Bland

Where does your data go?

Bland's own infrastructure — the strongest data-architecture story of any voice platform:

• Bland owns its stack. Not a wrapper around third-party APIs: the models are custom-built for phone calls, the telephony and AI infrastructure is in-house, and the platform's claim is "your data never goes through third parties. No sudden changes to your model, pricing, or terms of service." That's a materially different architecture from every other voice listing — and it removes the subprocessor-chain risk the other reviews document.
• Compliance is in the base rate, not a tier: SOC 2 Type I and II, HIPAA (agreements on request), PCI DSS, GDPR — all included in standard pricing. Compare Synthflow (HIPAA is Enterprise-only) and Vapi (zero data retention is a $1,000/month add-on).
• Self-hosted and on-premises deployments are available for the most sensitive workloads — the only voice platform reviewed offering that.
• The honest gap, same as Retell: no EU-region hosting statement. GDPR is certified; residency is not stated.

What happens when it's wrong?

Bland publishes no call-accuracy benchmarks, but its architecture addresses the failure surface structurally:

• Pathways, not freeform chat: conversations run on visual, deterministic flows rather than open-ended improvisation. The trade — less flexibility than Retell's approach — is also the guardrail: predictable call behavior is a compliance feature, not a limitation, in regulated industries.
• The scale evidence is the stress test: infrastructure documented at up to a million simultaneous calls; Fortune 500 banks, insurance carriers, and healthcare systems as customers. Claims backed by enterprise deployment rather than benchmarks.
• The failure mode reviewers flag: inbound is supported but not its strongest use case — it's an outbound-first platform. For reception-style flows, Retell's conversational quality scores higher (8.6 vs 9.4 on independent voice-quality testing).
• Compliance is designed in: debt-collection flows built with FDCPA and TCPA handling, PCI DSS for payment calls — the regulated verticals are first-class use cases, not afterthoughts.

What's it actually bad at?

• Inbound reception is not its strength. Independent testing: 8.6/10 voice quality (vs Retell's 9.4) and 850ms median response times (vs Retell's 680ms). Competent, not category-leading.
• The plan-fee structure needs arithmetic. Start is $0/month with $0.14/min; Build is $299/month with $0.12/min; Scale is $499/month with $0.11/min. The crossover points: Build beats Start around 15,000 minutes/month; Scale beats Build near 20,000. Below those thresholds, the free Start tier is the rational choice.
• Transfer-time billing: calls that transfer to a human bill extra on Bland-provided numbers — bringing your own phone setup removes it. A hidden cost for receptionist use cases specifically.
• Developer-shaped despite the visual builder. More approachable than raw API platforms, weaker than Synthflow's guided experience for non-technical operators.
• Concurrency limits on self-serve tiers: 10/50/100 calls across the tiers — unlimited waits for Enterprise.

What does it cost really?

The comparison that makes Bland's case: at 10,000 calls/month, the platform stack costs vary 4–5x across the category — Retell ~$2,800/month all-in, Bland ~$4,900/month, and Vapi's "cheapest" $0.05/min platform fee becomes $10,000–13,200/month once every passed-through vendor is counted. Bland's $0.11–0.14 is the only rate in the category you can take at face value. The compliance math: bundled HIPAA saves a regulated business $12,000–50,000/year versus competitors' add-ons and tier gates.

What does it replace — and what do you still do?

It replaces the outbound call workload: appointment reminders, reactivation campaigns, lead qualification, claims intake, collections — at a volume where per-call economics beat human staff. Pathways makes the flows deterministic and repeatable. You still own: designing the flows (a real design job), compliance posture for your jurisdiction, and the integration work for your CRM stack.

What the ecosystem offers

• Pathways — a visual conversation-flow builder for complex multi-step calls (claims intake, loan qualification, medical intake) without code; deterministic flows instead of open-ended improvisation.
• In-house models, custom-built for phone calls — sub-second latency across 40+ languages, 23 real-time translation languages.
• Contact-center telephony integrations — Genesys, Five9, NICE CXone, Amazon Connect, Talkdesk — for businesses replacing rather than adding phone infrastructure.
• CRM and calendar integrations: HubSpot, Cal.com, SMS confirmations, warm transfers with caller context.
• A developer integration for Claude Code and Cursor — the API-first team's path into Bland.
• Compliance as architecture: debt-collection flows with FDCPA/TCPA handling, PCI DSS for payment calls, HIPAA for patient scheduling — the regulated use cases are documented first-class.

If I stop paying, what do I lose?

• Agents and flows: built in Bland's builder, not portable in any documented way — the same lock-in shape as Synthflow and Retell. GitHub export exists for versioning, not migration.
• Numbers: bring your own phone setup and your numbers are portable — same rule as the other voice listings.
• The lock-in reality: moderate. What's genuinely different is the platform architecture: Bland owning its stack means no subprocessor surprises, but also no third-party escape hatches.

Who is this NOT for?

Simple inbound receptionists — Retell's conversational quality wins that job, and turnkey receptionists win it for non-technical owners. Anyone below 15,000 minutes/month who'd be paying platform fees for discounts they don't hit — Start's $0/month with $0.14/min beats Build below that threshold. And non-technical owners who want the absolute simplest path — the turnkey products do that better.