Solutions

  • Solutions
  • Sovereign AI
INGUARD-SOVEREIGNAI
Your data. Your servers. Your AI.

InGuard-SovereignAI keeps enterprise and public sector data inside your own network.
Document analysis, retrieval, reasoning, report writing and workflow automation all run within your infrastructure.

Explore features
95%

Knowledge retrieval accuracy with hybrid search

Air-gapped

Runs and updates without any internet connection

30+

AI models selected automatically per task

hwp

Korean office documents parsed with table structure intact

WHY SOVEREIGN AI
You want AI. You cannot send your data outside.

Security policy rules out cloud AI for many organizations.
Building it in-house means people, time and operating know-how you may not have.
Four problems come up again and again.

  • Security blocks it

    The moment a question or an attached file leaves your network, it breaks internal policy.
    Projects stop at the review stage.

    01
  • Nobody can find anything

    The document exists somewhere inside the company, but no one knows where.
    So it gets written again from scratch, and archives never become assets.

    02
  • Answers cannot be trusted

    Plausible but false answers make the system unusable for real work.
    Without visible sources, verification takes longer than doing it manually.

    03
  • It never reaches the work itself

    Question in, answer out, and that is all. br Collecting, organizing, writing and sending still fall on people.

    04
WHAT WE SOLVE
All four problems, solved inside your network

InGuard-SovereignAI unifies multiple AI models, tools, data and permissions
into a single operating layer. It is an AI operations platform, not one more chatbot.

BE SECURE
Data never crosses the boundary

Parsing, retrieval, inference, sessions and logs stay inside your organization.

  • On-premise deployment and air-gapped operation
  • Real-time detection and masking of personal identifiers
  • TLS 1.3, AES-256, role-based access control, audit logging
BE SMART
The right model is selected automatically

It automatically assigns the most suitable model based on the nature of the task, security level, cost and response time.

  • Sensitive data uses local models; general tasks use external models
  • Replace the model in settings and apply new models immediately
  • Repeated queries are answered instantly from the cache
BE PRODUCTIVE
It delivers results, not just answers

From research and analysis to document creation and delivery, the entire process is completed in a single request.

  • Automatically generate reports in hwp, pdf, docx and pptx formats
  • Connect to core systems via API and carry out actual operations
  • Automatically execute recurring tasks on schedule
FEATURES
KEY FEATURES

We provide ready-to-use business modules along with API engines that can be integrated into existing systems.

WHAT WE SOLVE

Conversations powered by internal documents

Answers are generated using both internal documents and web searches, with referenced document names and pages shown as sources. File attachment analysis, summarization, translation and image description are supported.

AI DOCUMENT

Automated report generation

Creates business plans, weekly/monthly reports and analysis reports based on templates. Automatically generates charts and tables, then lets you download them as hwp·pdf·docx·pptx files or send them by email.

KNOWLEDGE RAG

Next-generation knowledge search

Combines three types of search—similarity, question and keyword search—and determines the chunking method based on document characteristics. Automatically determines the optimal approach. Smart Tune recommends the best settings so even non-experts can adjust the quality.

FORM BUILDER

Turn documents into input forms

Recognizes scanned documents and images and reconstructs them as input forms. It identifies table, image and signature areas, then maps fields for direct integration with business systems.

AGENT

Self-executing agents

Breaks complex requests into multiple steps and processes them. It uses more than 100 tools, and can generate and execute code in an isolated environment when needed.

FINE-TUNING

Models tailored to your organization

Trains the model on internal terminology and business context. AI automatically generates training question-and-answer datasets, so the process can be carried out without specialized personnel.

GOVERNANCE

Measure performance and usage

Measures response quality with evaluation sets and compares models. Monitor usage, permissions, spending limits, and security events from the administrator dashboard.

OPEN API

Works directly with existing systems

All features are provided through Open API 3.0 standard REST APIs. ERP, MES, groupware and portals can call AI functions directly from their existing interfaces.

INGUARD-SOVEREIGNAI
Our search accuracy determines answer quality

A single vector search is not enough to achieve sufficient hit rates for real-world business documents.
We combine three types of search and train organization-specific embeddings to improve accuracy.

75% → 95%

Search accuracy before and
after applying 3-level hybrid search

0.005 sec

Response time on a cache hit

6 stages

RAG pipeline from query
decomposition to verification

Source citation

Reference documents and
pages shown in every answer

ARCHITECTURE
All processing within the customer's infrastructure is completed

The reasoning, knowledge and governance engines are integrated into a single operating system that allows only authorized external communication according to policy.

Users · Customer systems

The three engines for reasoning, knowledge and governance are integrated into a single operating system.

Reasoning · Orchestration

Automatic model selection · Tool execution
Multi-step task orchestration

Memory · Knowledge

Document understanding · Hybrid
search · Knowledge graph · Embeddings

Governance · Control

Personal data masking · Access control
Audit logs · Spending limits

Models · Data

Local LLM · Vector database · Knowledge graph · Cache · Relational DB · File storage

Deployment - On-premises · Air-gapped · Private cloud · Hybrid

Within the customer's data center boundary

WHAT WE SOLVE
Start small and scale

You do not need GPU servers from the start. As you move to the next stage, the knowledge base and settings built in previous stages carry over unchanged.

START Start with document search

Upload internal documents to start knowledge search and Q&A. A single standard server is enough without a GPU.

Configuration

API server, basic RAG, database, vector storage

Model

Combination of RAG and external LLM

GROW Establish data sovereignty

Deploy a local LLM internally to operate without external communication. Add document generation and workflow automation.

Configuration

Report generation server, job queue, cache server, web search integration

Model

Internal LLM-centered, inference GPU server

ADVANCE AI tailored to your organization

Operates models tailored to your organization's terminology through domain-specific embeddings and fine-tuning. Extends to multimodal capabilities.

Configuration

Document AI, knowledge graph, multi-agent system, model operations

Model

Proprietary trained models, training GPU servers

INDUSTRIES
Industries

Accurately interpreting the structure of Korean documents to maximize the value of document-centric work is our key differentiator in the Korean market.

  • Public sector · Administration

    Civil service support, administrative document drafting, guideline and legal research, and press release writing. Convert accumulated HWP assets into a knowledge base.

  • Finance

    Internal policy search, risk report drafting, Support for consultations in closed networks. Data prohibited from external transfer is processed internally.

  • Manufacturing

    On-site support based on work instructions and equipment manuals
    Q&A, quality analysis and process knowledge search

  • Healthcare

    Clinical guideline search, health screening data analysis,
    Protect sensitive information through internal network processing

  • Transportation · Urban

    Retrieve control-room response guidelines, check facility history, and automatically generate situation reports

  • Construction · Plant

    Search design documents and specifications, safety regulations
    and inspection report automation

PENTAGATE ADVANTAGE
Perception AI and decision-making AI from a single provider are supplied by a single provider

Pentagate has supplied deep-learning video analytics and smart infrastructure to the field. With Sovereign AI added, events occurring in the field and the decision criteria accumulated by the organization are connected in a single workflow.

Video intelligence

Detects situations across CCTV, ITS and smart factory environments.

  • Safety violations and entry into restricted areas
  • Traffic flow and unexpected incidents
  • Early detection of disasters such as wildfires
Sovereign AI

Makes decisions based on the organization's policies and history regarding detected situations.

  • Provides response procedures and action guidelines
  • Automatically generates situation reports
  • Notifies relevant departments and updates connected systems
A single internal network

Both capabilities operate within the customer's infrastructure.

  • Neither surveillance video nor document assets leave the organization
  • Integrated delivery without additional integration development
  • Implementation and maintenance through a single point of contact
FAQ
Frequently Asked Questions

Yes. It can be used even in a closed network with no internet access at all.

It interprets HWP and HWPX at the structural level. It preserves table row and column relationships, paragraph hierarchy, and the positions of images and formulas when extracting content, so questions such as "What is the budget in table 3" can be answered accurately.
It can also recognize scanned documents and photos.

No. It can flexibly integrate with various models without being dependent on any specific AI model.

Yes. It can integrate with existing ERP and groupware, and can be configured using APIs to fit your enterprise environment.

It can be configured so that you can also verify the documents and sources supporting each answer.

Implementation time varies depending on the enterprise environment and scope of integration, and we will provide an appropriate implementation schedule through consultation.