Notes on the Capability Economy.
We don't write to a content calendar. We write when we have something specific to say β a customer's question that turned out to be everyone's question, an architectural decision worth showing our work on, a pattern from the design-partner motion that's compounding.
Find what to read by audience and stage β three audiences, four stages, twelve questions.
The grid is the navigational primitive of the resource surface. Columns are audiences; rows are buying stages. Each cell is a question-shaped headline that points at one or more blog posts that carry the answer.
What is AI coaching? Why do most AI rollouts stall?
The failure-mode taxonomy primer for buyers β why 95% of generative-AI pilots show no measurable P&L impact, and what an operating layer fixes that distribution alone doesn't.
What does an AI coach actually do for me at work?
A walkthrough of a day with Lucy β what she sees, how she nudges, what you build, and why the capability stays with you instead of with the tool.
How does AI coaching fit a public-sector workforce literacy mandate?
An EU AI Act Article 4 framing of workforce AI literacy β what the mandate actually asks for, and how a coaching surface satisfies it without surveillance.
Is Lucy different from Copilot or training programs? What does the trust architecture look like?
The routing-judgment thesis plus the VΒ·CΒ·C primer β why general AI tools answer prompts, training teaches out of context, and Lucy's trust architecture is structurally different.
What does Lucy see at my desk? What can my manager see?
The Signal / Context / Content boundary explained for the person doing the work β what Lucy observes, what she keeps, and the aggregate-only line manager-side data never crosses.
Does Lucy align with EU AI Act Articles 4 + 14? With NIST AI RMF 1.0?
The trust-architecture-as-compliance essay β how aggregate-only manager data, refused capabilities, and human-oversight defaults map to Articles 4, 14, and the NIST AI RMF 1.0 functions.
What does a pilot look like? How do you measure outcomes?
The pilot deal-shape walkthrough β how design-partner engagements scope, what the measurement plane looks like, and which signals tell you the operating model is actually compounding.
What's a Personal Coaching Plan? What does the first useful win look like?
A walkthrough of the Personal Coaching Plan β how Lucy sequences your first agent build, what counts as the first useful win, and how the pattern compounds across your week.
What does the Paid Pilot variant look like for procurement?
The Paid Pilot variant explainer β the deal shape designed for public-sector procurement, the deliverable structure, and how the engagement closes into an at-scale rollout.
How does Lucy scale beyond the pilot team? What's the bottom-up moat?
The bottom-up scaling moat essay β why capability-formation data is structurally different from task-completion data, and why the moat compounds the further the rollout goes.
How does Lucy scale across teams without losing the trust posture?
The capability-stays-with-you essay β how cohort gating, aggregate-only manager surfaces, and refused capabilities keep the trust posture intact as adoption widens.
How does Lucy roll out across a public-sector workforce at scale?
The public-sector workforce-literacy-at-scale essay β what a multi-agency rollout looks like, how the trust architecture lands with works councils and regulators, and how the cadence holds.
Latest from the Lucy Labs blog.
Reverse-chronological. Audience and stage tags reference the grid so you can read across or read down.
Trust β Personal value β Adoption β Capability β Scale
The five-stage framework for actually changing behavior with AI. Why every other order fails β and why "rollouts" that skip Trust never make it past Adoption.
Business buyer Β· DiscoveryTrust architecture: building AI people will actually use
Why "privacy by design" can't be a marketing line. The eight subsystems of trust, how they stack, and what breaks when you miss one.
Business buyer Β· EvaluationUniversal AI skills are the leverage point most rollouts skip
The 70% of AI value lives in the muscle, not the model β the universal skills that turn AI from a curiosity into a working partner. And almost every rollout skips past it.
Individual / employee Β· DiscoveryWhy predicting the future is harder than you think
People are bad at exponentials. Not a little bad. Really bad. The paper-folding problem, applied to your AI roadmap.
Business buyer Β· ScaleThe alignment problem nobody is solving
AI quietly renamed our core framework. Here's the document alignment system we built to catch drift before it compounds.
Business buyer Β· Implementation