Business Strategy and AI Project Management
What this domain tests
Whether you can justify an AI investment in financial terms, sequence a portfolio of use cases, run delivery that survives contact with stakeholders, and connect model metrics to a business outcome.
Core rules
- Build the business case on value (revenue, cost, risk, experience) against total cost of ownership: data acquisition, labelling, compute, integration, monitoring, and change management.
- Use NPV, payback period, and sensitivity ranges rather than single-point forecasts; state assumptions explicitly.
- Prioritise on value versus feasibility, with data readiness and regulatory exposure as gating factors. A high-value use case with no usable data is not a first project.
- Define the baseline and success metric before building, tied to a specific process step and decision owner.
- Delivery: CRISP-DM or agile sprints, timeboxed proof of concept, then pilot, then production - with written kill criteria at each gate.
- Map stakeholders: executive sponsor, product owner, data owner, risk and compliance, and the end users whose workflow changes.
Worked example
Two candidate projects. Project A saves 2% of a $50M cost base, so $1M, with a feasible 90% accurate model. Project B is 99% accurate on a $2M process, so about $20k at 1% savings. Reasoning: rank on value, not accuracy - A is worth roughly fifty times B. Then confirm A's data readiness and integration cost before committing, because accuracy never appears on a P&L.
Traps that cost marks
Reporting model accuracy to executives instead of financial impact. Omitting running and change-management costs. Launching a pilot with no funded path to production. Ignoring adoption, so a technically working model sits unused. Hiding uncertainty instead of presenting a range.