Voice AI is transforming how businesses handle phone communications. Brooke recently sat down with Will Bodewes from Phonely to discuss how they're using voice agent simulations & evaluations to ensure reliable customer interactions at scale.

Building Voice AI That Enterprises Can Trust

Phonely helps businesses deploy voice AI agents through a no-code interface. Their platform integrates with existing business tools, from scheduling software to CRM systems, enabling automated phone interactions that consistently meet quality standards through rigorous evaluation and simulation.

"If you run a business and you're trying to get voice AI integrated into that to answer the phones, that is exactly what we do," explains Bodewes. The solution has proven particularly effective in high-call-volume industries like insurance, Medicare, and healthcare, where systematic AI agent evaluations are crucial for maintaining quality.

Moving Beyond Basic Prompts with Workflow-Based Voice AI

What makes Phonely's approach unique is their focus on structured workflows rather than simple prompts. This system enables more reliable voice agent performance through systematic evaluation and testing.

"We realized that a prompt wasn't going to solve the problem," Bodewes notes. "These businesses have a series of questions that need to be asked or steps that need to be followed." This insight led to their workflow system, which uses structured prompt chains combined with continuous LLM evals to handle complex conversations effectively.

Ensuring Quality Through Voice Agent Simulations

A major challenge in deploying voice AI is maintaining consistent performance given the inherent variance in large language models. Phonely addresses this by integrating Coval's voice agent simulation capabilities directly into their product, allowing users to systematically test and validate their voice agents.

"When you have a phone call that goes really well, you can add that to a simulation test," Bodewes explains. "If you change something else above that conversation and want to see if that call is still going to run well, you can rerun that whole conversation using large language models to simulate it."

This evaluation and simulation approach enables businesses to:

• Test changes before deployment using voice agent simulations

• Track version control with systematic AI agent evaluations

• Validate performance across diverse scenarios

• Identify and resolve issues through automated testing

How Voice AI Is Evolving

The combination of voice AI and systematic evaluation is opening new possibilities for customer interaction. "Voice AI gives you all these new controls that you can adjust to change the conversation," Bodewes explains. "You have so much more data and insight into conversations than you ever did with traditional phone systems."

This enhanced control through evaluation and simulation enables:

Testing different conversation flows with automated evaluations

• Using LLM evals to measure customer sentiment

• Quick iteration on scripts with immediate feedback

• Data-driven optimization through systematic testing

Building Better Voice AI Through Evaluation

The collaboration between Phonely and Coval shows how voice AI agent simulations are becoming essential for enterprise voice AI adoption. As the technology evolves, systematic testing and quality assurance remain critical for successful deployment.

"It's been interesting to see how voice AI evaluation has become increasingly important," reflects Bodewes. "We're seeing rapid adoption of both voice AI and the tools needed to test it effectively."

For more information about implementing simulations and evaluations for your AI agents, reach out to brooke@coval.dev