Equipment Maintenance AI Agent 'EMLink Intelligence' Significantly Updated — Now Capable of Autonomous Execution of Complex Tasks and Building Custom RAG in as Little as 1 Minute Using In-house Knowledge

Setsubi Hozon Sogo Kenkyusho Co., Ltd., a provider of the equipment maintenance cloud service 'EMLink' for factories and infrastructure, has significantly updated its AI agent function, EMLink Intelligence. It has evolved from an 'AI that answers questions' to an AI agent that autonomously executes multiple steps by decomposing tasks simply by being given a goal. It also supports graphical display of analysis results and reporting output to Excel and PDF.

Furthermore, it will be possible to easily build a custom RAG by simply incorporating the company's standard manuals and glossaries (scheduled for release), and by combining it with the autonomous agent, it will derive company-specific answers (unique solutions) that are not general theories. It autonomously executes everything from search, aggregation, graphing, analysis, to report creation, and in our verification, it reduces maintenance work time by up to 95%.

It supports the entire process from information gathering necessary for preventive maintenance to action plan formulation and prioritization integrating past know-how.

Limitations of Single-Instance Analysis Using Generative AI

Until now, the use of generative AI in the maintenance field has primarily been 'inquiry-based,' where users read Excel or CSV files and receive one answer per question. However, maintenance field operations consist of a series of multiple steps such as data normalization, search, aggregation, graphing, analysis, and reporting. With conventional AI utilization, humans had to continuously provide detailed instructions for each step.

Additionally, the difficulty in verifying the basis of AI responses and their consistency with internal maintenance procedures and standards has been a factor hindering adoption in the field.

EMLink Intelligence Update

Main Updates

1. Evolution to an 'AI Agent' that Operates Autonomously

By simply communicating a goal, the AI autonomously decomposes the tasks to be performed and sequentially executes multiple steps: search → aggregation → graphing → analysis → report creation. The execution progress and results of each step can be confirmed on the screen, allowing humans to focus on the final decision.

2. Graphical Display of Analysis Results

Aggregation and analysis results are visualized in various formats such as bar graphs and matrix diagrams. Since they can be grasped as diagrams rather than mere numbers, trends and priorities can be intuitively understood.

3. Reporting Output (Excel/PDF)

Generated tables and analyses can be output and shared as Excel (multiple sheets) or PDF (A4 report). It also supports copying to the clipboard, allowing direct use in internal reports and reviews.

4. Easily Build Custom RAG — Combine with Agent for 'Unique Solutions' (Scheduled for Release)

A custom RAG can be easily built by simply incorporating the company's standard manuals, maintenance procedures, various standards, and glossaries. By combining this with Intelligence's autonomous agent, it derives 'unique solutions' that align with the company's site and standards, rather than general theories.

Perspective

Generic AI

EMLink Intelligence (Custom RAG x Agent)

Response

Stays within general theories

Unique solutions based on company manuals and standards

Technical Terms

Prone to misinterpretation

Correctly interprets abbreviations, common equipment names, and model numbers using glossaries

Basis for Response

Unclear

Specifies the source (which document, which part), useful for audits and training

Construction

(Supplemented with prompts as needed)

Builds custom RAG in 1 minute or more just by importing company documents (PPT, WORD, PDF, TXT)

For example, if asked, 'Referencing the vibration standards for centrifugal pumps from our company's maintenance procedures, check if there are any dangerous signs in the inspection results from the past year,' it will provide an answer with the relevant section cited as the basis.

These tasks can be autonomously executed just by 'assigning' them:

From an RBI (Risk-Based Inspection) perspective, calculate risk from impact (CoF) x probability of occurrence (PoF) and graph it using a risk matrix.

For equipment used over a long period, compare replacement and continued repair using NPV/IRR (Defender–Challenger) and present the top 5 replacement priorities.

Organize failure modes from past defects, inspection history, and accumulated trouble cases within the company, and extract equipment that requires重点 inspection.

Aggregate maintenance costs for the past few years by equipment type and identify outliers and abnormal costs.

For inspection results of specific machinery, cite the company's maintenance procedures and determine OK/NG with a basis (RAG).

Identify instances where spare parts inventory falls below safety stock levels and create an order proposal for long-lead-time parts.

Reduce Work Time by Up to 95%

These analyses and report creations traditionally required the person in charge to go through steps like 'search → filter → aggregate → graph → consider.' By having EMLink Intelligence autonomously execute these series of tasks, our verification shows a reduction in work time of up to 95%. This allows personnel to focus on 'decision-making' tasks rather than 'creation' tasks.

What is EMLink?

'EMLink,' an Asset Management Cloud, is an equipment management service for plants and factories. Aimed at cost reduction, resource efficiency, and strengthening management systems, it is being introduced nationwide, primarily to process manufacturing industries such as chemicals, materials, pharmaceuticals, and food, as well as infrastructure operators. As a 'platform-type EAM for manufacturing sites' that connects site data with management data, it can be introduced starting from 120,000 yen per month per factory (50 users).

About Setsubi Hozon Sogo Kenkyusho Co., Ltd.

EML is a maintenance tech company that provides the asset management cloud 'EMLink' for plants, factories, and infrastructure. With the mission 'to support social and industrial infrastructure from behind the scenes,' since its establishment, it has provided 'EMLink' to over 100 leading companies in sectors such as oil, chemicals, automotive, steel, power infrastructure, and local governments.

Company Overview

Company Name: Setsubi Hozon Sogo Kenkyusho Co., Ltd.

Established: September 2022

Representative: Shogo Aihara, Representative Director & CEO

Headquarters: 3-18-16 Megurohoncho, Meguro-ku, Tokyo

Business Description: Development and sales of the asset management system 'EMLink' for factories and infrastructure, and related consulting.

Website: https://em-labo.co.jp/