---
id: "lib-2026-10-23-40-ai-a-40-year-braised-pork-brand-moves-to-an-ai"
type: "library"
title: "靠一鍋滷豬腳飄香 40 年，如何走向 AI 智慧工廠？…「組織＞流程＞系統」/ A 40-year braised-pork brand moves to an AI factory: organisation before process before system"
lang: "ZH"
org: "領導影響力學院 (鄧師傅功夫菜)"
source_type: "Podcast"
date: "2026-10-23"
week: "2026-W43"
url: "https://nosov.ai/library/lib-2026-10-23-40-ai-a-40-year-braised-pork-brand-moves-to-an-ai/"
original_url: "https://www.youtube.com/watch?v=eBeccjsaIAs"
tier: "B"
tier_name: "Useful, limited"
tier_reason: "An interview with the CEO of a Taiwanese family food group giving a concrete, credible sequence from process mapping to ERP to sensors and AI tools; all figures are his own, and the channel promotes a paid course."
summary_basis: "transcript"
original_minutes: "37"
tags: ["business transformation", "process before ai", "food manufacturing", "erp"]
---

# 靠一鍋滷豬腳飄香 40 年，如何走向 AI 智慧工廠？…「組織＞流程＞系統」/ A 40-year braised-pork brand moves to an AI factory: organisation before process before system

Tier B (Useful, limited): An interview with the CEO of a Taiwanese family food group giving a concrete, credible sequence from process mapping to ERP to sensors and AI tools; all figures are his own, and the channel promotes a paid course.

Summary based on the full transcript.

## In short

A 37-minute Mandarin interview (23 October 2026) on the Global Views leadership academy channel with Bill Deng, CEO of Deng Shifu, a Taiwanese restaurant and packaged-food group founded in 1984 and now run by two brothers. His rule is "organisation before process before system". The firm spent more than two years writing down and standardising processes before ERP, then added sensors and AI tools in a new food factory. It has no in-house IT staff and runs everything as cloud services.

## Key numbers

- 2+ years: time spent standardising every process from purchasing to shipping before the ERP went live
- 3 years 4 months to 1.5 years: training time for a mid-level chef before and after process simplification (as little as 3 months in some cases)
- 32: refrigerators fitted with temperature and humidity sensors that replaced handwritten hygiene logs
- ~300 / ~400 / 1,000+: raw materials, finished goods and semi-finished items covered by an AI assistant on ERP data

## What it reports

- Before the handover, management was "rule by person": whatever the founder said. Deng came back from three years at Panda Express in the US, where he rose from store manager to area manager.
- In 2008 the firm centralised core dishes in a central kitchen; costs rose, but flavour became consistent and accountability clear.
- Process simplification cut training of a mid-level chef from about 3 years 4 months to 1.5 years, in some cases 3 months, and allowed more days off, which he credits for better retention.
- ERP preparation: inconsistent units (Taiwanese catties versus kilograms) and data silos; standardising processes from purchasing to shipping took more than two years. Some staff left.
- Common mistake he names: buying a system because others use it, then finding nobody can run it. Written processes were also needed because many staff are foreign and do not speak Mandarin.
- Executives must map processes themselves; one employee assigned alone stalls on cross-team friction.
- Factory of about 1,000+ ping: sensors on 32 refrigerators and workrooms send readings to the cloud for hygiene audits. AI form apps capture multilingual production and quality records for staff from six or seven countries. An AI assistant on ERP data answers cost and stock questions across about 300 raw materials, 400 finished goods and 1,000+ semi-finished items.
- Traceability down to the purchase date of an ingredient batch helped win large retail contracts.

## How they got it

A hosted interview; the guest describes his own company. No financials or before-and-after metrics beyond those above. The digest is based on uploader Traditional Chinese captions.

## Limits

- Self-reported by the CEO; no outside verification.
- The channel runs an advertisement for its own management and AI course inside the episode.
- AI tools are described at a general level; no usage or accuracy data.

## Why it matters for companies

A clear small-company case of the order that works: align the organisation, write the processes, then choose systems, and only then add AI on top of clean data. The training-time reduction shows that the payoff from process work can arrive before any AI is involved. The point that executives must lead process mapping in person applies to any firm whose AI projects stall on cross-department hand-offs.

## Key passages in translation

> 你流程有出來後我們才可以問下一個問題是那請問接下來我要買什麼系統來優化這個流程

Only once the process exists can we ask the next question: which system should we buy to improve this process?

> 以前培養一個廚師需要三年四個月現在變成是一年半甚至到三個月就可以培養出一個中間幹部的廚師出來

Training a chef used to take three years and four months; now it takes a year and a half, even three months, to train a mid-level chef.

> 你可以做錯然後去做修改可是你不能什麼都沒有做

You can get it wrong and then revise it, but you cannot do nothing.

Original: https://www.youtube.com/watch?v=eBeccjsaIAs