skills/financial-modeling/references/technology-and-tco.md

# Technology, automation, and total cost of ownership

## Cost the whole operating model

Compare solutions at the same service level, volume, geography, horizon, and demand assumptions. Include implementation, integration, migration, testing, dual-running, training, ongoing licenses, usage charges, support, maintenance, data work, and exit costs when applicable. Identify what is included in a vendor quote or loaded labor rate before adding allocations.

For cloud and AI workloads, model usage units and unit prices separately. Include retries, review effort, exception handling, evaluation, monitoring, and quality remediation where material. Model variable and fixed costs appropriately; do not treat every technology investment as capex or every cloud contract as purely variable.

Distinguish cash expenditure from internal opportunity cost. Both may affect a decision, but a reallocated salaried employee does not automatically create a cash saving. Present a cash view and, when useful, an economic-cost view without mixing the two.

## Establish the benefit mechanism

Build from measured or explicitly estimated drivers: eligible work × time change × sustained adoption × usable share, adjusted for review and rework. Check whether the measured result applies to the same tasks and population. Model the ramp and quality alongside speed.

Separate these outcomes:

- **Task time released:** observed or estimated minutes saved. This is not yet a financial benefit.
- **Capacity available:** time that can be scheduled or redeployed to useful work.
- **Cash savings:** spending actually avoided or reduced through an identified staffing, outsourcing, overtime, or purchasing change. Include transition costs.
- **Incremental contribution:** additional fulfilled demand × contribution margin, constrained by demand, bottlenecks, and sales capacity. Revenue is not the same as profit or cash flow.
- **Risk reduction:** change in exposure supported by a defensible mechanism and frequency/severity assumptions. Show scenarios when probabilities are unknown.

Do not count the same hours as both a reduction in payroll and capacity that produces additional sales. Automation savings, prediction quality, and network effects each require evidence; none is proven by its category.

## Model uncertainty that matters

Test adoption, quality, review load, unit usage cost, benefit timing, and switching costs as the case requires. A pilot should define what evidence will justify rollout, what failure would stop it, and the cost of learning. Treat future use cases and options to change technology as contingent opportunities, not invented value added to rescue a negative case.

The decision should state whether benefits become cash, capacity, service improvement, or risk reduction, who can realize them, and how they will be measured after implementation.