Course 1 of The AI Decision Course
Proving AI's value: use cases and ROI
For leaders who have to decide which AI projects to fund and show they paid off: naming the problem, setting the bar before a pilot, reading the numbers, and when to stop.
27 short lessons. Each one is a five-minute read with a card your team fills in and three AI prompts to test it.
Lessons in this course
- Before you fund an AI project, make someone name the problem
- Start from a problem the people it's for will admit to having
- Look past the chatbot to where AI pays
- Spot the work that can't be staffed, and size the case
- A safe first AI project or a differentiating one
- Pick the first function by what you sell
- A 95% accurate model still has to show the business number
- Set the baseline and the bar before the pilot starts
- How to read a headline that says most AI pilots fail
- Test the hours-saved claim before you sign for 200 seats
- Build the AI case on numbers the business already trusts
- Find use cases by watching the work
- Two ways to cut a long list of AI ideas
- Small bets or big ones: setting the mix of your AI portfolio
- Cut a big problem down to the size of the first project
- Choose the model by how the mistake gets caught
- Start with the data you have, and fund the work it still needs
- Budget for the system around the model
- Test it before your customers do
- Before you choose or switch a model, run it on your own work
- Who checks the output: rules, a model judge, or a person
- Before you approve the rollout, find the faulty part
- Before you fund the case, rerun it at surge prices
- Ask AI for a decision, and say what a good one is
- Four tests before an AI proposal gets money
- Stop, restart or keep funding
- Name the edge that doesn't come with an API key