The artificial intelligence-powered voice platform Bland Inc. today announced Norm, an AI assistant that allows technical teams to build reliable voice agents from a conversational prompt.

According to Bland, building voice AI agents is extremely difficult and requires considerable expertise. Unlike normal chat-based systems that merely respond to voice activation, AI voice agents must handle human-like responses, including interruptions and tone, while operating at extremely low delays that feel natural.

Most people aren’t capable of engineering the systems that build these systems under the hood, writing the prompts, constructing the necessary application programming interfaces, connecting the data and wiring all the elements.

According to Chief Executive Isaiah Granet, Norm changes all this. “What we do is the most complicated phone calls in the world,” Granet told SiliconANGLE in an exclusive interview.

Bland doesn’t target companies attempting to replace phone trees or businesses seeking to have voice agents answer simple questions. Instead, Norm can build complex scenarios such as instructional guidance for products, appointment scheduling around flexible time constraints and more.

Ideally, Bland wants to target scenarios that require nuance and not necessarily simplicity. An agent could run 45 minutes and interpret ambiguity, troubleshoot devices for a patient with a blood pressure cuff and escalate when needed.

For example, a user could provide a simple instruction to Norm: “Build me a full scheduling agent and integrate with my Cal.com.”

The system would then go to work and automatically generate the prompt, persona, agent, pathways, validation conditions, extraction rules and API integrations. The engineering team can then thoroughly test the agent before deploying it live.

All changes go to a safe branch; users can review changes between the original and the updated prompts.

This is what the company’s Forward Deployed Engineers used to do by visiting people in call centers and on the phone. Now Norm is making that process self-serve. “The idea with Norm is what we’re making accessible to everybody is essentially what our Forward Deployed Engineers have been doing for our biggest customers,” Granet said.

Users don’t need to get a Twilio account or deal with connecting APIs and all the other components, he explained. Norm can sort that out in one shot or a few shots. It’s designed to provide customers with a full-service experience, with a voice assistant on the other side.

When customers approach Norm, they see what is essentially a chat-style window where they type what they want to build, such as the scheduling assistant mentioned above. Norm then asks follow-up questions to clarify, handles naming, integration needs and publishing, and can deploy across web, phone and SMS texts.

Granet said the voice agent industry is reaching a point where AI is becoming part of the customer service stack. It’s replacing phone trees, but it’s also providing a different way of interacting with companies. The problem, he said, is that many companies try to use voice agents to prevent people from connecting to customer service, ultimately blocking them from getting to a live person.

He said Bland’s mission is partly about avoiding what he calls the “AI pain machine,” where AI argues with the caller, blocks escalation and recreates the worst parts of legacy call automation. “We shouldn’t replace it with phone trees that argue with you,” he said.

Bland also said a major part of its differentiation comes from building its models fully in-house rather than relying on outside providers. Granet said that decision was driven by privacy and security for sensitive conversations, reliability for business-critical calls, and the need to produce better voice experiences by default.

“We’re not going to ever be OpenAI in the sense of general intelligence, but what we will be is phone call intelligence better than anybody else in the world,” Granet added.

Image: SiliconANGLE/Microsoft Designer