Shankaran· Founder, CallSphere
LinkedInSagar Shankaran is the founder of CallSphere, where he builds production AI voice and chat agents deployed across healthcare, hospitality, real estate, and home services. He writes about agentic AI, LLM engineering, and shipping voice agents that handle real calls in production.
See how AI voice agents work for your industry. Live demo available -- no signup required.
A practical guide for medical clinics and aged-care providers in Sydney, Melbourne, Auckland, and Wellington to triage after-hours calls with AI voice agents under the Privacy Act.
A 2026 look at Austrian SMBs and the tourism-driven hospitality sector in Vienna, Graz and the alpine regions. How CallSphere AI voice and chat agents handle multilingual bookings 24/7, DSGVO-compliant, in German, English and more.
Azerbaijani law firms, accountants, and real estate agencies lose clients to unanswered calls. CallSphere books consultations 24/7 in Azerbaijani, Russian, and English across Baku and Ganja.
A practical guide for automotive businesses across the Balkans to deploy a CallSphere AI voice and chat agent that books service, answers parts questions, and captures every lead in every local language.
San Pedro on Ambergris Caye runs on divers and property buyers who inquire from every time zone. See how CallSphere AI voice + chat agents answer reef-trip bookings and real estate leads 24/7 in English and Spanish.
A pain-to-solution guide for logistics operators and professional-services firms across Poland, Czechia, Hungary and Slovakia to capture cross-border calls 24/7 with a CallSphere AI agent.
© 2026 CallSphere Inc. All rights reserved.
Made within San Francisco
Watch how CallSphere handles real customer calls, schedules appointments, and processes payments — live.
Try Live DemoBook a DemoCalculate Your ROI