Interpretive Design Research for AI Products

AI product teams often frame user research around tasks: what does the user need to do, where is the friction, and which feature solves it? That approach is useful, but incomplete. Ilpo Koskinen’s Design, Empathy, Interpretation points to a deeper layer: users do not merely perform tasks; they interpret technologies and act according to the meanings they create.

The book articulates interpretive design research through the history of an empathic design research group in Helsinki. Human beings are understood through meaning: how they interpret themselves, others, objects, technologies, and abstractions in context.

For AI productization, this matters because AI systems are meaning-heavy technologies. They do not simply add functionality. They change how users understand expertise, control, responsibility, fairness, creativity, and work. A model recommendation may be interpreted as advice, command, surveillance, evidence, or noise. A chatbot may be seen as service, deception, convenience, or cost-cutting. These interpretations shape adoption more powerfully than many feature lists.

A product team using interpretive design research would not begin only with requirements. It would begin with context. Who are the actors? What practices already exist? What meanings attach to current tools? What does “good work” mean in this domain? What anxieties surround automation? What organizational language is used to describe AI? What forms of evidence are trusted?

From usability to meaning

A useful design journey runs from usability to user experience, empathic design, codesign, radical innovation, constructive design research, and beyond. Each stage adds depth. Usability asks whether the system can be used. User experience asks how it is experienced. Empathic design asks how it fits human lives. Codesign asks how users can participate in shaping it. Constructive research asks how prototypes and artifacts can generate knowledge.

In practical consulting, this can become an AI product discovery framework:

• Interpretive fieldwork with interviews, observation, workflow shadowing, and artifact analysis.
• Meaning mapping across stakeholder groups.
• Prototype conversations that surface assumptions, not only interface preferences.
• Codesign of governance: approval, escalation, transparency, and accountability rules.
• Pilot evaluation that measures trust, perceived agency, and organizational fit alongside usage metrics.

This approach is particularly valuable for enterprise AI, where failure often comes from organizational mismatch rather than technical impossibility. A technically strong model can fail if it violates professional norms. A simple automation can succeed if it aligns with user meaning and improves practice.

The productization message is straightforward: design research should not be reduced to UX polish. It is part of AI risk management and value creation. Interpretive methods help teams build products that users understand, trust, and integrate into real workflows.

The best AI products will not only answer “Does it work?” They will answer “What does it mean to the people who must live with it?”