HAI across the product lifecycle
Applying human-AI interaction practices **from discovery through launch and iteration**—not as a polish pass. HAX Toolkit guidance: use tools early; Workbook at requirements time; Playbook while prototyping NLP failures. PAIR chapters fo...
What it is
Applying human-AI interaction practices from discovery through launch and iteration—not as a polish pass. HAX Toolkit guidance: use tools early; Workbook at requirements time; Playbook while prototyping NLP failures. PAIR chapters follow product development flow with worksheets. Stanford HAI voices argue multidisciplinary partners must join at the start, not the end.
<!-- IMAGE: discover → define → design → ship → monitor → redesign -->
Visual Spec & Architecture Diagram
Product lifecycle with HAI checkpoints at discovery, design, launch, iterate—questions per gate.
Why it matters
Teams told Microsoft that guidelines often require engineering/data changes—“if the spec doesn’t have that built in, it’s too rigid.” Late HAI is expensive HAI. Lifecycle literacy also covers notifying users when capabilities change and adapting cautiously over time.
How it works (plain)
| Stage | HAI moves |
|---|---|
| Discover | Is AI the right tool? (PAIR value patterns) |
| Define | Workbook: pick guidelines; size UI/AI/data/eng |
| Design | Design Library / PAIR patterns; oversight modes |
| Prototype | Playbook failure scenarios; recovery UX |
| Ship | Onboarding, uncertainty, confirms |
| Operate | Feedback, monitoring, change notifications |
| Redesign | Recalibrate automation vs control |
Everyday example
Building a staircase with a handrail in the blueprint—not bolting a rope on after someone falls.
Try it
For one AI feature, write the lifecycle stage you are in and the single HAI artifact you will produce this week (Workbook rows, pattern storyboard, or failure play).
Myths
- ⚠️ Myth: HAI starts when visual design starts.
- ✓ Reality: It starts at problem selection and data design.
- ⚠️ Myth: Launch ends HAI work.
- ✓ Reality: Over-time guidelines (feedback, updates, global controls) own the long game.
Sources
- HAX Toolkit get-started: https://www.microsoft.com/en-us/haxtoolkit/ ↗
- HAX Workbook: https://www.microsoft.com/en-us/haxtoolkit/workbook/ ↗
- HAX Guidelines: https://www.microsoft.com/en-us/haxtoolkit/ai-guidelines/ ↗
- PAIR Guidebook: https://pair.withgoogle.com/guidebook-v2/ ↗
- Stanford HAI: https://hai.stanford.edu/news/how-do-we-design-and-develop-human-centered-ai ↗