Designing AI Decision Support for Tea Grading
Research into how AI could support consistency in tea tasting and grading — without replacing the person who actually understands the tea.
Tea grading is a judgement-heavy business process. At a small tea brokerage in Siliguri, much of that judgement lived in experience, memory and informal records. This research asked what an AI-enabled platform would actually need to do to make tea tasting and grading more consistent, without taking the decision away from the people who understand the tea.
A grounded account of what one real tea-trading business would need from an AI-enabled decision-support system. Rather than asking whether AI can grade tea, the research asks a more useful design question: what should AI do when the person making the decision has expertise the system does not?
- Tea grading sits at the centre of buying, pricing and buyer acceptance, yet the reasoning behind those decisions was never held in one structured system.
- The real question wasn't whether human judgement is subjective — it was what conditions (trust, accountability, data control, cost) would make tea professionals actually accept AI support.
- AI is framed as decision support, not automated judgement — structuring information, surfacing comparisons, preserving expert knowledge.
- The sensory judgement and accountability stay with the human expert; AI's job is making that expertise easier to structure, compare and reuse, not replacing it.
- Recommended structured grading support, historical comparison, and buyer-requirement matching as the platform's core jobs — not sensory automation.
- Recommended environmental and image standardisation to remove a major source of inconsistency before any AI layer is added.
- Deliberately scoped where not to automate — treated as the main design decision, not an afterthought.
- A grounded account of what one real tea-trading business would need from an AI-enabled decision-support system — research-grounded recommendations, not a built system.
- Validated directly with the three focal-company participants in a Stage-2 focus group, rather than finalised in isolation.
"The most important design decision was deciding where not to automate. In judgement-intensive work, the value of AI may come less from replacing expertise and more from making expertise easier to structure, compare, preserve and use."
Read the complete report submitted for this project, including the research, analysis, strategy, and final recommendations.