FlashLabs Overhauls AI Voice Agent 'FlashAI,' Evolving it into a Full-Fledged Platform for Inbound/Outbound Calls with Reusable AI Assistants and 4-Level Inference Depth

FlashLabs Inc. (Headquarters: Chiyoda-ku, Tokyo; CEO: Yoichi Hosoi) announces a significant enhancement to its voice AI campaign operations platform through three major updates to its AI voice agent, 'FlashAI,' implemented from May to June 2026. This update newly adds features such as a four-level configurable AI inference depth, reusable assistant configurations across campaigns, a live test function, and version control. With these changes, FlashAI evolves from a single-purpose AI calling tool into a 'Voice AI Campaign Operating System' for centrally building and managing call and response tasks with voice AI.

Background and Objectives

Since 2025, the use of generative AI in Japan's contact center industry has been rapidly shifting from 'pilot introduction' to 'full-scale operation.' According to industry research, the adoption rate of generative AI reached approximately 63% as of 2025, and companies responding 'no plans to use' plummeted from 29% to 3% (Source: Mobilus Inc. 2026 Contact Center Trends Forecast Seminar).

However, three key challenges have emerged in the full-scale operation of voice AI agents:

1. **Mismatch between Task Complexity and AI Performance** The complexity of calls ranges widely, from simple inquiries to advanced consulting sales. A single AI model can be overkill and costly for simple tasks, while being underpowered and leading to poor customer experience for complex ones.

2. **Operational Burden from Redundant Setups for Each Campaign** For companies running multiple voice AI campaigns for different purposes like sales, support, and appointment setting, the need to build prompts and voice settings from scratch for each campaign created a significant burden in terms of lead time and management costs.

3. **Insufficient Pre-Launch Verification** The response quality of an AI agent often can only be verified in actual calls. Deploying them without sufficient prior testing carries the risk of unexpected responses and a degraded customer experience.

To solve these issues, FlashLabs radically redesigned FlashAI's core architecture, rolling out three major updates sequentially from May to June 2026.

Overview of FlashAI's Major Updates

① **Four-Level AI Inference Depth: 'Core / Reflex / Insight / Strategist'** FlashAI now allows users to select from four levels of AI inference depth when creating a campaign. Higher levels (Insight / Strategist) enable more advanced AI agent behaviors, such as maintaining context over multiple turns, objection handling, and triggering follow-up workflows via CRM integration. Additionally, cost management transparency has been improved, allowing users to check the inference depth used and credits consumed per call under Settings → Billing & Plan.

② **Campaign Language Support for AI Summaries** FlashAI's call summaries are now automatically generated based on the language specified in the campaign settings. For example, if the campaign language is set to 'Japanese,' the AI summary in the call log will also be in Japanese, facilitating review by domestic teams.

③ **Reusable Voice AI Assistant Architecture** At the core of this update is a new architecture that structures the AI voice agent's settings as an 'Assistant,' making them reusable across multiple campaigns.

- **Expert Configuration Mode**: By switching from 'Base' to 'Expert' mode in the campaign settings, users can integrally configure the Transcriber, Model (conversational intelligence), Voice, First Message, System Prompt (agent role/logic), and Voice Settings. - **Assistant Reusability**: Once configured, an assistant can be reused across different campaigns. For instance, a sales assistant can be used for multiple outbound lead lists, or a support assistant can be applied to several inbound campaigns. This eliminates redundant setups and significantly shortens the time to launch new campaigns.

Value for Businesses

1. **Balancing Optimal AI Performance and Cost According to Task Complexity** The four-level inference depth allows businesses to allocate the optimal AI resources for each task—'Core' (low cost) for simple notifications and 'Strategist' (high performance) for advanced consulting sales. Paired with credit consumption visibility, this enables ROI-maximized operational design.

2. **Increased Speed and Consistency in Campaign Deployment** Reusing assistants drastically reduces the launch time for new campaigns. It allows for maintaining consistent AI agent quality across departments like sales, support, and appointment setting, while still allowing for easy individual tuning.

3. **Pre-Launch Quality Assurance and Continuous Improvement** Pre-launch testing with 'Talk Preview' and safe rollbacks with version control allow for a gradual transition to live operation while incrementally improving agent quality. The JSON export function also facilitates settings audits and backups.

Future Developments

FlashLabs will continue to evolve FlashAI from a tool for single campaigns into a reusable, testable, and version-controlled 'Voice AI Campaign OS.' Future plans include a unified operational foundation for both inbound and outbound calls, more advanced inference and tool integration, and enhancements toward operational automation and autonomy, continually supporting corporate calling and response operations.

A Comment from the CEO Yoichi Hosoi, CEO, FlashLabs Inc.

'2025 was the year generative AI began to be seriously tested in the contact center industry. 2026 is the turning point from testing to practical application. The 'Assistant Architecture' and 'Four-Level Inference Depth' we've introduced to FlashAI at this time are not just feature additions; they are the foundation for transforming voice AI agents from 'disposable tools' into 'organizational assets.'

By reusing and version-controlling assistants, both sales and support teams can systematically accumulate the results of their trials and continuously improve. The choice of inference depth implements a mature approach to AI utilization—not the simplistic idea that 'more expensive AI is always better,' but the selection of appropriate AI performance that matches business value.

We will develop FlashAI as a platform that ultimately leads sales and customer experience toward autonomy.'