How the toolkit works
The three layers
The toolkit has three layers. Together they produce traceable, auditable workforce planning outputs.
Calculation engine
The Workforce Model (Excel workbook) implements the seven-layer model — baseline, demand, supply, AI scenarios, gap, interventions, cost roll-up, outputs — as formulas that flow from inputs through assumptions to outputs.
This is the system of record. Every number traces back through formulas to source data and stated assumptions.
Methodology & templates
The Playbook (this site) describes the methodology: phases, steps, governance, decision gates, common pitfalls, glossary. The Templates give ready-to-fill structures for the major outputs.
This layer answers 'what do we do, in what order, and what should the outputs look like?'
Prompt library
Twelve structured prompts for AI tools. Each is matched to a specific step in the methodology. They accelerate work that is otherwise time-intensive.
The prompts work in any capable AI tool — Microsoft Copilot, Cowork, Claude, ChatGPT.
What 'agent' means in this toolkit
What the prompts are: Carefully designed instructions for an AI tool to produce a specific kind of output (a draft, a synthesis, an analysis). They include context, constraints, output format, and placeholder slots for inputs.
What the prompts are not: Autonomous agents. They have no memory between uses, no orchestration, no live data integration. They are templates the user runs in a separate AI tool, on demand.
This distinction matters because regulatory submissions require auditable calculation. An AI tool can draft narrative and analyse text; it cannot be the source of numerical truth in a defensible submission. The Workforce Model is.
How the layers connect
- Methodology → Model: The Playbook directs the working group to populate the model layer by layer.
- Prompts → Methodology: Prompts produce drafts that fill the templates the methodology calls for.
- Prompts → Model: Some prompts produce structured assumption packs that populate model inputs directly.
- Model → Templates: Model outputs are narrated in the templates (options paper, regulatory submission narrative).
Why this architecture
- Auditability. Regulators and auditors can examine the model and trace every output back to source.
- Durability. The methodology, model, and prompts persist. AI tools change rapidly; the toolkit does not depend on any single tool.
- Pragmatism. Building production AI agents that read from internal systems takes 6-12+ months. Prompts in user-driven tools work today.