ElevenLabs adds Architect assistant to its AI Agent platform, enabling it to read calls and run automated tests

Article

ElevenLabs equips voice Agents with a “AI engineer”: read calls, change configurations, and test themselves

Beating AI Briefing: ElevenAgents is ElevenLabs’ enterprise AI Agent platform. Companies can use it to build AI that handle phone calls, online chats, and customer service tasks. Now, ElevenLabs has added a built-in AI assistant, Architect, specifically to help teams build and maintain these Agents.

For example, if refund calls are frequently transferred to human agents, a team can ask Architect directly why. It reads real calls, Agent configurations, and test results, identifies the problem, and then modifies the prompt workflows, tools, knowledge bases, or guardrails.

After making changes, Architect automatically writes simulated tests to compare performance before and after the changes. All changes first enter draft status and do not directly affect production Agents. Once testing is complete, they are submitted to a human for review and deployment.

Architect can also build new Agents directly. After users explain their requirements and provide materials, it can configure prompt knowledge bases, tools, and tests. Teams can also invoke Architect through Claude, Claude Code, ChatGPT, and Cursor.

ElevenLabs says Architect can help Agents improve continuously. However, a human must still initiate tasks, and changes must receive human approval before going live.

Original link https://m.theblockbeats.info/flash/370586

Key points

ElevenAgents is ElevenLabs’ enterprise AI Agent platform, which can be used to build AI。
Architect is a built-in AI assistant added by ElevenLabs that can read real calls, Agent configurations, and test results.
Architect can modify prompt workflows, tools, knowledge bases, or guardrails.
After making changes, Architect automatically writes simulated tests and compares performance before and after the changes.
All changes first enter draft status; after testing is complete, they are submitted to a human for review and deployment.
Architect can also configure the new Agent’s prompt knowledge base, tools, and tests based on requirements and materials provided by the user.
A human must still initiate tasks, and changes must receive human approval before going live.

AI-assisted interpretation

The following is analysis, separate from reported facts. Verify important claims independently.

In simple terms, Architect is like a AI assistant that helps teams maintain customer service Agents. It can review calls and configurations, propose or execute changes, and then use simulated tests to check the results; however, human approval is still required before deployment.

Why it matters to readers

If these features operate as described, teams can handle Agent configuration, debugging, and testing within the same workflow. The original text provides no data on efficiency gains or customer adoption.

This shows that AI Agent does more than answer questions: it can also help build, troubleshoot, and test another Agent.

Risks and unknowns

The information comes from a summary of technology news; the evidence contains no official ElevenLabs announcement or independent verification materials.
The evidence does not specify Architect’s official availability scope or pricing.
The evidence does not specify whether simulated test results can represent real-world online customer service performance.
The evidence does not provide data on actual enterprise adoption or maintenance efficiency.

ElevenLabs adds Architect assistant to its AI Agent platform, enabling it to read calls and run automated tests

Article

AI-assisted interpretation

Why it matters to readers

Risks and unknowns

Related Developments

Related concepts

Keep following

Security Center