Where (and Why) We Really Need AI in the UX Workflow
The Problem
Because of my passion for AI tools, I often find myself in conversations with designers, leaders, and ones who are just starting their journey in UX and product design. Almost everyone asks the same question: “How should we integrate AI into the design process?”
It’s a fair question, but maybe not the right first one. Instead of asking how, perhaps we should start with where and why. Where do we actually need AI? And why are we introducing it in the first place?
Why This Matters
The question of where AI can add real value isn’t one-size-fits-all: it depends on each team’s specific challenges and culture. Across the various UX product teams I’ve worked with, I’ve seen how dramatically their pain points can differ, even within the same organization.
In general, two factors determine where AI is most useful:
- Which parts of the workflow are time-consuming or ripe for automation.
- Where you can trust AI’s output, and have the expertise and time to review it when needed.
When I gathered data from different designers about how much time each design stage takes, the diversity of answers was striking:
- For some, handoff was the biggest bottleneck because of limited communication with engineers.
- For others, discovery took the most time because every step required stakeholder alignment.
- Some teams needed help with design craft due to weak or inconsistent design systems.
- Continue this list yourself…
So before jumping into AI adoption, we must clearly define what exactly we’re trying to improve … and why.
Real Examples and Scenarios
Let’s imagine a simple brainstorming exercise: exploring common pain points where AI could make a difference.
1. Requirement Changes
This is one of the most common and most frustrating problems.
Requirements change constantly, forcing designers to loop back and redo earlier work. Because these updates are often poorly documented, it can be difficult to understand what changed and why.
An AI agent could help by capturing requirement updates from meetings, Figma files, or communication channels like Slack, then automatically compiling them into a shared Confluence page. With enough historical data, it might even estimate the impact of new changes based on past discussions and external references.
But! and this part really matters! the hardest challenge isn’t building the AI agent (you could even use an existing tool like https://www.read.ai/).
The real challenge is ensuring team alignment: agreeing on a single feedback channel, trusting AI-generated output, and keeping update processes clear and consistent. AI can support the workflow, but humans still have to orchestrate the collaboration.
2. Hand-off Documentation
Another common bottleneck is the handoff stage: the moment when design meets engineering.
Depending on a team’s maturity and how well designers and developers communicate, this step can be seamless or painfully slow. Some teams require designers to spell out every interaction, behavior, and spacing detail to avoid misunderstandings; others can resolve most questions with a quick sync.
In reality, documentation often gets pushed to the margins, squeezed in at the end of already tight timelines. Designers end up copy-pasting notes, renaming frames, adding annotations, and filling out endless tables just to ensure nothing gets lost in translation. It’s important work, but it’s rarely creative or satisfying.
This is where AI could offer real value. Imagine an assistant that automatically generates handoff documentation from your design files: linking directly to the relevant Figma components, prototypes, and specifications. It could summarize user flows, spot missing components, and flag issues with naming conventions or spacing. If the team agrees on a standard structure, the process becomes semi-automated: the AI drafts, and the designer reviews and refines.
The goal isn’t to take control away from designers or egineers, it’s to offload the repetitive work so they can focus on clarity and communication rather than formatting. Teams could save hours each week while producing more consistent, reliable documentation.
3. Component Creation
Design system work: components, tokens, structures, documentation, is essential for scalability but often slow and manual, especially without a dedicated system designer. Even experienced teams struggle to keep components current when requirements change quickly.
AI can act as a design-system co-pilot: scanning Figma files, identifying patterns, suggesting new components, generating tokens, and flagging duplicates or naming inconsistencies. Designers could even describe components in natural language (“Create a modal with two button variants”), and the AI could generate the initial structure instantly.
We still need someone to organize, validate, and shape the output, but AI can function like a junior design-system designer, supporting the Design System Lead and accelerating the work.
This shifts time away from mechanical maintenance and toward improving usability and accessibility. Early non AI plugins like Instancer and Bloq already explore this space, but it could not be used as one-for-all solution.
If you know other solutions, I’d love to hear them.
How I Use AI With My Team
As a curious UX Manager, I’m not only exploring AI myself , I’m also making sure the team enjoys using these tools and gets real value from them.
Figma Make
This has become essential for speeding up discovery. Instead of gathering the whole team for every brainstorming session, designers can co-create with AI to generate ideas quickly.
The team also uses it to turn mid-fidelity or scattered high-fidelity screens into clickable prototypes and collect early feedback.
We haven’t formalized this as a process, but the team picked it up from each other, and now it’s a natural part of our ideation flow.
Copilot
A very intuitive way to summarize meeting notes and convert them into clear, structured actions. We extract notes constantly and even replaced the “note-taker” role during design reviews.
This came from our wider org : Copilot is used beyond design, so it integrated naturally into our workflow.
ChatGPT
Team members use it to summarize long email or Slack threads and turn them into actionable steps. This is especially helpful for our international team: we can prompt it to rewrite instructions in simple, clear language.
I introduced this when I noticed that some feedback and specifications were being misunderstood. As a non-English speaker myself, I often rely on it to clarify new terms or idioms.
Of course, we still validate all AI output. These tools can hallucinate, and reviewing the results still requires context and expertise: but it’s much faster overall.
What’s Next
Before adopting any AI tool, we should always ask:
- What problem are we actually solving?
- Does that problem truly exist, and could it be solved more simply?
- Do we have enough expertise to validate and refine the AI’s output?
AI won’t magically fix your design process. But used thoughtfully, it can remove friction, save time, and help designers focus on what humans do best: creativity, empathy, and problem-solving.
Let me know if you have any other cases! Let’s discuss.
Where (and Why) We Really Need AI in the UX Workflow was originally published in UX Planet on Medium, where people are continuing the conversation by highlighting and responding to this story.
المصدر: المصدر الأصلي
{“@context”:”https://schema.org”,”@type”:”NewsArticle”,”headline”:”Where (and Why) We Really Need AI in the UX Workflow”,”description”:”
The Problem
Because of my passion for AI tools, I often find myself in conversations with designers, leaders, and ones who are just starting their journey in UX and product design. Almost everyone asks the same question: “How should we integrate AI into the design process?”
It’s a fair question, but maybe not the right first one. Instead of asking how, perhaps we should start with where and why. Where do we actually need AI? And why are we introducing it in the first place?
Why This Matters
The question of where AI can add real value isn’t one-size-fits-all: it depends on each team’s specific challenges and culture. Across the various UX product teams I’ve worked with, I’ve seen how dramatically their pain points can differ, even within the same organization.
In general, two factors determine where AI is most useful:
- Which parts of the workflow are time-consuming or ripe for automation.
- Where you can trust AI’s output, and have the expertise and time to review it when needed.
When I gathered data from different designers about how much time each design stage takes, the diversity of answers was striking:
- For some, handoff was the biggest bottleneck because of limited communication with engineers.
- For others, discovery took the most time because every step required stakeholder alignment.
- Some teams needed help with design craft due to weak or inconsistent design systems.
- Continue this list yourself…
So before jumping into AI adoption, we must clearly define what exactly we’re trying to improve … and why.
Real Examples and Scenarios
Let’s imagine a simple brainstorming exercise: exploring common pain points where AI could make a difference.
1. Requirement Changes
This is one of the most common and most frustrating problems.
Requirements change constantly, forcing designers to loop back and redo earlier work. Because these updates are often poorly documented, it can be difficult to understand what changed and why.
An AI agent could help by capturing requirement updates from meetings, Figma files, or communication channels like Slack, then automatically compiling them into a shared Confluence page. With enough historical data, it might even estimate the impact of new changes based on past discussions and external references.
But! and this part really matters! the hardest challenge isn’t building the AI agent (you could even use an existing tool like https://www.read.ai/).
The real challenge is ensuring team alignment: agreeing on a single feedback channel, trusting AI-generated output, and keeping update processes clear and consistent. AI can support the workflow, but humans still have to orchestrate the collaboration.
2. Hand-off Documentation
Another common bottleneck is the handoff stage: the moment when design meets engineering.
Depending on a team’s maturity and how well designers and developers communicate, this step can be seamless or painfully slow. Some teams require designers to spell out every interaction, behavior, and spacing detail to avoid misunderstandings; others can resolve most questions with a quick sync.
In reality, documentation often gets pushed to the margins, squeezed in at the end of already tight timelines. Designers end up copy-pasting notes, renaming frames, adding annotations, and filling out endless tables just to ensure nothing gets lost in translation. It’s important work, but it’s rarely creative or satisfying.
This is where AI could offer real value. Imagine an assistant that automatically generates handoff documentation from your design files: linking directly to the relevant Figma components, prototypes, and specifications. It could summarize user flows, spot missing components, and flag issues with naming conventions or spacing. If the team agrees on a standard structure, the process becomes semi-automated: the AI drafts, and the designer reviews and refines.
The goal isn’t to take control away from designers or egineers, it’s to offload the repetitive work so they can focus on clarity and communication rather than formatting. Teams could save hours each week while producing more consistent, reliable documentation.
3. Component Creation
Design system work: components, tokens, structures, documentation, is essential for scalability but often slow and manual, especially without a dedicated system designer. Even experienced teams struggle to keep components current when requirements change quickly.
AI can act as a design-system co-pilot: scanning Figma files, identifying patterns, suggesting new components, generating tokens, and flagging duplicates or naming inconsistencies. Designers could even describe components in natural language (“Create a modal with two button variants”), and the AI could generate the initial structure instantly.
We still need someone to organize, validate, and shape the output, but AI can function like a junior design-system designer, supporting the Design System Lead and accelerating the work.
This shifts time away from mechanical maintenance and toward improving usability and accessibility. Early non AI plugins like Instancer and Bloq already explore this space, but it could not be used as one-for-all solution.
If you know other solutions, I’d love to hear them.
How I Use AI With My Team
As a curious UX Manager, I’m not only exploring AI myself , I’m also making sure the team enjoys using these tools and gets real value from them.
Figma Make
This has become essential for speeding up discovery. Instead of gathering the whole team for every brainstorming session, designers can co-create with AI to generate ideas quickly.
The team also uses it to turn mid-fidelity or scattered high-fidelity screens into clickable prototypes and collect early feedback.
We haven’t formalized this as a process, but the team picked it up from each other, and now it’s a natural part of our ideation flow.
Copilot
A very intuitive way to summarize meeting notes and convert them into clear, structured actions. We extract notes constantly and even replaced the “note-taker” role during design reviews.
This came from our wider org : Copilot is used beyond design, so it integrated naturally into our workflow.
ChatGPT
Team members use it to summarize long email or Slack threads and turn them into actionable steps. This is especially helpful for our international team: we can prompt it to rewrite instructions in simple, clear language.
I introduced this when I noticed that some feedback and specifications were being misunderstood. As a non-English speaker myself, I often rely on it to clarify new terms or idioms.
Of course, we still validate all AI output. These tools can hallucinate, and reviewing the results still requires context and expertise: but it’s much faster overall.
What’s Next
Before adopting any AI tool, we should always ask:
- What problem are we actually solving?
- Does that problem truly exist, and could it be solved more simply?
- Do we have enough expertise to validate and refine the AI’s output?
AI won’t magically fix your design process. But used thoughtfully, it can remove friction, save time, and help designers focus on what humans do best: creativity, empathy, and problem-solving.
Let me know if you have any other cases! Let’s discuss.
Where (and Why) We Really Need AI in the UX Workflow was originally published in UX Planet on Medium, where people are continuing the conversation by highlighting and responding to this story.
“,”datePublished”:”2025-12-05 11:54:50″,”author”:{“@type”:”Organization”,”name”:”العربية”},”publisher”:{“@type”:”Organization”,”name”:”Newsly”,”url”:”http://wordpress-hr2d6.wasmer.app”},”image”:”https://cdn-images-1.medium.com/max/1024/1*WKsonpXEIyJ2sKWJHZs-KA.png”}
Post Comment