PKSHA AI Solution
: How to use it, features, and the business problems it solves
Add bookmark
What is PKSHA AI Solution?
PKSHA AI Solution is PKSHA Technology's (TSE Prime) customized AI development service. It covers consultation and requirements definition through model building, API development, and production deployment, with many cases in retail, finance, and insurance. Pricing is not disclosed (custom quote). 4,400+ client companies (as of October 2025).
Business problems it solves
About "PKSHA AI Solution"
What is PKSHA AI Solution?
PKSHA AI Solution is a customized AI solution service that develops and implements AI algorithms tailored to each company's challenges, provided by PKSHA Technology Inc., a company listed on the Tokyo Stock Exchange Prime Market. The company describes it as "not simply contract development," supporting clients end to end from problem definition through system design to integration into business processes.
PKSHA researches and develops algorithms for natural language processing, image recognition, and machine learning/deep learning. Its business runs on two pillars: "AI Solution" (customized) and "AI SaaS" (packaged). AI SaaS products are productized results of AI Solution work, and PKSHA can propose combinations of both. It is an option for large companies in retail, finance, insurance and similar industries that want to build their own forecasting, optimization, fraud detection, or LLM systems.
Engagement Overview
How the Engagement Works
The official "Implementation Steps" consist of five stages. The durations are official guidelines.
Consultation
Understanding challenges and discussing development policy (about 1 month)
Requirements definition
Initial analysis and decision on development policy (1–3 months)
Model building
Feature design and model building, with cycles of accuracy evaluation and tuning (3–6 months)
API development
Optimizing system integration (1–2 months)
Production deployment support
Support for deploying into production
Adding up the guidelines for STEPs 1–4 gives about 6–12 months (our own total; production deployment support is not included).
A distinctive feature is the time spent refining accuracy at the model-building stage. The industry pages state that PKSHA "focuses on commercial use, not PoC," meaning projects are designed to reach production rather than end at a proof of concept. Integration with existing systems is also emphasized: for credit card fraud detection, PKSHA embedded its algorithms into the fraud detection systems clients were already using.
Service Menu
01Industry solutions
Algorithms for forecasting, optimization, image recognition and more, built to fit each industry's challenges.
- Retail & distribution — Demand forecasting and inventory optimization, automatic shelf planogram creation, automated shift scheduling. Product suite: "PKSHA AI Suite for Retail"
- Consumer credit & banking — Credit card fraud detection, AI credit scoring
- Insurance — Detection of fraudulent insurance claims, product recommendation, automated payment decisions, risk early-warning
- Automotive & urban development — License plate recognition, parking DX (group company I-TECH), dangerous-driving scoring
- Education — Personalized question selection, automatic question generation, automated grading
- Medical & healthcare — Diagnostic imaging support, pharmaceutical marketing support, extraction of adverse-event information
02Function-based solutions
Cross-industry capabilities packaged as solutions.
- AI Security (PKSHA Security) — Fraudulent-use detection, insurance claim fraud detection, monitoring of inappropriate posts
- PKSHA LLMS — Integrated, company-specific customization of large language model (LLM) technologies such as GPT (released March 2023)
- PKSHA Private AI Agents — AI agents that run open models, including Japanese LLMs, in the client's own on-premises environment, with support from model selection through operational improvement (launched September 2026)
03Combining with AI SaaS
Use SaaS where packaged products are sufficient, and fill the gaps with customization.
- Customer experience (CX) — PKSHA ChatAgent, PKSHA FAQ, PKSHA VoiceAgent, PKSHA Speech Insight, PKSHA Knowledge Stream
- Employee experience (EX) — PKSHA AI Helpdesk, RoboOperator (ロボオペレータ), Yomel by PKSHA
If you are considering a chatbot on its own, see our PKSHA ChatAgent page as well. The PKSHA group also runs an "AI Powered Worker" business that combines AI with professional talent, and in 2026 launched "AI Powered PMO" (with group company X Capital) and an AI BPO service (with group company Triumph).
Pricing
Pricing is not disclosed (custom quote). As of October 2026, the official site does not list plans, monthly fees, or PoC cost guidelines. Because it is customized development, costs depend on the target operations, data, and duration.
| Item | Details (as of October 2026) |
|---|---|
| Price | Not disclosed (custom quote) |
| Price guideline | Not stated officially |
| Duration guideline | About 6–12 months for implementation steps 1–4 (sum of official guidelines for each stage) |
| Contact | Contact form on the official site |
Track Record and Case Studies
The scale figures published officially are: 4,400+ client companies (as of October 2025) and 7,000+ AI agents (no date given). The industry pages still show an older figure, "over 2,000 companies in 10 years since founding," so 4,400+ should be read as the latest. In its disclosures as a listed company, the AI Research & Solution business posted revenue of JPY 9,988.66 million for October 2025–June 2026 (up 31.4% YoY).
01Fraud detection (credit, insurance, online services)
- Credit Saison — Algorithms provided to fit the "ACEPlus" fraud detection system. The official site notes that in July–September 2020, about two years after introduction, card fraud losses were 49.8% of the previous year
- JACCS — Fully adopted after a proof-of-concept confirmed fraud-suppression effects beyond expectations
- E.design Insurance — Adopted for detecting fraudulent auto insurance claims
- Matching service operator — About 80% reduction in workload for post monitoring compared with human-only review (company name undisclosed)
02Forecasting and optimization for retail & distribution
- Asahi Group Holdings — Introduced the shelf planogram solution; the official site says the time required to create planograms is expected to be cut by about 65%
- Daytona International — "Marketing automation optimization AI" that predicts the best delivery timing and channel for each member; production operation scheduled to start October 1, 2026
- Effect examples (companies undisclosed) — Inventory efficiency of several hundred million yen or more per year via inventory optimization (retail); tens of millions of yen per year in automation via staffing optimization (services)
03Joint development in insurance design and healthcare
- Tokio Marine & Nichido Fire Insurance — PKSHA algorithms introduced into the insurance design process for "Cho Hoken"
- MedPeer Group (Mediplat) — Jointly developed an AI system that helps doctors draft answers in the occupational health service "first call"
Who It Suits and Who It Doesn't
It suits large companies that want to build forecasting, optimization, fraud detection, or LLM systems using their own data and take them all the way into production. In industries with many official case studies, such as finance, insurance, and retail, you can consult while checking comparable track records. For companies that cannot send confidential information to the cloud, PKSHA Private AI Agents running on-premises is another option.
It does not suit companies that want to compare prices in advance and place an order quickly, or those mainly seeking generative AI training for employees or hands-on help adopting tools. No price guidelines are published, and the official implementation steps proceed on a monthly scale. Training and in-house capability building menus are not described officially either. Since there is no mention of SMBs, smaller inquiries should confirm the scope of support when contacting PKSHA.
FAQ
Q. How much does PKSHA AI Solution cost?
It is not disclosed (custom quote). As of October 2026, the official site lists no amounts or PoC cost guidelines.
Q. Can small and medium-sized businesses use it?
There is no official mention of SMBs. The official site shows "Enterprise adoption: 70% or more," indicating support centered on large enterprises.
Q. Can development be outsourced as well?
Yes. The official implementation steps cover consultation, requirements definition, model building, API development, and production deployment support.
Q. How long does it take?
Adding up the official guidelines, it takes about 6–12 months before production deployment: consultation about 1 month, requirements definition 1–3 months, model building 3–6 months, and API development 1–2 months.
Q. How is it different from PKSHA's AI SaaS?
AI SaaS consists of packaged products for general needs, while AI Solution is customized to each company's specific challenges. PKSHA also proposes combinations of the two.
The information on this page is based on what was published on the official website as of October 2026. Please check the official website for the latest pricing and details.
