Knowledge preservation powered by ChatGPT

成立于1876年， DNP 全名为 Dai Nippon Printing Co., Ltd.，是全球最大的印刷公司之一，全球雇员超过37,000人。公司业务涵盖 Smart Communication、Life & Healthcare 与 Electronics 等领域，以品牌口号 " Creating future standards. " 为指引，致力于连接人和社会并推进可持续发展。

作为承诺的一部分， DNP 长期拥抱新兴技术。2023年4月，公司决定在全组织范围采用 AI；到5月，已构建起企业级的安全使用环境；2025年2月， DNP 在十个核心部门部署了 " ChatGPT Enterprise "。在三个月内取得的成果包括：

• 90% 的使用场景出现可量化成效
• 100% 的每周活跃使用率
• 87% 的时间缩减自动化率
• 70% 的知识复用率（自定义 GPT）
• 处理量增长 10 倍

通过战略部署加速普及

为充分发挥生成式 AI 的价值， DNP 将首批落地目标锁定为影响力最大的十个部门，并设立明确基准：每位员工每周至少使用 " ChatGPT Enterprise " 100 次，且任务时间缩减的自动化率要超过50%。

“我们通过让使用情况可视化来推动普及，” ICT Control Office、R&D and Engineering Management Division 总经理 Hiroyuki Otake 说。每个团队反复试验、分享经验并不断迭代，这种势头形成了可扩展的影响力。结果是，个别改进通过自定义 GPT 和共享用例在团队间传播，形成了当前推动业务转型的核心模式。

在引入 " ChatGPT Enterprise " 的部门中，影响最大的当属 ICT 研究与开发部。 Advanced Business Center 下 P&I Innovation Research and Development Unit 总经理 Yohei Ishida 带领团队用 AI 自动化并优化了专利检索与申请策略，取代了大量手工工作。

他的团队用 " ChatGPT Enterprise " 构建了以下流程：

• 专利检索：自动化搜索、摘要与分类，将检索时间缩短95%，覆盖面扩大10倍
• 申请策略：识别 DNP 技术与竞争对手专利的关键差异，降低被驳回风险并减少修改次数
• 竞争分析：自动生成初稿报告，将准备时间缩短80%

Ishida 指出，影响不仅限于效率，还提高了质量。“过去专利申请高度依赖个人判断，不同人和部门的标准各异。有了 ' ChatGPT Enterprise '，我们能做出更客观的决策，提升了申请的数量与质量。”通过提升知识产权战略， DNP 正在夯实产品独特性与长期竞争力的基础。

零基础构建 Python 脚本

负责推进生产技术 QCD（质量、成本、交付）创新的研究部门，利用 " ChatGPT Enterprise " 在需要高级分析与评估的工作上，大幅压缩了传统所需时间，例如实验设备操作、测量与数据分析。

关键成果包括：

• 将英文专利与设备原理的信息结构化从数月缩短到3天
• 让无 Python 经验的员工通过 " ChatGPT Enterprise " 生成并运行代码

一个突出的案例是，完全不懂 Python 的员工在零学习成本下生成代码并完成数据分析。传统上需要一年多的开发工作，在几天内就得以实现。将这些能力与研究人员的专业知识结合后，带来了新的洞见，并对全部门产生了显著影响。

“即便是不擅长或对 IT 有抵触的人，也能有效使用 ChatGPT 并取得显著成果，” Technology Development Center 下 Integrated Manufacturing Innovation Laboratory 主管 Takamasa Yoshizawa 与 Evaluation and Analysis Research Institute 的 Michiko Ito 指出，他们认为以 Python 代码进行数据分析是最成功的用例。

强化 IT 合规与云端运维

DNP 正利用 " ChatGPT Enterprise " 现代化 IT 治理。 Information Innovation Operations、ICT Center 下 System Infrastructure Development Division 总经理 Masahiro Kobayashi 强调，许多曾经手工且不一致的任务得到了改善：

• 外部安全审计：审计比对时间从30分钟降至5分钟；密码套件选择时间从3小时降至1小时
• 云安全：约100项未符合 " CIS Benchmark " 的初步检查，从两人天缩短到10分钟完成
• 评审支持：通过参考设计政策与历史记录，将需求评审时间从1小时缩短到30分钟

“模型在收集相关资料并生成清晰输出方面表现出色，” Kobayashi 说，“这让团队能把注意力放在决策上，而不是文件比对。”他同时强调 AI 无法取代人类监管：“验证与最终把关仍然是人的责任。”

用 AI 保存组织知识

知识流失是 DNP 面临的重大挑战：专业知识常常停留在资深员工脑中或埋藏在纸质文件里。 Advanced Business Center 下 AI Business Development Unit 的 Technology Development 总经理 Isaku Osawa 带领团队，用 AI 直接应对这一问题。

他们利用 " ChatGPT Enterprise " 将纸质手册和历史质量记录等非结构化资料进行结构化与数字化，一旦被摄入，这些记录就成为可通过自定义 GPT 访问的内部知识库。定义数据架构所需时间缩短了90%，可审阅的技术论文数量也翻了一番。

“我们的目标是把代际传承的知识变成数字化的劳动者，” Osawa 说。这不仅能缓解人手短缺，也为长期创新能力打下基础。

为 AI 原生业务运作夯实基础

“AI 代理将无缝融入各种场景，让每个人在几乎不自觉的情况下从中受益，” Otake 预见未来将从人机协作，走向由 AI 与 AI 间交互承担部分业务的基础层面。随着机器人技术进步，这一趋势会加速，最终形成物理层面的 AI 在现实世界中运行的局面。

展望未来， Otake 强调知识保存的关键性：“我们必须把为人设计的信息转化为 AI 能理解的格式，确保知识被保存与共享。目标是在劳动力萎缩的背景下提升生产力。”其核心是在前线将经验与质量记录编码为结构化数据，使 AI 代理与未来的物理 AI 能学习并应用，减少对个体专长的依赖，从而转化为持久的竞争优势。

秉承品牌口号 " Creating future standards. "， DNP 希望在印刷与信息技术优势的基础上，转型为一家 AI 原生企业，推动新的社会标准。

Founded in 1876, Dai Nippon Printing Co., Ltd. (DNP) is one of the world’s largest printing companies, employing over 37,000 people globally. With a portfolio spanning Smart Communication, Life & Healthcare, and Electronics, DNP is guided by its brand statement, “Creating future standards.” and a commitment to connect people and society while advancing sustainability.

As part of this commitment, DNP has long embraced emerging technologies. In April 2023, the company made a strategic decision to adopt AI across the organization. By May, DNP had built a secure environment for enterprise-wide use. In February 2025, the company launched ChatGPT Enterprise across ten core departments. Within three months, results included:

• 90% of use cases with ChatGPT Enterprise showed measurable results
• 100% weekly active usage rate
• 87% automation rate in time reduction
• 70% knowledge reuse rate (custom GPTs)
• 10x increase in processing volume

Accelerating adoption through strategic deployment

To fully realize the benefits of generative AI, DNP targeted ten departments with the highest potential impact. The company established clear benchmarks: each employee should use ChatGPT at least 100 times per week, and over 50% automation rate for task time reduction.

“We drove adoption by making usage visible,” says Hiroyuki Otake, General Manager of ICT Control Office, R&D and Engineering Management Division. “Each team experimented, shared learnings, and iterated. That momentum created a scalable impact.” As a result, individual improvements spread across teams through custom GPTs and shared use cases, forming core patterns now driving business transformation.

Cutting patent research time by 95%

In the departments where ChatGPT Enterprise was introduced, the greatest impact was seen in the ICT research and development division. Yohei Ishida, General Manager of P&I Innovation Research and Development Unit, Advanced Business Center, led his team to automate and improve patent research and filing strategies, replacing manual tasks.

His team built the following workflows using ChatGPT Enterprise:

• Patent research: automated search, summarization, and classification, cutting research time by 95% and expanding coverage 10x
• Application strategy: identified key differentiators between DNP’s technology and competitors’ patents, reducing rejection risk and minimizing revisions
• Competitive analysis: generated first-draft reports automatically, reducing preparation time by 80%

Ishida notes that the impact goes beyond efficiency to quality. “In the past, patent applications depended heavily on individual judgment, with standards varying by person and department. With ChatGPT Enterprise, we can now make objective decisions, which has improved both the volume and quality of our filings.” By elevating IP strategy, DNP is strengthening the foundations of product uniqueness and long-term competitiveness.

Building Python scripts with zero prior experience

DNP’s research division promoting production technology advances QCD (quality, cost, delivery) innovation to enhance the value of existing products and services, and pursues the development of new products and services. In areas that require advanced analytical and evaluation techniques, DNP has significantly reduced the time traditionally needed for tasks such as operating experimental equipment for material evaluation, conducting measurements, and performing analyses by leveraging ChatGPT Enterprise.

Key outcomes include:

• Structuring information from English patents and equipment principles in three days, down from several months
• Enabling employees with no Python experience to generate and run code through ChatGPT Enterprise

A particularly notable use case involved employees with no prior experience in Python, who were able to generate code and analyze data without any learning cost. Development work that would traditionally take more than a year was implemented within just a few days. By combining these capabilities with researchers’ expertise and knowledge, new insights were discovered, delivering significant impact across the entire division.

“Even those unfamiliar with IT or hesitant about it have used ChatGPT effectively and achieved significant results,” said Takamasa Yoshizawa, Director of Integrated Manufacturing Innovation Laboratory, Technology Development Center, and Michiko Ito from Evaluation and Analysis Research Institute. They highlighted data analysis with Python code as the most successful use case.

Enhancing IT compliance and cloud operations

DNP is modernizing IT governance with ChatGPT Enterprise. Masahiro Kobayashi, General Manager of System Infrastructure Development Division, ICT Center, Information Innovation Operations, highlighted improvements in tasks that were once manual and inconsistent:

• External security audit: cut audit comparison time from 30 minutes to 5 minutes; reduced cryptographic suite selection from 3 hours to 1 hour
• Cloud security: completed the initial check of ~100 CIS Benchmark noncompliance items in 10 minutes instead of two person-days
• Review support: shortened requirement reviews from 1 hour to 30 minutes by referencing design policies and past records

“The model excels at collecting relevant data and generating clear output,” says Kobayashi. “That allows our teams to focus on decision-making instead of document comparison.” He adds that AI will not replace human oversight: “Verification and final checks remain the responsibility of people.”

Preserving institutional knowledge through AI

One of DNP’s biggest challenges is knowledge loss. Expertise often lives in the minds of experienced employees, or buried in analog documents.

Under the leadership of Isaku Osawa, General Manager of Technology Development at the Advanced Business Center’s AI Business Development Unit, DNP is now using AI to address this issue head-on.

His team uses ChatGPT Enterprise to structure and digitize unstructured data from paper manuals to historical quality logs. Once ingested, these records become part of an internal knowledge base that anyone can access via custom GPTs. The time required to define the data architecture was cut by 90%. The team also doubled the number of technical papers they could review.

“Our goal is to turn generational knowledge into digital labor,” Osawa says. That shift not only offsets labor shortages but builds long-term capacity for innovation.

Building a foundation for AI-native business operations

“AI agents will blend seamlessly into various situations, allowing everyone to benefit from AI without even being conscious of it,” says Otake. He envisions a shift from human and AI collaboration to a foundation where parts of business run through AI to AI interaction. As robotics advances, this trend will accelerate, leading to a future where physical AI works in the real world.

Looking ahead, Otake emphasizes that knowledge preservation will be critical: “We must convert information created for people into information AI can understand and ensure that knowledge is preserved and shared. Our goal is to improve productivity as we prepare for a shrinking workforce.” The aim is to codify frontline know-how and quality records into structured data so that AI agents and future physical AI can learn and apply them, reducing reliance on individual expertise and turning it into an enduring competitive advantage.

Under its brand statement, “Creating future standards,” DNP seeks to expand strengths in printing and information technologies and transform into an AI-native company that generates new standards for society.

Interested in learning more about ChatGPT for business?

Generated by RSStT. The copyright belongs to the original author.

Source