Briefings
Independent analyst briefings on enterprise AI deployment by Jason A. Milne, organized on the project manager’s iron triangle and its shadow: cost (The Money Ledger), schedule (The Time Ledger), quality (The Trust Ledger), and risk (The Safety Ledger, home of AI Fault Lines). Same discipline in every book — every claim sourced, every figure dated, verdicts marked as the briefing’s own.
The Money Ledger · Cost
General ledger: deployment cost · Subledgers: physical · token · model · human · equity · Edition One live, roadmap through Jan 2027
Why enterprise AI pilots die broke — and the financial discipline behind the ones that don’t. The thesis edition.
What AI actually costs from the substation up: power, cooling, silicon depreciation, and the facility economics under every token.
Worked unit economics: cost per resolved task, caching and routing math, and the self-host-versus-API breakeven.
Deprecation and depreciation: model half-lives, the migration tax of retired versions, fine-tune write-offs, and amortizing an asset that ages in months.
Named, independently verified playbooks from deployments that reached P&L — and the habits they share.
Forward deployment and change management as line items: what the engineers cost, what adoption costs, and contracting for graduation metrics.
Payback curves, capex versus opex, and the investment case for enterprise AI — including its failure modes.
The Safety Ledger · Risk
AI Fault Lines · Standing briefing, updated by edition · Began as an audit of Bain’s 2024 sovereign-AI call
The sovereign-AI ledger two years after Bain’s call, the open-weights escalation, two safety precedents that cannot be un-set, and the awareness problem.
Earlier editions (Jul 2 · Jul 3, 2026) in archive; available as PDF on request.
The Time Ledger · Schedule
Why AI deployments slip: pilot purgatory, integration timelines, and the schedule economics of shipping.
Opens after the Money Ledger’s early subledgers ship.
The Trust Ledger · Quality
Evals as the measure of done: accuracy, reliability, and the cost of quality — what it takes for an AI system to earn trust.
Opens after the Money Ledger’s early subledgers ship.
Theses
The ledgers audit the present — dated, verified, marked to market. Theses argue where it points. Occasional long-form.
The firm as a feedback system — and what happens when the feedback runs itself.