An open playbook for leading AI and data work from a business decision to a reliable production workflow. Built for heads of data, AI leaders, managers, and practitioners who need results people can trust.
Use the operating models, templates, and synthetic examples to choose useful work, establish decision rights, validate outcomes, and improve the next run. This repository remains the home of AI+Data; its existing paths stay available.
| What you need to do | Start here | Leave with |
|---|---|---|
| Lead an AI transformation program | AI leadership operating guide | A staged portfolio, evidence gates, and review cadence |
| Establish your first 90 days | CDO operating system | A roadmap, KPI tree, and governance model |
| Choose what deserves a pilot | Workflow investment brief | A bounded decision, baseline, owner, and stop rule |
| Make ten recurring metrics trustworthy | Ten-metric pilot | A small trust layer with an inspectable finish line |
| Turn analysis into a decision | Decision memo loop | Evidence, options, a recommendation, and follow-through |
| Check a complete example | Synthetic funnel | Sample data, SQL, and a worked analysis |
| Explain progress to leadership | Board brief | Outcomes, limitations, tradeoffs, and decisions needed |
For a short guided read, use the executive reading path. For a practical exercise, use the synthetic funnel example.
Clone or open this repository, then give your file-capable agent a bounded request:
Read docs/ai-leadership-operating-guide.md and
toolkits/workflow-investment-brief.md.
Help me choose one recurring workflow to pilot.
Ask for the decision, available evidence, permissions, and constraints.
Use placeholders for unknowns. Do not invent current company facts.
Draft the investment brief in my private workspace, with a baseline,
review gate, stop rule, and the next action I can take this week.
Keep real company context and completed briefs in your own private workspace. These files guide the work; they do not connect to business systems or grant authority to act.
- Choose the decision. Start with a recurring business problem and the person who must act on the answer.
- Define trusted context. Establish metric definitions, approved sources, permissions, freshness, and caveats.
- Prove one workflow. Run the real path, compare against a baseline, and inspect errors and review effort.
- Earn wider use. A working solo run comes before a team handoff; a working team pilot comes before shared infrastructure.
- Make corrections durable. Review the miss, change the relevant definition or rule, and rerun a regression case.
These are reusable operating practices, not a claim that one architecture fits every organization. Examples labeled synthetic are teaching material, not company results.
Playbooks · Toolkits · Projects · Examples · Leadership docs · Research · Public writing
The public writing archive records 34 channel publications through August 3, 2026. Its dates and scope are retained; this repository refresh does not make the archive current through today. Inspect the catalog.
| Project | Use it for |
|---|---|
| World-Class AI + Data Operator — this repository | Leadership, operating models, governance, and adoption |
| ChatData | Hands-on analytical skills, runnable checks, and reusable analysis records |
| Your Voice | Writing that preserves the writer's meaning and voice |
| Company Stack | Principles and workflows for operating a digital product business |
Read the evidence boundary, source inventory, and publication standard. Public source pointers may identify their original publishers; the reusable leadership guidance does not disclose private employer plans, interviews, or customer details.
Contribute a concrete failure, a generalized operating improvement, or a synthetic example with a check another person can repeat. Explain what changed, what supports it, and where it does not apply. Keep private data and completed company templates out of issues and pull requests.
Curated by Paras Doshi. Read Insight Extractor for the public writing behind these ideas, or the editorial perspective for the scope of this collection. MIT license.