Bland AI

New

Bland AI is an enterprise platform for phone, SMS and web voice agents that runs on its own self-hosted speech models, with prompt pathways, versioned releases, evaluation gates, voice cloning and per-minute plans.

bland.ai

Bland AI

Explore features, practical uses and pricing below.

Bland AI builds voice agents for phone calls, SMS and the web, and pitches itself at regulated, complex call work handled end to end. Its distinguishing claim is the stack underneath: the company says it self-hosts its models, uses no third-party model providers, and gives each enterprise customer isolated infrastructure.

Prompts, pathways and voices

An agent's node prompt sets the goal for the current step, and a system prompt sets the tone across the whole agent. When one prompt stops being enough, the agent builder lets you lay out pathways: steps you place yourself, joined by the routes a call can take. Tools and knowledge sources only reach the agent when you @mention them in the prompt, and the docs warn that a tool built without a tag fails silently. You can pick up to 20 voices, which real calls rotate through at random.

Changes move through versions. Publishing sends a change from development to staging, and promoting sends staging to production. Bland keeps every version for comparison and rollback, and an unpromoted agent won't answer calls.

Evaluations that gate each release

Bland's evaluations rely on judges, which are questions asked about a conversation with a fixed set of verdicts, and test cases that describe a simulated caller. A run can grade recent production calls, selected calls or a batch of simulated conversations. Bland can also read up to 50 real calls and write one test case from each. A conversation passes only when every judge passes it, and a promotion gate scores each judge before a version goes live.

Bland Speech and voice cloning

The agents talk with Bland Speech v3, the company's own text-to-speech model, which is also available in a studio and an API at $0.04 per 1,000 characters. It performs bracketed tags such as [laughs] or [clears throat] rather than reading them out, streams over HTTP or WebSocket, and Bland quotes 199 ms to first audio on its streaming endpoint. An instant clone takes about 10 seconds of audio, while a professional clone is fine-tuned on 30 minutes or more. Cloning requires $5 of loaded credit and confirmation that you have the right to use the voice.

Example: overnight claim intake for an insurer

An insurer wants first notice of loss taken by phone overnight. The team writes the prompt, adds a pathway for the strict part (policy number, date, location, injuries), tags a claims-system webhook as a tool and connects Salesforce. After a soft launch on a small share of calls, they have Bland write test cases from up to 50 of those calls and add judges such as "Did the agent collect the policy number before ending the call?" Each staging version runs against those cases, and only a version that clears the gate is promoted to the main claims line.

Who Bland suits, and its trade-offs

Bland is built for enterprises with compliance requirements and enough volume to justify dedicated infrastructure. Its homepage lists SOC 2 Type II, HIPAA with BAAs on Enterprise, self-attested PCI DSS 4.0.1 and GDPR DPAs, and it lists 29 languages its agents speak. The self-serve plan's daily call cap is low. Teams that want to mix speech and model vendors should look at Vapi, because Bland's models are deliberately its own. More options are in our voice agents category.

Plans and per-minute rates

The pricing page lists three plans. Start has no platform fee and charges $0.14 per connected minute, with 10 concurrent calls, 100 calls a day and one voice. Build costs $299 a month plus $0.12 a minute, with 50 concurrent calls and 2,000 calls a day. Enterprise is custom and adds unlimited daily calls, in-call SMS and web chat, BAAs, SSO and data residency. Phone service is billed separately, through your own carrier or Bland's at pass-through cost.

Quick answers about Bland AI

Can I try it without a card?

The pricing page says the Start plan needs no card, though connected minutes are still charged at the plan rate.

Can the agent use my own voice?

Yes, through a Bland Speech clone, after loading $5 of credit and confirming you have rights to the voice.

Similar tools

Synthflow builds AI voice and chat agents for contact centers and multi-location businesses, with a visual Flow Designer, CRM and calendar actions, telephony connectors, batch calling and simulations on annual contracts.
Retell AI is a platform for phone voice agents that answer and place calls, with single-prompt or node-based flows, warm transfers, batch campaigns, simulation testing and pay-as-you-go per-minute pricing.
Cartesia offers Sonic text-to-speech and Ink speech-to-text models through a low-latency API, with instant and pro voice cloning, 44-language speech output and Managed Agents for building phone voice agents.