Practice set: families, use cases, value case and questions

P1. A system whose behaviour comes from rules people wrote is:

• Rules-based automation
• A predictive model
• Generative AI
• A search engine

P2. A good use case is one that is:

• Frequent, data-rich, error-tolerant and measurable
• Fashionable
• Rare and secret
• Expensive

P3. 1,000 documents at 4 minutes and £0.50 per minute cost:

• £2,000
• £200
• £4,000
• £500

P4. New cost £1,800 against old cost £2,000. The saving is:

• £200
• £1,800
• £3,800
• £2,000

P5. 5% of 1,000 parts are faulty. Calling every part good gives accuracy:

• 95%
• 5%
• 50%
• 100%

P6. Which is NOT one of the five questions?

• Which logo should it use?
• Who owns the result?
• What is the evidence?
• What data does it need?

P7. The cost of correcting AI errors should be:

• Included in the value case
• Left out
• Added to the benefit
• Ignored if small

P8. A claim should be tested on:

• Your own data against a baseline
• The vendor’s demo data only
• Nothing
• Last year’s brochure