Last CallLast Call Pool
Future of design

How AI-Powered UX Research Streamlines the Design Process

September 4, 2023 2:45 PM
Power Talk
🇬🇧 English
Bratysława 1

About

I‘m currently working on a tool that aims to automate the repetitive tasks of user research. I would like to talk about the process that I used to build such a tool, what it’s capable of, why designers should embrace AI, and my prediction for the impact of AI on UX in general.

I‘m working as a Product Designer for AI tools since 2019 and have a deep understanding of what is currently possible and where the limitations are. I would like to share my knowledge about the intersection of AI, design and business and how we as designers should handle the big shifts that are now reaching our industry.

Talk by

Photo of Bruno Recht, Product Designer at Porsche Digital
Bruno Recht
Product Designer at Porsche Digital

Watch the full talk

Watch this WaysConf session, then continue with related talks or explore the current programme.

Explore the 2026 agendaMore WaysConf recordings
From the recording

Talk in brief

At WaysConf 2023, Bruno Recht presents how generative AI and large language models (LLMs) are redefining product design and user experience research. While current LLMs lack seamless workflow integration, AI tools are quickly shifting from basic content generators to specialized assistants that automate tedious design tasks. Recht outlines three key predictions for the field: text-to-interface generation from design systems, natural language to functional code, and autonomous user research. To navigate this transition, product teams must adopt cross-functional collaboration and systematic frameworks. Recht introduces the AI Incubation Canvas to help teams validate AI-specific problems, manage dual build-measure-learn loops, and structure business, design, and technical requirements before introducing his concept for UserFlix, an AI research co-pilot.

Key takeaways

  1. 01

    Solve Problems Unique to AI

    Teams should only integrate AI when a problem cannot be efficiently solved through traditional software methods. Introducing machine learning adds technical complexity and maintenance overhead that is only justified by unique capability gains.

    Watch from 12:33
  2. 02

    Integrate AI into Existing Workflows

    Current large language models often fail to maximize productivity because users must constantly switch context between standalone chat interfaces and core design software. True efficiency relies on embedding AI directly into native design tools and workflows.

    Watch from 3:08
  3. 03

    Run Dual Iteration Loops

    AI product development requires running two simultaneous build-measure-learn cycles for user experience and machine learning models. Product teams must continually evaluate model outputs to eliminate hallucinations and algorithmic bias alongside traditional user testing.

    Watch from 13:02
  4. 04

    Form Cross-Functional AI Teams

    Effective AI design projects require a balanced mix of expertise across design, business, engineering, and data science. Uniting these four perspectives prevents single-discipline blind spots during product development.

    Watch from 13:32
  5. 05

    Automate Monotonous Design Tasks

    Machine learning should handle repetitive operational tasks like layer naming, component styling, and data clustering. Delegating administrative overhead to AI assistants liberates human designers to focus on strategic thinking and creative execution.

    Watch from 9:59
Read edited transcript highlights

These concise notes were edited from automatic captions and checked against the talk structure. They are not a verbatim transcript.

Contextualizing Generative AI and Interface Workflows

Large language models have established a strong baseline for text and content generation, but current implementations still suffer from severe interaction friction. Most popular generative tools operate as standalone web interfaces, forcing designers to constantly copy and paste outputs between disconnected software applications. This context switching neutralizes much of the speed gained through automated content creation. For artificial intelligence to deliver true operational value across product design teams, machine learning capabilities must be embedded directly into primary workspace environments rather than isolated inside generic chat windows.

Watch from 3:01

Surveying Machine Learning Plugins for Designers

A new wave of specialized design software is demonstrating how machine learning can streamline daily project work. Canvas companions like Gem offer automated sticky-note clustering and synthesis during whiteboarding workshops. Infrastructure utilities like Diagram focus on eliminating tedious administrative maintenance in design files, automatically standardizing layer naming and applying component styles. Meanwhile, speculative platforms like Galileo AI aim to convert text descriptions straight into complete interface screens. These early experiments highlight a clear shift away from generic chat prompts toward embedded utilities tailored to specific design actions.

Watch from 5:28

Shifting Human Roles from Execution to Curation

Emerging artificial intelligence tools will not replace human designers, but they will fundamentally restructure daily job responsibilities. Human professionals excel at higher-level conceptual reasoning, strategic vision, and critical evaluation, whereas algorithmic models excel at rapid execution and pattern processing. In future workflows, designers will act primarily as directors and curators who guide machine outputs, refine generated layouts, and enforce quality standards. Delegating repetitive production tasks to automated assistants frees design professionals to spend more time addressing complex user problems.

Watch from 8:32

Strategic Foundations for AI Product Development

Developing successful artificial intelligence features requires following three foundational rules. First, teams should only deploy machine learning models when solving a problem that traditional programmatic logic cannot address, avoiding unnecessary system complexity. Second, product creators must manage two parallel build-measure-learn loops: one focused on overall user experience and another dedicated to model performance, data training, and hallucination reduction. Third, AI projects demand balanced cross-functional collaboration across business strategy, experience design, software engineering, and data science.

Watch from 12:16

Structuring Projects with the AI Incubation Canvas

To systematically plan machine learning ventures, the AI Incubation Canvas divides project requirements into distinct business, user experience, and technical dimensions. Design teams manage user problem definitions and interface touchpoints, business leads map monetization models and market viability, and technical specialists handle data logistics, model selection, and fine-tuning. Clearly categorizing whether a project should fine-tune open-source models or build proprietary architecture ensures that technical complexity directly aligns with business goals before engineering begins.

Watch from 14:07
Explore WaysConf 2026

View the agenda

Get tickets