Why Generic AI Fails Companies: Building AI With Real Philosophy
Nishiyama Asako (THA) and Ozawa Kensuke (AICX Association) argue that AI without embedded company philosophy produces hollow results, and that context engineering will define the next era of enterprise AI.
- Ozawa: the vast majority of generative AI services launched after ChatGPT’s November 2022 debut were built without philosophy and have not demonstrably doubled any company’s revenue.
- THA’s AI Shachou encodes a CEO’s values, decision criteria, and 5–10 year vision through deep interviews — not just document uploads — so the AI reflects the company’s philosophy, not generic LLM behavior.
- Nishiyama reports THA itself grew 3x while using AI Shachou internally; new employees send daily reports to both a human mentor and the AI, which returns value-aligned feedback.
- Google’s Gemini Enterprise now lets admins control all employee-facing prompts centrally, signaling a broad industry shift toward corporate-personalized AI over raw chatbots.
- Ozawa predicts the next bottleneck for enterprise AI is not model intelligence (GPT-5 is already sufficient) but the cleanliness and organization of internal company data.
- Ozawa coins the term context engineering — pre-loading AI with company background, dialogue history, and live internal documents — as the successor to prompt engineering, enabling vague questions to get precise answers.
- THA’s moat is extracting tacit, offline, unlabeled knowledge (philosophy, judgment calls) that does not exist in Google Drive, requiring high-touch CEO interviews rather than automated ingestion.
- A construction company case: THA interviewed 5 senior foremen, extracted their combined tacit expertise, and encoded it into a single AI so junior staff could access knowledge those foremen were too intimidating to approach directly.
2025-12-05 · Watch on YouTube
