Volume I: Architecting Work in the Ai-Native Enterprise
Ten Foundational Papers, published in four acts. The order is intellectual, not chronological: break the old model, build the economic case, rebuild work, then earn permission to deploy it. Each paper does one distinct job and leaves behind a piece of usable architecture before the next one builds on it.
Every paper below is pre-publication. None has shipped yet — this index is the honest state of the programme, not a preview of what's coming soon.
The paper index
Field code, title, and the one job each paper is written to do — the set as it stands, sheet by sheet.
Act I — Break the old model
The 40-Hour Workweek Turns 100
Why the architecture of work deserves reconsideration.
The Human Workday
Ai is changing what human time is for.
Act II — Build the economic case
The Productivity Dividend
When Ai gives us capacity back, where does it go?
The Human Capacity Problem
Work can be standardized. Human capacity cannot.
Act III — Rebuild work
The Death of the Job Description
The job description can no longer serve as the operating architecture of work.
The Workforce Without Headcount
Why capacity should come before FTEs.
The New Manager
From supervising people to governing human–machine systems.
Act IV — Earn permission to deploy it
The Accountability Architecture
What replaces activity when humans and machines produce the work together.
The Human Override
Where machine authority ends and human accountability begins.
The Human Workday Constitution
Thirteen binding rules for architecting work in the Ai-native enterprise.
Paper X is the Constitution. It publishes last, deliberately — every Article is meant to read as inevitable by the time it arrives, not as a philosophical opening argument.
Two forms, one adversarial case
Every paper in this index is being prepared in two forms: the Foundational Paper — long-form, referenced, 3,500–6,000 words — and a two-page Executive Brief for a reader with ten minutes rather than sixty. Every paper also carries its own adversarial case built into the document itself: the strongest argument against its position, printed inside it, not left for a critic to raise later.
How a claim earns its grade
Every substantive claim in a Human Workday document carries a level, printed in the text itself — not filed away in an appendix a reader will never open.
- A · Established
- Multiple high-quality studies, or regulatory or technical consensus. Stated plainly, and cited.
- B · Emerging
- Credible empirical research with limited replication or scope. Stated with its qualifiers attached, not as settled fact.
- C · Hypothesis
- A logical extension from the evidence, not yet demonstrated. Framed explicitly as a hypothesis, never as a finding.
- D · Design principle
- A normative Human Workday decision. Presented as a decision, with reasons — never dressed up as an empirical fact.
A research house that grades its own claims — and says plainly when a position is a decision rather than a finding — is more useful to a skeptical reader than one that presents every sentence with equal confidence. The grade is not a hedge. It is where the claim keeps its receipts.
The applied evidence system
We don't ask you to trust the diagram. Every consequential architectural claim should be challengeable.
The scale above grades a claim about the world. Applying an architecture inside a real organization adds a second decision: who or what actually does a piece of work — a person, a system, an Ai model, or some mix of them. That decision earns its own grade, its own record of what argues against it, and its own condition for being reopened — not just a citation.
- Evidence Grades
- How strong is the underlying evidence? The A–D scale above, applied to a claim about the world.
- Delegation Grades
- How established is the proposed allocation? A separate scale — E0 assumption, E1 indicative, E2 observed, E3 established under stated conditions — for a design decision about who does the work, not a fact about the world.
- Counter-cases
- What evidence argues against the architecture? Recorded at full strength, including the case that doesn't have a clean answer yet — never a strawman built to be defeated.
- Durability Conditions
- When should the decision be reconsidered? Established does not mean permanent — a model change, a workflow change, or an error rate crossed all count.
- Architecture Decision Records
- Why was the work designed this way? The decision, its evidence grade, the alternatives that were rejected, and the condition that would reopen it — on the record, not asserted from memory.
Inspect the research → Private repository
The Research Ledger, the counter-cases, and the Architecture Decision Records behind these grades live in a private repository while the corpus is still being built. This becomes a working link when it opens — not before.
We publish our weaknesses.
When a critic mounts a serious challenge to a claim or an architecture element and it survives scrutiny, the architecture changes — and the change is disclosed, not buried.
The public record
Two instruments make the evidence discipline checkable rather than declared. The Research Ledger logs every source Human Workday cites, alongside its methodology, sample, and limitations — not just the finding it's used to support. It updates in the open, and changes to it are visible in its own history.
The adversarial audit series records the other half of the discipline. When a critic — internal, external, or the Adversarial Review Panel — mounts a serious challenge to a claim or an architecture element, the audit is numbered (HWR-001, HWR-002, and onward) and records the challenge at full strength, the analysis, and the outcome: withstood, revised, or still open. A revision cites the audit that caused it.
Unresolved questions
Redesign creates new problems too.
The Five HR Jobs series — the applied work this evidence system was built to carry — closes each redesign with its own counter-case section, and names one entry the strongest of the set: if transactional work is how junior generalists build the pattern library that makes senior judgment good, delegating that work erodes the pipeline that produces the judgment it was training.
The apprenticeship problem
If Ai consumes apprenticeship work, where does expert judgment come from ten years later?
The series calls this counter-case unresolved, not answered: sampled review and routine-tier validation work are partial substitutes for that exposure, but whether they carry the same formative weight is an open question it says out loud rather than settles by assertion.
Human Workday doesn't consider an inconvenient question evidence that the architecture failed. It considers it a research obligation.
The question doesn't stop at HR. Wherever junior work has trained senior judgment — accounting, law, engineering, medicine, finance, consulting — the same problem is a candidate research programme, not a solved one.
New papers are announced on this page first. There is no mailing list and no email capture here — none exists.
The document that governs all of this work is public.