A subject-matter expert brings substantial knowledge of a problem area, industry, workflow, or user population. That knowledge helps an AI team find worthwhile opportunities, understand users, define useful features, and avoid costly misjudgments. Experts may know little about technical feasibility, so their strongest contribution is usually domain judgment rather than implementation guidance. section 2.1.2 section 12.1.5
Use subject-matter expertise across the product lifecycle:
- Identify promising AI applications by exposing real workflow needs, user problems, and industry-specific opportunities. section 2.1.2 section 12.1.5
- Define features and success criteria that fit how people actually work, rather than optimizing technical performance in isolation. section 12.1.5
- Validate outputs and provide targeted feedback, while recognizing that limited availability may require data annotators, knowledge engineers, or active users to extend the work. section 12.1.5
Collaboration works best when technical and domain perspectives are deliberately connected. Experts can prevent dead ends, but they may focus on surface-level issues, misunderstand AI as a rules-based system, or prioritize workflow integration differently from the AI team. Shared metrics, simple AI education, structured feedback, think-aloud sessions, regular alignment, and an AI-domain liaison make their knowledge easier to use. section 12.1.5
Logistics product
A logistics company can involve a dedicated expert in product questions, consult additional logistics specialists, and later recruit active power users as co-creators for detailed feedback. This blends domain knowledge with AI capability as the product matures. section 12.1.5
Table 12.2 Common challenges when collaborating with domain experts
| Challenge |
Problem |
Solution |
| Shallow feedback |
Experts focus on surface-level problems (typos, minor errors, layout problems) instead of systemic improvements. |
Use structured feedback frameworks, group problems into layers, and conduct think-aloud sessions to uncover deeper insights. |
| Misguided assumptions about AI |
Experts misunderstand AI logic, treating it like a rules-based system or expecting direct, manual control over outcomes. |
Provide simple AI education, encourage feedback on outcomes rather than mechanisms, and show how AI improves over time. |
| Divergent priorities and communication styles |
AI teams focus on technical performance, while domain experts prioritize usability and workflow integration. |
Define shared success metrics, appoint AI-domain liaisons, and hold regular alignment meetings. |
| Resistance to AI and change aversion |
Experts fear AI will replace their judgment, don’t trust AI’s logic, or resist workflow disruptions. |
Position AI as a copilot, enhance transparency with explainability features, and introduce AI in low-risk, high-value use cases first. |
Use the challenge-and-solution table to diagnose shallow feedback, mistaken assumptions about AI, and mismatched priorities or communication styles. section 12.1.5
