Why this repo exists.
anorthernman-developer/frontierorg-diary is not a product dump and not a finished transformation. It is a public lab for a fictional infrastructure holding group: sample data you can inspect, ideas argued in the open, methods you can reuse, and frameworks that refuse a fake win.
Open the repository Read the diary
Why GitHub
The site tells the story. The repository is the evidence. If a sentence on frontierorg.work cannot be traced to a file, a fixture, or an explicit “not measured”, it does not belong in a board pack.
- Provide sample data. Synthetic humans, cases, and invoices. Regenerate them. Do not treat a row count as a saving.
- Explore ideas in the open. The diary records the argument. The issues record what is still unresolved.
- Create methods. Gates, scorecards, an org map, and occupational bridges. They are working papers, not a certification.
- Share frameworks. Approaches to AI adoption that hold cost, risk, and engagement and enablement in the same view. Pause is an allowed outcome.
Sample data
Everything named as a person, a company, or a system of record in this work is fictional. Emails stay on example.com. The directory is simulated only. There is no live tenant.
| Path | What it is | What it is not |
|---|---|---|
| data/sample/ | One service-promise join, scenario SCN-61003, seed 20261003. | not a measured outcome |
| data/sample_large/ | 2,000 synthetic humans and matching CRM- and SAP-style rows. | not a company |
| data/sample_full/MANIFEST.md | Row counts from a local 50,000-human run. The CSVs are not in git. | not a directory |
| data/onet/ and data/generated/ | O*NET 31.0 slice, task text, and keyword workload labels. | not a job evaluation |
| data/esco/ | Official ESCO v1.1.0 crosswalk to O*NET-SOC 2019, where a row exists. | not ESCO v1.2.1 |
Regenerate from the repository root. Another seed changes identifiers. It does not rewrite the diary.
Attribute the O*NET database (CC BY 4.0) if you reuse occupational text. See data/onet/ATTRIBUTION.md and data/esco/ATTRIBUTION.md.
Methods
These are the artefacts a consultant can lift. None of them is permission to deploy an agent.
- Engagement gates. A step does not start because a tool exists. Engagement and docs/engagement/gates.md.
- Scorecards. C-suite, operational, and business-area views. Actuals stay not measured until a cell cites a fixture and says it is a count. Scorecards.
- Org map. Occupations and tasks, not a reporting line. Department tags are a heuristic. Org map.
- O*NET and ESCO bridges. Codes and titles so work can be compared. A URI is not a consultation. People.
- Controls. Licence, region, and retention named or left unknown. Controls.
- Diary and issues. The argument, including the pauses. Diary. Standing issues: GitHub issues.
Frameworks
AI adoption here is judged on three weights at once: cost, risk, and engagement and enablement. A hole in one is a pause, not a footnote. This page does not report a return on investment. There is not one.
| Weight | The question | Where it lives | Actual |
|---|---|---|---|
| Cost | Can finance defend the spend, including the unknowns? | Scorecards · COST-001 | not measured |
| Risk | Are licence, region, retention, and safety-critical work named before anything acts? | Controls · RISK-001 · RISK-002 | not measured |
| Engagement and enablement | Do named people know what would change, and have they been asked? | People · PEOPLE-001 · PEOPLE-002 | not measured |
How we use those weights
- Pause beats a fake win. An empty cost envelope, an unnamed region, or a workforce that has not been consulted stops the story. It does not get a percentage.
- Human gates. A draft may be written for a person. It may not send, book, pay, or switch a network. Safety-critical and payment work stays human-only unless a named reviewer says otherwise in writing. That has not happened.
- ADKAR-style, not an ADKAR claim. Awareness, desire, knowledge, ability, and reinforcement are a useful order of questions. We have not run that programme. Day 12 is a scenario script. Courses run: zero. Workforce consultations: zero.
- The board triad still stands. Customer outcome, employee transition, and risk and cost weigh the same. Engagement and enablement is how we talk about the people weight. It is not a fourth score we have calculated.
How to reuse
- Read the simulation banner. Do not present FrontierOrg, Northwater Distribution, or the personas as a client.
- Clone the repository. Regenerate a sample with the published seed if you want the diary’s identifiers.
- Lift a method — a gate, a scorecard shell, a workload label — and keep the “not measured” cells until you have your own evidence.
- Do not copy a keyword label into a redundancy plan. Do not backfill a workshop.
- Attribute O*NET and ESCO. The narrative and the generators are MIT. See the licence.
- Maintainers: follow docs/HANDOVER.md after each day. Craig approves diary words before they are the public entry.