Home / Seasons / AI for Business
Praxis Learn™
AI for Business
The numbers that tell you whether an AI system works, what it costs and what it returns. A season is five episodes, one metric each, with its practice pack and one certificate at the end.
Seasons
Choose a season
Each season is five episodes on metrics that belong together, with the practice pack for every metric and one certificate. Launch prices are shown.
Season 1 · 5 episodes
Answer quality
Measure whether an AI assistant's answers are right, grounded and good enough to trust.
- Hallucination rate
- Groundedness rate
- Verified answer accuracy
- Evaluation pass rate
- Weighted quality score
Season 2 · 5 episodes
AI agents
See whether an AI agent finishes the job, how many steps it takes and when people have to step in.
- Task completion rate
- Steps per task
- Tool-call success rate
- Human intervention rate
- Autonomous resolution rate
Season 3 · 5 episodes
Token costs
Work out what each task costs in tokens, and what caching and volume do to the monthly bill.
- Tokens per task
- Cost per task
- Cached-token saving
- Monthly AI usage cost
- Cost per successful task
Season 4 · 5 episodes
AI versus people
Compare what a task costs when AI does it, when people do it, and when people check the AI's work.
- Human cost per task
- Human review rate
- AI cost per task including review
- AI cost per resolution
- Saving per resolution
Season 5 · 5 episodes
AI customer service
Know how much of customer service the AI resolves on its own, and how customers feel about it.
- Automated resolution rate
- Escalation rate
- Repeat contact rate
- AI customer satisfaction (CSAT)
- True resolution rate
Season 6 · 5 episodes
Safety
Count guardrail triggers, policy breaches, data leaks and injection attempts per 10,000 conversations.
- Guardrail trigger rate
- Policy violation rate
- Personal data leakage per 10,000 conversations
- Prompt-injection success rate
- Safety incidents per 10,000 conversations
Season 7 · 5 episodes
Time saved
Turn minutes saved per task into hours a month, net of rework, and into capacity freed.
- Time saved per task
- Hours saved per month
- Rework rate
- Net hours saved
- Capacity freed (FTE)
Season 8 · 5 episodes
Adoption
See who uses the AI, how often, and what each active user costs.
- Adoption rate
- Weekly active usage rate
- Uses per active user
- Licence utilisation
- Cost per active user
Season 9 · 5 episodes
Business impact
Measure the lift in conversion, revenue, cost and errors, and the profit the AI adds.
- Conversion uplift
- Revenue uplift
- Cost reduction %
- Error reduction %
- Incremental profit from AI
Season 10 · 5 episodes
Return
Put the AI project's total cost against its benefits: net benefit, ROI, payback and NPV.
- Total cost of ownership
- Annual net benefit
- AI ROI
- Payback period
- NPV of the AI project
Season 11 · 5 episodes
Predictive AI accuracy
Judge a predictive model by accuracy, precision, recall, F1 and the cost of its mistakes.
- Accuracy
- Precision
- Recall
- F1 score
- Expected error cost per 1,000 decisions
Season 12 · 5 episodes
Fairness
Check whether an AI decides differently for different groups, and by how much.
- Selection rate by group
- Approval gap
- Error rate gap
- False positive rate gap
- Disparate impact ratio
Season 13 · 5 episodes
Speed and reliability
Measure how fast and how reliably the AI answers: response time, uptime, failures and service level.
- Time to first answer
- Average response time
- Uptime %
- Failure rate
- Response service level
Practice packs
Just practice
A practice pack is the self-marking practice tools of one topic: new numbers every attempt, marked as you type, with the working when you want it. No course and no certificate.
Practice pack
Answer quality
Measure whether an AI assistant's answers are right, grounded and good enough to trust.
- Hallucination rate
- Groundedness rate
- Verified answer accuracy
- Evaluation pass rate
- Weighted quality score
Practice pack
AI agents
See whether an AI agent finishes the job, how many steps it takes and when people have to step in.
- Task completion rate
- Steps per task
- Tool-call success rate
- Human intervention rate
- Autonomous resolution rate
Practice pack
Token costs
Work out what each task costs in tokens, and what caching and volume do to the monthly bill.
- Tokens per task
- Cost per task
- Cached-token saving
- Monthly AI usage cost
- Cost per successful task
Practice pack
AI versus people
Compare what a task costs when AI does it, when people do it, and when people check the AI's work.
- Human cost per task
- Human review rate
- AI cost per task including review
- AI cost per resolution
- Saving per resolution
Practice pack
AI customer service
Know how much of customer service the AI resolves on its own, and how customers feel about it.
- Automated resolution rate
- Escalation rate
- Repeat contact rate
- AI customer satisfaction (CSAT)
- True resolution rate
Practice pack
Safety
Count guardrail triggers, policy breaches, data leaks and injection attempts per 10,000 conversations.
- Guardrail trigger rate
- Policy violation rate
- Personal data leakage per 10,000 conversations
- Prompt-injection success rate
- Safety incidents per 10,000 conversations
Practice pack
Time saved
Turn minutes saved per task into hours a month, net of rework, and into capacity freed.
- Time saved per task
- Hours saved per month
- Rework rate
- Net hours saved
- Capacity freed (FTE)
Practice pack
Adoption
See who uses the AI, how often, and what each active user costs.
- Adoption rate
- Weekly active usage rate
- Uses per active user
- Licence utilisation
- Cost per active user
Practice pack
Business impact
Measure the lift in conversion, revenue, cost and errors, and the profit the AI adds.
- Conversion uplift
- Revenue uplift
- Cost reduction %
- Error reduction %
- Incremental profit from AI
Practice pack
Return
Put the AI project's total cost against its benefits: net benefit, ROI, payback and NPV.
- Total cost of ownership
- Annual net benefit
- AI ROI
- Payback period
- NPV of the AI project
Practice pack
Predictive AI accuracy
Judge a predictive model by accuracy, precision, recall, F1 and the cost of its mistakes.
- Accuracy
- Precision
- Recall
- F1 score
- Expected error cost per 1,000 decisions
Practice pack
Fairness
Check whether an AI decides differently for different groups, and by how much.
- Selection rate by group
- Approval gap
- Error rate gap
- False positive rate gap
- Disparate impact ratio
Practice pack
Speed and reliability
Measure how fast and how reliably the AI answers: response time, uptime, failures and service level.
- Time to first answer
- Average response time
- Uptime %
- Failure rate
- Response service level
65 AI for Business micro courses
Every metric is also a micro course on its own: five scenes, a practice tool and a certificate, R295 · US$19, lifetime access.
How a season works
Learn it, practice it, prove it
- 01
Learn each metric
Five episodes, one metric each: what it measures, how to work it out and how to read the result.
- 02
Practice until it clicks
Cases from different businesses, with new numbers every time. Your answer is marked as you type.
- 03
Get help when stuck
Hints point to the step that went wrong, never just the answer. Want a person to talk it through? Book a tutor session as an optional add-on.
- 04
Earn the certificate
Finish every episode and the practice pack in the season and you get one certificate for the lot.
Just one metric
Only need one metric?
Every micro course is also available on its own, with its certificate, at a launch-offer price of R295 or US$19.
One whole series
Want more than one season?
Full series access gives you every season and micro course in one series of your choice for 12 months, with the practice tools built into every season. Tutoring is an optional add-on.