Builder update: What we've shipped for AI voice agents and beyond

As organizations expand AI across more channels, languages, and customer journeys, they're running into a new challenge: turning early AI success into a repeatable operational capability. Scaling AI customer service requires the ability to continuously improve, validate, and expand AI interactions over time.

Over the past few months, we've expanded the capabilities that power Ada ACX, from AI voice agents and Simulations to MCP Server enhancements and email operations. Each release is designed to help enterprises build, deploy, analyze, and continuously improve AI agents across every channel.

These releases are designed to help organizations scale customer experiences with greater confidence, whether that's supporting customers in dozens of languages, validating complex workflows before deployment, or managing customer interactions across channels.

Here's what we built.

Powering richer voice experiences

AI voice agents improve customer satisfaction by reducing friction during interactions. Background noise, response delays, and language barriers can all undermine customer confidence before a resolution ever happens.

The challenge with voice is that customers evaluate every interaction in real time. They notice the pauses, the interruptions, when an experience feels natural, and when it doesn't.

This spring, we shipped four updates that help AI voice agents deliver more reliable, natural customer experiences at scale.

Background noise cancellation

Real calls don't happen in quiet studios. Customers call from moving vehicles, open offices, and busy households.

Without advanced noise suppression, background sounds can trigger unintended utterances, stall workflows, corrupt transcripts, and ultimately reduce automated resolution rates.

Ada’s Voice AI now incorporates noise cancellation directly into the voice processing architecture before speech detection and transcription occur. Noise is removed at the source rather than filtered downstream.

The result is cleaner input, more reliable conversations, and stronger performance in the environments where customers actually call.

Latency improvements, Smart Interruptions, and interstitial audio

Customers judge AI voice agents in two ways:

1. Was the issue resolved?
2. Did the interaction feel natural?

The second question gets answered in the pauses. When an AI agent is retrieving information from a business system, silence creates uncertainty. Customers assume the agent didn't hear them, the call dropped, or something went wrong.

• Latency improvements address this at the source, with faster infrastructure for call processing, speech synthesis, and knowledge retrieval, reducing response times across every voice interaction with no configuration required.
• Interstitial audio addresses the moments that can't be eliminated. When a response exceeds a configurable threshold, a brief audio cue lets customers know the AI agent is still working on their request. ACX managers can choose the audio experience and trigger timing to align with their customer experience strategy.
• Smart Interruptions help conversations flow more naturally. The AI agent can now distinguish between meaningful interruptions and conversational signals like "ok," "sure," or "uh-huh." Instead of restarting its response, the agent pauses and continues where it left off.

Together, these improvements help AI voice agents deliver more natural conversations while reducing friction during service interactions.

Expanding multilingual support

Enterprise organizations rarely support a single language. More commonly, they need to support customers across:

• Multiple regions
• Multiple brands
• Multiple languages
• Multiple service channels

Managing separate AI agents for each combination creates operational complexity that doesn't scale.

Ada’s Voice AI now supports 42 languages natively: English, French, Dutch, German, Italian, Portuguese, Spanish, Swedish, Albanian, Arabic, Bosnian, Bulgarian, Catalan, Chinese (Simplified), Chinese (Traditional), Croatian, Czech, Danish, Estonian, Filipino (Tagalog), Finnish, Greek, Hebrew, Hindi, Hungarian, Indonesian, Japanese, Korean, Lithuanian, Macedonian, Malay, Norwegian, Polish, Romanian, Russian, Serbian, Slovak, Slovenian, Thai, Turkish, Ukrainian, and Vietnamese.

ACX teams can configure an AI agent once and deploy it across every supported language without maintaining separate experiences. This means organizations can automatically support multilingual customer service through a single AI agent rather than building and managing separate deployments for every market.

Driving deeper resolution: Simulations and Coaching

Most organizations believe AI agents can handle increasingly complex customer interactions. The question is whether they can verify those experiences before customers encounter them.

Without end-to-end testing, teams can't confidently validate that workflows execute as intended. Conditional logic breaks, integrations fail, and edge cases emerge only after deployment. That's the gap Simulations is designed to close.

ACX teams can now test complete customer journeys across voice, messaging, and email before changes reach production. Teams can validate:

• Conditional logic
• Multi-step workflows
• Customer confirmations
• Backend actions and integrations

For voice experiences, teams can play back the full simulated interaction as a customer would experience it, with transcripts, audio, and reasoning steps visible in one place.

But identifying issues is only half the process.

Coaching for Simulations

When a simulation uncovers a problem, ACX teams need a fast path to improvement. Coaching for Simulations brings that workflow directly into the testing experience. Teams can review what happened, understand why the AI agent responded the way it did, and apply Coaching without recreating the scenario or switching tools.

The updated behavior can then be validated immediately through another simulation. The result is a continuous improvement cycle of simulate, identify, coach, validate, and ship.

Every issue discovered in testing becomes an opportunity to improve performance before customers ever experience it. This is what an agentic customer experience looks like when it's built for continuous improvement.

Empowering ACX teams: MCP server improvements

ACX managers increasingly analyze performance inside AI workspaces like Claude, ChatGPT, and Copilot, alongside CRM, BI, and operational data.

The challenge is that insights are only valuable when teams can act on them.

Until recently, that meant identifying an issue in one tool and switching to another to make changes, run tests, or validate improvements. The improvement loop—analyze, fix, test, measure—couldn't close where the work started.

Ada’s MCP server update introduces three major improvements:

• Knowledge write tools: Create and update knowledge articles directly from your AI workspace the moment an opportunity is identified.
• Simulations tools: Retrieve test cases, review results, create new scenarios, update existing tests, and trigger validation runs without switching tools.
• Guided reporting: Automatically generate performance reports using orchestrated prompts and tool workflows that pull metrics, conversation samples, and recommendations together.

Together, these capabilities help ACX teams move from insight to action without losing context. The result is a continuous improvement loop that lives in a single workspace: analyze, act, test, and measure.

Improving messaging: Multi-inbox support for email

Enterprise email operations are rarely simple. Most organizations manage dozens of inboxes across brands, regions, business units, and languages. Historically, supporting those inboxes with AI meant creating and maintaining separate AI agents for each one.

That's where complexity starts to compound.

Ada supports omnichannel customer service across voice, messaging, and email through a unified AI agent experience. These inboxes frequently include:

• Regional support addresses
• Department-specific queues
• Brand-specific inboxes
• Language-specific email channels

The traditional approach creates operational overhead: workflows must be duplicated, configuration changes must be replicated, reporting becomes fragmented, and expansion requires rebuilding what already exists.

Support for multiple inboxes eliminates that overhead. Organizations can now connect regional, departmental, language-specific, and brand-specific email addresses to a single AI agent while maintaining independent inbox identities.

With support for multiple inboxes, teams can:

• Connect multiple service email addresses to a single AI agent
• Maintain unique sender names, branding, headers, and footers for each inbox
• Control launch settings independently by inbox
• Automatically respond from the correct email address every time
• Analyze performance by individual inbox or across the entire operation from a unified view

Adding an address for a new region, department, or brand takes minutes. The result is omnichannel customer service that scales without creating additional operational burden.

What this adds up to

Ada's 2026 Agentic CX Report found that 92% of organizations expect to increase investment in AI customer service over the next year. As organizations expand AI across more channels, languages, and customer journeys, the challenge shifts from deployment to operations.

That's what ties these releases together, helping organizations create a stronger foundation for agentic customer experience—one that can adapt, improve, and expand as customer expectations evolve.

These releases help enterprises:

• Deliver more consistent experiences across voice, messaging, chat, and email
• Scale AI agents across languages, regions, and brands
• Validate and improve customer experiences before changes reach production
• Create continuous improvement loops that compound over time
• Reduce operational complexity as AI programs grow

There's more to come.