company-operating-system

18 · The AI Era: AI Practices for Every Layer (AI-Native Organizations)

In 2026, the management paradigm is shifting from “people + tools” to “people + AI + tools.” This chapter answers two questions: ① Within this book’s eight-layer framework, where can each layer adopt AI right now? ② What does an “AI-native organization” look like, and how do you get there step by step?

The author is himself an AI-organization practitioner (this repo is maintained jointly by a human founder + AI collaborators, with 20+ automated tasks deployed).


0. The Honest Bottom Line


1. The Eight-Layer Framework × AI Practice Map

For each layer: what AI can do / which tools / maturity (🟢 production-ready · 🔵 semi-mature · 🟠 early stage)

① Strategy & Steering (strategy / budget / KPIs & performance / review)

AI practice What it does Example tools Maturity
Market intelligence The macro/industry/competitor views of the Five Views — AI auto-aggregates news, research reports, and competitor moves Intelligence radar (TrendRadar), AI search, research report summaries 🟢
Scenario planning AI generates 2-4 future scenarios + contingency plans for each LLM + prompt frameworks 🔵
Competitive monitoring Auto-tracks competitor pricing / new products / hiring / funding Crawlers + LLM summaries 🟢
Budget simulation What-if: change one parameter and see the impact on net profit BI + natural-language LLM queries 🔵
Review minutes Monthly operating meetings auto-transcribed + variance attribution + action-item extraction Speech-to-text + LLM 🟢

Case: within the “Five Views” of Huawei’s (Chinese technology giant) BLM, AI can significantly automate the initial gathering and summarization of macro, industry, and competitive information; information verification, customer insight, internal capability assessment, and strategic trade-offs still require humans.

② Governance (equity / ring-fencing / compliance)

AI practice What it does Maturity
Contract review AI scans equity agreements, articles of association, and contracts for abnormal clauses 🟢
Compliance checks AI checks whether related-party transaction pricing deviates from market prices (transfer-pricing red line) 🔵
Board materials Auto-generates board reporting decks / monthly operating reports 🟢
Cross-border regulatory monitoring AI tracks regulatory changes in target countries (e.g., PMTA/TPD for e-cigarettes) 🟢

③ Value Chain (R&D / production / supply / sales / service)

Stage AI practice Example tools Maturity
Sales/business AI auto-follows up leads, writes quotation emails, builds customer profiles CRM AI, email AI 🟢
R&D AI-assisted design (structure/circuit/code), patent search, competitor teardowns Copilot, AI retrieval 🟢
PMC AI predicts production scheduling, delivery dates, and material-kit completeness warnings Forecasting models + ERP data 🔵
Production AI visual QC (defect detection), predictive equipment maintenance Industrial vision models 🟢
QC Auto-generated inspection reports, customer-complaint classification and attribution LLM + vision 🟢
Customer service 24/7 AI customer service + automatic complaint classification/escalation Chatbots 🟢

Highest-priority pilots for manufacturers: visual QC + complaint classification + delivery-date prediction (highest ROI, fastest to implement).

④ Organization (departments & collaboration)

AI practice What it does Maturity
AI-agent employees AI agents join departments as “virtual employees” (e.g., an “AI buyer,” an “AI order tracker”) 🔵
Process automation RPA + AI handles documents / approvals / reconciliation automatically 🟢
Organizational knowledge base A “company brain”: all SOPs, documents, and experience fed to AI; employees Q&A anytime 🟢
Meeting minutes Weekly/review meetings auto-minuted + task assignment 🟢

⑤ Resource Foundation (people / finance / materials / data)

Resource AI practice Maturity
People AI resume screening, interview-assessment support, onboarding Q&A 🟢
Finance AI invoice recognition and booking, reconciliation, cash-flow forecasting, cost-anomaly alerts 🟢
Materials AI sourcing and price comparison, supplier risk monitoring, supply-interruption warnings 🔵
Data Natural-language BI queries (ask a question, get a report), automated data-quality checks 🟢

Case (this book’s own practice): the author’s team has turned “data” into an AI cockpit — revenue/margin/inventory/cash flow are aggregated automatically every day and pushed to Feishu (Lark, the Chinese workplace collaboration platform), where AI generates a daily operating report with anomalies flagged in red.

⑥ Moat

AI practice What it does Maturity
R&D speed AI-assisted R&D = patents/new-product iteration at double speed 🟢
Data moat Business data + AI analysis = insights competitors don’t have 🔵
Operating efficiency AI cost reduction = cost advantage = one of the hardest moats 🟢
Personalization AI delivers one-to-one service/product configuration for every customer 🔵

Note: AI is a “tool”; moats are still built from “people + organization + data.” AI strengthens a moat but cannot conjure one out of thin air.

⑦ Risk & Compliance

AI practice What it does Maturity
Anti-fraud / anomaly detection AI flags abnormal transactions, expense claims, and inventory shrinkage 🟢
Compliance monitoring AI scans contracts/documents for compliance 🔵
Sentiment / crisis early warning AI monitors brand sentiment, competitor negatives, and industry risks 24/7 🟢
Audit support AI sample selection, discovery of anomalous audit leads 🔵

⑧ Decision Engine

AI practice What it does Maturity
Decision advisor AI aggregates information → presents options/trade-offs/evidence chains to the CEO 🟢
Red-team challenger AI plays devil’s advocate: points out biases and blind spots in your decisions 🔵
Real-time cockpit AI answers “what’s the state of the company right now” anytime 🟢
Post-decision analysis AI tracks outcomes and tests assumptions after a decision 🔵

Key point: AI is the advisor, not the commander. Final decision authority, accountability, and trust must remain with people.


2. AI-Native Organizations: Three Forms

The three forms are not a timeline or a linear progression — a single company can combine them: customer service on AI agents, finance still on copilots, R&D already at process-level automation. The distinguishing dimension is “how deeply AI participates in the business.”

Form 1: AI-Enhanced (tool augmentation)

Traditional organization + AI tools
Humans do everything; AI makes them faster
e.g., employees use AI to write emails, build reports, look things up

How to tell: AI is a personal tool — no persistent state, no direct calls into business systems, decision authority unchanged.

Form 2: AI-Collaborative (role-level collaboration)

Humans + AI agents form the team together
AI is a "member": it has its own role, tasks, and deliverables
e.g., an AI sales assistant auto-follows leads, an AI QC inspector watches the line 24/7,
an AI finance assistant reconciles accounts

How to tell: AI has a role and long-term context, can call into some business systems to execute bounded tasks, and key actions require human review.

This repo’s author’s practice: this repo is maintained jointly by a “human founder + AI collaborators” — AI handles content drafting, deployment reviews, and document generation; humans handle judgment, direction, and final approval. It is a minimal specimen of Form 2.

Form 3: AI-Native (process/organization redesign, exploratory stage)

Core processes, data, and accountabilities are designed around human-AI collaboration
AI drives bounded processes; humans supervise, decide, and handle exceptions
e.g., AI executes bounded tasks within functional processes like sales/customer service/finance,
while humans manage goals, permissions, and red lines

How to tell: core processes are redesigned around human-AI collaboration; AI has persistent state, can execute actions, and operates within permission boundaries with audit logs.

⚠️ Honest warning: Form 3 today is mostly “demos run, production unverified.” Production deployment should start with a single process, least privilege, and reversible actions — don’t let AI autonomously run cross-department operations from day one.

Open-Source Exploration Projects (snapshot 2026-08-12; ⭐ = that day’s value)

Repo Claimed positioning Actually public components Evidence level
Claw-Company/clawcompany 583 AI company OS: 38 roles / 6 templates / 4-layer memory Role templates + memory system, TypeScript README/demo
getnao/sylph 178 Company brain: AI agents + skills + self-improvement MCP integration + agent framework, JavaScript README/demo
Lifecycle-Innovations-Limited/claude-ops 114 Claude Code enterprise operations: 57 skills / 21 agents Claude Code skill pack + integration README/demo
yomidenzel/BOS 130 Turn Claude into a COO (10 AI skills) Claude Code skill pack (French) README/demo
felipeluissalgueiro/hive 20 Run a company with Claude Code squads Multi-agent harness, HTML Early prototype
10Legs/ceo-f500-harness 8 15 executive agents + decision-permission matrix Agent prompts + workflows Early prototype

⚠️ A note on evidence: ⭐ is a same-day snapshot and does not indicate adoption or production maturity. The “evidence level” column is based on public READMEs/directories/examples; whether something is production-ready must be verified by you (tests, deployment docs, production case studies).


3. The Four Steps from Traditional to AI-Native Organization

Step 1: The Data Foundation (0-3 months)

Without data, AI spins its wheels. Build the “data” foundation first (Ch. 05):

Step 2: Pilot 3 High-ROI Scenarios (3-6 months)

Prioritize scenarios that are “high-frequency, repetitive, data-backed, and fault-tolerant”:

Scenario Why first Time to results
AI customer service / complaint classification High frequency, data already available 2-4 weeks
AI daily operating report / review Simple, immediately useful 1-2 weeks
AI visual QC Highest ROI 1-3 months
AI sales-lead follow-up Direct revenue upside 1-2 months

Step 3: From “AI Tool” to “AI Team Member” (6-12 months)

Step 4: Build AI Governance (in parallel with pilots, not after)

⚠️ Governance is not something you start at step 4 — establish minimal governance before the first pilot, or AI will enter the business ahead of the rules. Step 4 is about systematizing governance.

Cross-Cutting AI Capabilities (infrastructure shared by all layers)


4. Ready-to-Use Templates: AI Practice Checklists

Department-Level AI Pilot Application Form

Department: __________
Pain point (one sentence): __________
What AI can do: __________
Data required: __________
Estimated savings in people/time: __________
Risks & fault tolerance: __________
Pilot duration: __________
Reviewer: __________

Human-AI Division of Labor

Task Human AI assists AI fully automated Reviewer
Strategic direction ✅ intelligence
Customer quotes ✅ final call ✅ price suggestion CEO
Complaint classification QC department
Daily operating report Finance
Large purchases ✅ approval ✅ price comparison General manager
Contract initial review ✅ final review ✅ pre-review Legal / owner

5. Risks and Boundaries (must be honest)

Risk Description Countermeasure
AI hallucination AI confidently makes things up Key decisions must be human-verified + give AI data sources
Data security Trade secrets fed to cloud AI Use local/private deployment for sensitive data (e.g., Ollama)
Over-reliance Human judgment atrophies Schedule regular “no-AI days” to exercise judgment
Agent going rogue AI does things it shouldn’t Permission boundaries + audit logs + red-line checklist
Fake efficiency Demos look good but aren’t really used Pilots must quantify ROI; cut them if they fall short

6. One-Line Summary

An AI-native organization is not about “replacing people with AI” — it’s about using AI to free human energy from repetitive work and focus it on judgment, relationships, and innovation. Start with the data foundation, pilot 3 high-ROI scenarios, then make AI a team member, and finally build AI governance. A company that skips AI practice in 2026 is like a company that skipped smartphones in 2010.


Contributing

PRs welcome: