Bland AI
Also known as: Bland, Bland Speech, Norm
Enterprise voice AI platform for high-volume and regulated phone workflows, running its own telephony, speech, language and text-to-speech stack behind a node-based pathway builder, with SMS, RCS and iMessage on the same agents.
Bland AI is an enterprise platform for AI phone calls, built so companies can automate the bulk of their inbound and outbound call volume with agents that hold real, human-sounding conversations. Founded in 2023, it targets the calls that used to require staffed call centers, support, sales, scheduling, intake, and identity verification, and is designed to scale to very high volumes with low latency and enterprise-grade stability.
What sets Bland apart is that it owns the entire stack. Rather than stitching together third-party speech, language, and transcription services, the company built and hosts its own text-to-speech, inference, and transcription models, all tuned specifically for phone calls and run on dedicated infrastructure. The practical consequence is that a customer's call data never passes through outside providers, which matters for regulated industries, and a single per-minute rate covers the model, speech, and telephony together.
To keep agents reliable, Bland built its own approach to scripting conversations called Conversational Pathways, which splits a prompt into a graph of nodes with routing, transfers, and conditional logic rather than relying on one long, open-ended prompt that can drift or hallucinate. Agents can clone a voice from a short audio sample, speak in dozens of languages with the ability to switch mid-call, and operate across voice, SMS, and chat with a shared memory so context carries between channels. They plug into existing telephony through SIP and providers like Twilio and connect to CRMs, scheduling tools, and contact-center platforms so an agent can actually take actions, not just talk.
One capability worth knowing about if your calls have to get through other companies' phone systems: Bland agents can work their way through a third party's touch-tone menu, inferring which keypad digits to play and playing them, and they detect voicemail and respond to it on outbound calls. That is shipped as a named tool and configurable inside a pathway.
Bland leans heavily on control and observability: teams test agents against edge cases before going live, watch calls in real time, and review post-call analysis, sentiment, and scoring. Security is a core selling point, with encryption, audit trails, access controls, and regional deployment options aimed at procurement and compliance review. The platform is developer-oriented, giving technical teams deep control to build, integrate, and expand voice agents, and is typically rolled out gradually, starting with high-volume repetitive calls and routing complex cases to humans with full context.
Vendor details
Canonical URL
https://www.bland.ai
Category
Customer support agent
Company status
independent
Use cases & customers
Target customers
Deployment options
In practice
You run high call volume in a regulated industry and can't send call data through third-party providers. Bland hosts its own speech, language, and transcription models on dedicated infrastructure, so a caller's data never leaves its stack.
An open-ended prompt makes your voice agent drift and hallucinate on real calls. Bland's Conversational Pathways split the conversation into a graph of nodes with explicit routing and conditions, keeping the agent on script.
You want one agent to follow a customer across phone, SMS, and chat without losing the thread. Bland gives agents shared memory across channels and connects to your CRM, so it can take real actions, not just talk.
Sources & related URLs
Agentic Index coverage score
12.0 / 14 capabilities · 86%
| Integrations & Tool Calling | Full |
|---|---|
|
The v2 Tools API creates tools against built-in integrations including Cal.com, Salesforce and Slack, or against any HTTP endpoint for internal services, invoked automatically mid-call and mid-SMS with the response injected back into the live conversation, with credentials held in a secrets store. Webhook nodes, a custom code node and bring-your-own-Twilio extend it further. Breadth across scheduling, CRM, messaging, telephony and arbitrary internal systems. Sourcedocs.bland.ai/tutorials/tools/integration-tools.mdread 2026-09-05 |
|
| Workflow Orchestration | Full |
|
Conversational Pathways are node-and-edge conversation graphs with branching, conditional routing on tool output, variable extraction, inline tool execution on dialogue nodes, a custom JavaScript code node, and Flex Mode letting the agent choose its next node. Pathways carry versions that promote to staging or production, folders, and a generation endpoint that builds a pathway from a prompt. Sourcedocs.bland.ai/tutorials/pathways.mdread 2026-09-05 |
|
| Knowledge Grounding & RAG | Full |
|
Knowledge bases are built and held by the platform: created from an uploaded file, direct text or a website scrape with sitemap URL discovery, then listed, updated, versioned by id and queried directly through a chat-with-knowledge-base endpoint, and attached to pathway nodes by id. The same knowledge base serves every agent it is attached to, rather than context assembled per call. Sourcedocs.bland.ai/api-v1/post/knowledge-learn-web.mdread 2026-09-05 |
|
| Human Oversight & Guardrails | Full |
|
Guard Rails are a configurable, first-class object with their own create, update and delete endpoints, running continuously on every AI response to catch critical failures before they reach the caller, backed by blocked-number rules, live listen and stop on an active call, warm and live transfer to a human, and threshold alarms routed to email, phone or a custom tool. A vendor-owned gate on the response before it lands. Guardrails, warm and live transfers and alarms are sold on the Enterprise tier; the self-serve plans carry the pathway-level transfer node only. Sourcedocs.bland.ai/tutorials/guard-rails.mdread 2026-09-05 |
|
| Security, Identity & Governance | Full |
|
SOC 2 Type I and Type II, HIPAA-eligible with a signed BAA, GDPR and PCI DSS, with compliance documentation obtainable under NDA on Enterprise, and FedRAMP certification announced on the vendor's own blog; a trust and security page, DPA and published subprocessor list sit in the footer. Customers get SSO, JWT request signing, HMAC webhook validation, organization member permissions, a secrets store and a paginated audit-log endpoint for SIEM export. The identity controls are sold on the Enterprise tier. Sourcebland.ai/trust-securityread 2026-09-05 |
|
| Observability & Auditability | Full |
|
An event-stream endpoint returns the sequence of events that occurred during a call, alongside per-execution tool logs with aggregated statistics, call detail with transcripts and corrected transcripts, recordings, citation schemas extracting structured data from transcripts, configurable alarms, and audit logs for SIEM export. Reconstruction of what the agent did step by step, not a metrics dashboard alone. Sourcedocs.bland.ai/api-v1/get/event-stream.mdread 2026-09-05 |
|
| Memory & State Persistence | Full |
|
A dedicated Memory API: contact memory records scoped to a persona or agent number, holding a summary, structured facts and recent cross-channel messages, with endpoints to read the memory context a contact sees at conversation start, retrieve what memory changed during a specific call, merge new facts, enable or disable memory per pathway version, and reset one contact's memory without touching the rest. Scope and lifetime are both stated and the memory is separable from the call record. Sourcedocs.bland.ai/api-v1/post/memory-contact.mdread 2026-09-05 |
|
| Deployment & Data Residency | Full |
|
The plan comparison lists on-premises and VPC deployment and data residency as Enterprise line items, alongside a dedicated orchestration server, and the documentation offers self-hosted and on-premises deployment for regulated industries with region-scoped static IPs for outbound and SIP traffic. The controls are sold on the Enterprise tier rather than gated behind an access program. Sourcebland.ai/pricingread 2026-09-05 |
|
| Prebuilt Agents, Templates & Packs | Full |
|
Pathway templates the customer starts from, out-of-box test scenario templates with a clone-to-my-pathway endpoint, a shipped read-only library of eval agent templates alongside org-saved ones, and Personas as reusable agent identities attached to any number of phone numbers. Named assets the buyer adopts as units, browsable and clonable. Sourcedocs.bland.ai/tutorials/pathways.mdread 2026-09-05 |
|
| Triggers & Channel Coverage | Full |
|
Inbound and outbound calling over purchased numbers, SIP trunking with auto-discovery, test calls and number porting, or a bring-your-own Twilio account; plus batch outbound campaigns from a recipient list, a web voice and chat widget, and SMS, RCS and iMessage messaging running the same pathways as voice. Sourcedocs.bland.ai/enterprise-features/sip-integration.mdread 2026-09-05 |
|
| Model Flexibility & Routing | Not documented |
|
Bland owns the full stack it runs on, stating that telephony, speech to text, the language model, text to speech and pathway orchestration are all its own, and its per-minute price bundles the language model with no per-token charge. No model selection, custom model or bring your own model surface is documented. Sourcedocs.bland.ai/welcome-to-bland.mdread 2026-09-05 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
A versioned REST API across v1 and v2 covering calls, pathways, knowledge bases, memory, tools, personas, evals, SIP and organizations, plus a command line interface, an interactive dev terminal, a Web Agent SDK with token-based auth, HMAC-signed webhooks and post-call webhooks, and a hosted MCP server with a published tool reference and setup guides for Claude Code, Cursor, Codex and other clients. Sourcedocs.bland.ai/integrations/mcp/overview.mdread 2026-09-05 |
|
| Testing, Debugging & Optimization | Full |
|
Two shipped customer-facing test surfaces with readable results: Agent Testing runs scenarios with assertions against a pathway or persona, returning assertion results and scores, batch runs, simulation sets reporting pass rates and flakiness, and node failure heatmaps; and Evals grades batches of real calls with configurable LLM judges, returning per-judge verdicts and per-call results. The artifact under test is the customer's own agent. Sourcedocs.bland.ai/tutorials/scenarios.mdread 2026-09-05 |
|
| Browser & Computer Use | Not documented |
|
Bland agents navigate third party IVR menus that take DTMF input by inferring and playing keypad tones, shipped as a play-keypad-touch-tone system tool and as pathway configuration, and they detect and respond to voicemail systems on outbound calls. That is telephone keypad control; no browser, desktop session or remote computer control is documented. Sourcedocs.bland.ai/api-v1/post/calls.mdread 2026-09-05 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Recent platform changes
Bland AI released adaptive resumption and node-scoped interruptibility for its voice agents. The system now automatically measures a caller's speaking pace and pause lengths to dynamically adjust its response timing, replacing the manual resumption speed setting.
Bears on: Agent capability
View sourceBland announced an additional $50M in funding, bringing total raised to more than $100M. Named backers include Scale, Emergence, HubSpot, Dell Technologies Capital, and Upfront.
View sourcePricing
Start $0.14/min with no platform fee; Build $299/mo at $0.12/min; Scale $499/mo at $0.11/min; Enterprise custom
hybrid (plan tier sets the per minute rate, then per connected minute usage)
Included quota
Start (free): 100 calls/day, 100 calls/hour, up to 10 concurrent, one voice clone, $0.14/min. Build ($299/mo): 2,000 calls/day, 50 concurrent, $0.12/min. Scale ($499/mo): 5,000 calls/day, 100 concurrent, $0.11/min. Enterprise: unlimited caps, custom rates, dedicated support.
What is public
Plan prices ($299 Build, $499 Scale), per minute rates by tier, transfer and SMS rates, and the free Start tier are public. Enterprise rates are custom.
Billing mechanics
In December 2025 Bland moved from a flat $0.09/min to plan based per minute rates: Start (free) $0.14/min, Build ($299/mo) $0.12/min, Scale ($499/mo) $0.11/min, Enterprise custom. One per minute rate covers the language model, speech to text, text to speech, and telephony. Transfers on Bland numbers add about $0.03 to $0.05/min (free with bring your own Twilio), SMS is about $0.02 each, and very short outbound calls carry a $0.015 minimum.
Cost watchouts
Transfers on Bland numbers add a per minute surcharge, warm transfer hold time bills at the plan rate, SMS and text to speech bill separately, number rental varies by region, and international numbers add per minute cost. The monthly plan fee does not include any minutes.
Variable cost rationale
Cost is driven by connected voice minutes, so spend scales directly with call volume and duration. The all inclusive per minute rate simplifies the stack, but transfers, SMS, and higher concurrency add on top.
Additional watchouts
The monthly plan fee is infrastructure access only, not bundled minutes, so model the per minute usage plus transfers separately. Real deployments often run higher than the sticker per minute rate.
Overage / add-ons
Usage is billed per connected minute at the plan rate; hitting a plan's call caps rejects further calls rather than incurring surprise charges. Higher plans lower the per minute rate and raise the caps.
Sales call required
Mixed (some tiers require a call)
Free / trial
Start plan: no platform fee, 2 credits and an inbound number included, no card required
Lowest paid plan
Build $299/mo (plus $0.12/min)
Commercial notes
Developer first outbound and inbound voice platform running on its own infrastructure. Enterprise adds dedicated infrastructure, SIP connectivity, US, EU, and APAC deployment options, and security controls including AES-256 at rest and TLS 1.3 in transit, with HIPAA support via a signed BAA.
Key ambiguities
Enterprise per minute rates and volume discount tiers are custom and not published; real per minute cost lands a bit above the headline once transfers and add ons are included.
Related vendors
- Decagon — AI agent platform purpose-built for customer support, with deep…
- 5.Y — Singapore early-stage platform whose GLUCOSE product deploys…
- Ada — Agentic customer experience platform whose AI agents autonomously…
- Aisera — Enterprise agentic AI platform that orchestrates specialized agents…
- ASAPP — Generative AI for enterprise contact centers that plugs into…
- Assembled — AI customer support orchestration platform that unifies autonomous…
Alternatives to Bland AI
The closest documented capability profiles to Bland AI among customer support agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Delight.ai11.5 / 14A lighter documented profile than Bland AI
- Fin11.0 / 14A lighter documented profile than Bland AI
- Kore.ai13.0 / 14Adds documented Model Flexibility & Routing
- Omilia11.0 / 14A lighter documented profile than Bland AI
- ASAPP11.5 / 14Adds documented Model Flexibility & Routing
- ElevenAgents12.5 / 14Adds documented Model Flexibility & RoutingBland AI vs ElevenAgents →
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded