Technica Inc., now Technica AI Inc., released AIKNOW 2.0 on June 30, 2025, introducing agentic RAG. This approach allows the system to plan searches, use approved tools, and review information over multiple steps before preparing an answer.
This English adaptation is based on the June 30, 2025 press release. The former company name, Technica Inc., product scope, and company profile reflect the time of publication. See the current product page for the latest capabilities.
The announcement also introduced a free webinar scheduled for July 24, 2025. Registration is now closed; the original details are retained below for reference.
Challenges in a Basic RAG Pipeline
Retrieval-augmented generation (RAG) retrieves relevant information for a language model to use when preparing an answer. A basic RAG workflow lets users search enterprise information in natural language. Answer quality depends on the retrieved material, the reliability of its sources, and how the model uses it.
A workflow that searches only once can encounter several limitations:
- A single search may miss information needed for a multi-step query or relationships spread across sources.
- Retrieved passages may be redundant, irrelevant, or insufficient to resolve conflicting information.
- Ambiguous questions and business-specific terminology may require clarification or additional retrieval.
AIKNOW combines knowledge graph and vector search with re-ranking to select relevant information. The effectiveness of this approach needs to be evaluated using representative business questions.
Features of agentic RAG
AIKNOW 2.0 introduced search planning, branching, and repeated retrieval. These features describe the approach taken in this release; multi-step retrieval can also be used in other RAG systems.
1. Multi-step reasoning and action
When a task requires it, the system can retrieve more information, call configured APIs, or use calculation tools within defined permissions and execution limits.
2. Understanding intent and planning flexibly
The AI agent identifies what information to retrieve, reviews the results, and can adjust its search before preparing a response. It works within the configured workflow and available data sources.
3. Integration with knowledge graphs and workflows
Knowledge graphs connect information about entities across sources, supporting questions that require several retrieval steps. These relationships do not, on their own, establish cause and effect.
4. Evidence review and additional retrieval
The system can check whether the retrieved information is sufficient to answer the question and seek additional material if needed. This helps strengthen the evidence but does not replace independent assessment of answer accuracy.
Looking ahead
The release outlined the following intended benefits. Results depend on the application, configuration, and evaluation.
1. Flexible handling of complex inquiries and tasks
Repeated retrieval and branching can help address questions that draw on multiple sources. Unresolved questions and exceptions still require clarification or human review.
2. Greater efficiency and better use of knowledge
Planning searches and organizing evidence can reduce repetitive research. Time savings need to be measured in the actual workflow rather than inferred from the system’s design.
3. Better consistency, accuracy and quality
Connecting evidence and checking responses can improve consistency. Accuracy and service quality need to be evaluated against agreed test data and metrics.
AIKNOW 2.0 expanded the platform’s retrieval capabilities. The current product page describes how AIKNOW puts enterprise knowledge to work in decisions and workflows with defined controls.
Free webinar
EndedThe original webinar details are listed below. Registration is closed.
Title: The arrival of next-generation RAG: how agentic RAG is changing business operations
— Planning search strategies autonomously and integrating diverse data to support faster decision-making —
Date and time: Thursday, July 24, 2025, 12:00–13:00 (JST)
Format: Online (Zoom)
Fee: Free (advance registration required)
Original registration page (archived reference): https://technica.ai/ja/event/webinar1/
Recommended for:
- Organizations seeking to make better use of their accumulated knowledge
- Teams working to share business knowledge held by individual employees
- Organizations looking to streamline non-routine work
- Teams whose existing RAG or chatbot deployments have not met expectations
- Organizations looking to move beyond proof-of-concept trials into practical AI use and revenue generation
Speakers:
Speaker:
Satoki Tsuyuri (Representative Director and CEO, Technica Inc.)
According to the biography published with the announcement, Satoki Tsuyuri studied machine learning at Tokyo Institute of Technology and holds a master’s degree. His experience includes business consulting, AI system development, and enterprise deployment.
Facilitator:
Takuma Fujimoto (Section Manager, System Services Division, Treebell Co., Ltd.)
The announcement described Treebell’s engineering and software development services, from system proposals through operation and maintenance, as well as plans to offer AI and cloud services.
The webinar was designed for teams exploring enterprise knowledge retrieval and practical AI deployment.
About Technica Inc.
According to the company profile published in 2025, Technica had approximately 40 AI scientists and engineers providing AI consulting and system development services. The About Us page describes the current organization.
Company name: Technica Inc.
Representative: Satoki Tsuyuri, Representative Director and CEO
Founded: February 2024
URL: https://technica.ai/
Media inquiries
Technica Inc.
Email: office@technica.ai
Thank you for your interest in Technica AI and AIKNOW.
This release was distributed via PR TIMES.