Designing AI that Enhances Human Value and Usability: A Framework for Developing AI Products and Features in SaaS Services
Burianová, Hana (2026)
2026
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-202604298395
https://urn.fi/URN:NBN:fi:amk-202604298395
Tiivistelmä
AI features are being shipped into SaaS products faster than the design field has developed tools to handle them. The result is a growing gap. Designers are making high-stakes decisions about user control, transparency, and accessibility without a shared structure for reasoning about any of it.
This thesis addresses that gap directly. It asks how AI products and features can be designed in a human-centered way in Software-as-a-Service (SaaS) services, and what kind of design framework that requires.
The research was conducted as a qualitative, literature-based study combining thematic analysis of academic research with synthesis of current practitioner knowledge. The analysis surfaced four recurring failure patterns in AI feature design: value ambiguity, loss of user control, cognitive overload, and accessibility failures. These were translated into five concrete structural design requirements.
The outcome is a five-dimension human-centered design framework for AI features in SaaS services, covering Human Value & Purpose, Interaction & Usability Structure, Human Role, Agency & Control, Transparency & Trust, and Accessibility & Inclusion. Each dimension is defined by a central design question, a set of actionable design considerations, and the key tension designers must navigate within it. The framework is designed for early-stage AI feature work and extends existing UX methods into the territory they were not built for.
The core conclusion is that human-centered AI design is a structural problem, not a checklist. Existing heuristics and accessibility standards are necessary but insufficient. Accessibility, transparency, and user control must be built into AI features as foundational obligations before the interaction design begins, not after it is complete.
This thesis addresses that gap directly. It asks how AI products and features can be designed in a human-centered way in Software-as-a-Service (SaaS) services, and what kind of design framework that requires.
The research was conducted as a qualitative, literature-based study combining thematic analysis of academic research with synthesis of current practitioner knowledge. The analysis surfaced four recurring failure patterns in AI feature design: value ambiguity, loss of user control, cognitive overload, and accessibility failures. These were translated into five concrete structural design requirements.
The outcome is a five-dimension human-centered design framework for AI features in SaaS services, covering Human Value & Purpose, Interaction & Usability Structure, Human Role, Agency & Control, Transparency & Trust, and Accessibility & Inclusion. Each dimension is defined by a central design question, a set of actionable design considerations, and the key tension designers must navigate within it. The framework is designed for early-stage AI feature work and extends existing UX methods into the territory they were not built for.
The core conclusion is that human-centered AI design is a structural problem, not a checklist. Existing heuristics and accessibility standards are necessary but insufficient. Accessibility, transparency, and user control must be built into AI features as foundational obligations before the interaction design begins, not after it is complete.