DNP cuts patent research time by 95% with ChatGPT Enterprise

Dai Nippon Printing (DNP), the 148-year-old Japanese printing giant, is showing what aggressive enterprise AI adoption actually looks like — and the numbers are frankly hard to ignore. After a strategic pivot in April 2023, the company deployed ChatGPT Enterprise to ten core departments and tracked metrics that most firms only dream about: a 100% weekly active usage rate and measurable results in 90% of use cases within three months.

The most dramatic gains came from the ICT R&D division, where patent research — traditionally a manual slog through databases — saw a 95% reduction in time spent while simultaneously expanding coverage by 10x. Yohei Ishida, who leads the P&I Innovation unit, told the company that patent applications previously “depended heavily on individual judgment, with standards varying by person and department.” ChatGPT Enterprise changed that, enabling objective decision-making that improved both the volume and quality of filings. Competitive analysis reports that once took hours are now generated as first drafts automatically, an 80% time savings.

But the real signal here isn’t just speed — it’s accessibility. DNP reported that employees with zero Python experience were able to generate and run code through the platform, compressing development work that would typically take over a year into just a few days. Researchers combined these new capabilities with domain expertise to surface insights they wouldn’t have found otherwise. The company also set surprisingly concrete usage benchmarks: 100 interactions per employee per week and a 50% automation rate for task time reduction.

On the IT governance side, Masahiro Kobayashi’s infrastructure team slashed security audit comparison time from 30 minutes to 5 minutes, and a cloud security check covering roughly 100 CIS Benchmark items went from two person-days to 10 minutes. DNP’s approach stands out because they made usage visible, encouraged experimentation, and let learnings spread organically through custom GPTs. The 70% knowledge reuse rate suggests these aren’t one-off wins — they’re compounding. Whether this pace is sustainable as the rollout expands beyond the initial ten departments is the question worth watching.