“Boost the Refund Resolution Rate”: ElevenLabs to Support AI Agent Operations Improvements
IT DAILY ·
✦ AI Summary
ElevenLabs on the 19th unveiled the alpha version of a new feature for ElevenAgents, ElevenAgents Architect.
The feature analyzes conversation logs when users describe requirements by voice or text, then presents change proposals and verification simulations, supporting both building and operational improvements in the same conversation flow.
Architect is available in Architect, Claude, Claude Code, ChatGPT, Cursor, and Grok, and the company plans to expand its use to customer experience, compliance, and product teams in the future.
Global AI audio company ElevenLabs on the 19th unveiled the alpha version of a new feature for ElevenAgents, ElevenAgents Architect. ElevenAgents is an enterprise voice and chat agent platform.
The feature is designed not only to support the building of enterprise AI agents, but also to help with conversational detection and improvement after deployment. A key feature of Architect is that it connects the initial agent creation process and the operational improvement process in a single conversational flow.
Users can describe the functions they want using voice or text. For example, they can ask questions such as how to improve refund processing resolution rates.
When a user explains a problem that needs improvement, AI analyzes the conversation logs. It then proposes changes and also presents simulation tests for verification.
This is intended to replace the process in which staff manually reviewed counseling records and adjusted individual settings. ElevenLabs said it built this feature around a workflow that handles the building of enterprise AI agents and operational improvements in the same conversation.
Architect is a feature that supports building and improvement based on interactions with ElevenAgents settings items such as knowledge bases, prompts, workflows, voice settings, procedures, guardrails, and external tools. Using failed tests, a surge in inquiries, and issues found in ElevenAgents Spotlight as starting points, Architect analyzes conversation logs and identifies areas that need changes. Its main function is to comprehensively analyze several thousand conversations and suggest improvement opportunities that staff may easily miss.
In a later stage, Architect generates change proposals and writes exception-handling simulations to support verification before deployment. It also supports setting up guardrails and exception simulations from the building stage. Under this concept, the role of staff shifts from searching for problems and working on individual settings to judging the appropriateness of proposed changes and deciding whether to deploy them.
Architect has the functional characteristic of changing the settings of agents that are already in operation. The feature includes change history and deployment control. On the dashboard, Architect saves revised settings as version-controlled drafts, and the publishing process requires user review before posting. The changes are also subject to standards defined by the team.
ElevenLabs said its feature is not limited to its own environment and can also be used in multiple external AI tools. ElevenLabs explained that the way it is applied when using external AI tools varies, and that changes made through external tools may be directly reflected depending on the approval settings of those tools.
The feature is available in Architect, Claude, Claude Code, ChatGPT, Cursor, and Grok. It said the standards for building and managing agents in external AI tools are user-provided information and connected tool lookup information.
ElevenLabs plans to expand the scope of the feature's use in the future to customer experience, compliance, and product teams. As an example of use for customer experience staff, it cited requests to revise procedures in line with policy changes.
As an example for compliance staff, it cited requests to confirm the scope of guardrail application. For product staff, it included requests to check resolution rates after changes.
The company said this workflow could reduce the need for on-the-ground staff to ask specialized personnel to handle tasks and then wait. It added that it also expects the feature to help technical experts and implementation partners focus on system integration, architecture design, and deployment strategies.
Karl Hughes, head of AI services at The Acon Group, said that in highly regulated sectors such as insurance, deployment speed is important, but it is also important to control an agent's behavior at each stage. He said Architect supports both deployment speed and stage-by-stage control of agents.
Source: IT DAILY · Lee Jae-young
Original: https://www.itdaily.kr/news/articleView.html?idxno=242285
References
This article was produced with the help of an automated content generation algorithm.
Source: IT DAILY
View originalThis article was summarized and organized by BizCrush based on the original article from IT DAILY. For exact quotations and full details, please refer to the original article.