Episode 051 Raw Data

Thanks for your interest in The Question!

Sign up to participate in future questions here.

This file is the raw data from Episode 51, deep dive hosted on April 17, 2025 with Ben Callahan and Vitaly Friedman.

There are 86 answers. The question this week was:

-----

Welcome, System Thinkers!

In a recent conversation with Vitaly, he shared an observation he’s had about design systems—most product teams are trying to figure out how to integrate AI into their products in a meaningful way, but most design systems aren’t giving any kind of AI guidance to their consumers. And if they do, the guidance typically refers to the appearance of AI features, rather than interaction patterns for designing better AI experiences.

As design system practitioners, it’s often our policy to “wait and see” what sticks out in the products before we bring those ideas into the design system. Vitaly’s observation made me wonder if that’s the right call in this case.

For episode 051, let’s dive into design system AI guidelines!

Have your consuming products shipped meaningful AI integrations?

Does your design system offer guidelines for how product teams should go about designing and embedding AI features into their work?

Share an example or a story about how your design system has or has not supported consuming team AI integrations.

Question 1: Have your consuming products shipped meaningful AI integrations?

Question 2: Does your design system offer guidelines for how product teams should go about designing and embedding AI features into their work?

Question 3: Share an example or a story about how your design system has or has not supported consuming team AI integrations.

We don't have any, but there have been growing conversations about how we might use AI in our system to generate visual assets like illustrations or how we might use it within plugins to generate designs for teams without designers.

I would love to hear how other teams use AI in their systems in a meaningful, scalable way.

They have: a conversational search, and a chatbot. The jury's out on whether they are meaningful yet as they're still new and results aren't in on whether they are providing users with a real solution.

The design system has supported in the build of a conversational AI search feature and a chatbot, but this was simply in a component capacity. We have yet to make these into sharable patterns, write any documentation on them, or write broader guidelines on AI. There are other Ai projects in the works at our company that could benefit from all of these things, but as usual - roadmaps/priorities get in the way.

In my opinion AI projects can fall victim of being very 'solution first' because its such a shiny new toy. I am interested to see in the next 6 months how much true value an AI search can provide users over a traditional search, for example.

n/a

Unfortunately I can’t share any example since at the time we had not taken any steps in that direction. But I’m very interested..!

We only use AI for content generation currently - no stronger integrations, yet, however more are on the roadmap for the upcoming quarters.

Yes

Lack of resources has deprioritized a lot of improvements to the design system which has actually hindered AI support with some legacy rigidity built in to the components

Yes

In many cases the corporate IT has set rules of which AIs are allowed and which not, based on their server premises (GDPR) or matter how they promise to use customer material on their own teaching. Also the question about IPRs is an issue.

This hasn’t been a thing for us yet, but it’s on the horizon for a couple clients/projects so I’m curious to learn from any others who have done this.

No example! But im curious to hear others

Yes

One area is an embedded experience in one of our more complex components that would require custom hooks to support the AI integration. In that case, we streamlined our normal process to get them something quickly while still keeping the integrity of the component.

We tried to support our primary AI experience at our company by documenting the chat experience and components, but in the end, it turned out due to how it was maintained and shipped it didn't meet our criteria of "Used in 3 places" that we generally require. Additionally, it was evolving too rapidly for inclusion in the design system just yet – before we were even done with the first round of docs they were fundamentally changing some components. So in the end we hid the docs, and left control of that piece of the UI with the team building the feature.

Yes

n/a

Yes

We have some guidelines on how to use related componentry like chat and we also have UX content guidelines. We have a team that manages AI use at Target so they are the gatekeepers of most of that.

As teams are experimenting with building new applications with GenAI, they will often leverage open source tooling. This makes it difficult to get them to use our DS as our DS is not open source so teams will often use Material UI in their experimentations. If those POCs/experiments accelerate into a production offering, we have to work with them to migrate to our Target DS.

N/A

none

So far, the only AI integration we have is help with documentation. I find this possibility very interesting, I wonder if we can use AI integration for onboarding and socialisation.

Yes

We're collaborating with the teams responsible to define those AI interactions to a) contribute said components and patterns into the DS, and b) house the guidelines and principles in the DS documentation, same as with Content or A11y guidelines.

Yes

Not yet, but it's something that we're starting to talk about.

We don't have AI integrations, but a Chatbot in the platform and other AI powered experiences.

Unfortunately, I don't have any! (yet)

Yes

There is guidance within the design system site, but it's not "owned" by the design system team and comes from UX.

We've supported teams peripherally (smaller assets for consumption or design feedback), but not directly with design system assets. Our DS team is small and has separate priorities, while those working on AI are moving much faster than design system processes or tooling currently does. It's not that we don't want to work on these things (we've asked a few times what's needed from us), but in all likelihood it'd take our full capacity to do so. There's also a lot more privacy and feature considerations that aren't suitable for a more public design system environment.

We are only just now starting to add AI integrations into our product, so we haven't quite reached the point of needing to support integrations. Just thinking about how to best support this stuff in the future.

Yes

I’ve got nothing, really sorry

I prefer to minimize the use of AI as much as possible. Too many products are trying to cram it into every feature, whether or not it's actually needed or helpful for users. More attention should be given to the ethical and environmental concerns around AI, rather than blindly adopting it's use.

Yes

The design teams are currently trying to adopt design systems so we haven’t anything out using the design system. However, there are ai products in the company that I’d like to utilize and I’ve heard of other companies using the design system to do this. I am very curious about this weeks topic.

Yes

We designed buttons and indicator icons to visually identify the features or functionality that use AI in our products, in contrast to those that do not.

Yes

We just invested in helping out AI/ML team create up AI loading affordances that will be implemented into the design system

Yes

We're just getting our feet wet with AI integration.

My company is both trying to jump on the AI bandwagon and is worried about legal issues with using it. We’ve hired an AI dev team, and so far the only live things we have are a chatbot on one product and some AI summaries of user reviews that are clearly labeled. The VP of tech is pushing hard for us to use more AI, but the editorial team is worried about losing credibility and the legal team is telling us to slow down and run everything through them. So lots of mixed messages.

Only example I can think of is not using public AI systems (like copilot) for recommendations on Design or Code improvements, as the models they have been trained on are imperfect and the suggestions are often not helpful/correct

Yes

We created buttons and indicator icons for our products to use to visually identify the features and functionality that incorporate AI (in contrast to those that do not).

Yes

Someone on my team is in the process of writing guidelines about it. Such a great timing for this episode of The Question.

The company I work for started to offer a LLM subscription for everyone to use. We're in the process of writing guidelines to align how/when people should use AI in an enterprise level.

We’re mostly using AI on internal processes, not a tech company so I think we’ll be slow to adopt AI

Yes

We built a tool for generating alt text as part of the greater team.

Yes

Our design system includes patterns about how to represent AI functionalities to our customers.

Yes

I'm not on the design system team, but rather a consuming team. It was determined that the AI features being added to our products are too much in flux for them to be designed by the design system team. The team building the AI features operates much like a start-up and they are constantly iterating. We tried having the design system team take over that work but they felt they were a bottleneck and therefore decided not to get involved until the AI features are more stable.

Our team has not supported AI integrations because our product does not yet have AI integrations yet. However, I think this is something we need to be considering for the future. I've been looking at how Carbon defines design system AI best practices and am getting some inspiration from that. Would be interested to see other examples of how companies/products are handling that.

We currently have no support for AI integration.

Yes

In progress!

We have one experimental AI product right now, and I'm working closely w/ their designer on how to include AI design patterns into our system. They're using the core DS and as he is designing new flows and ideas we're discussing what could be included.

In our case, I don't know how much some of these patterns might be useful for other teams, but I do strongly feel that if you're a company doing AI integration across multiple products that the DS should have an opinion.

We're also working to add design system language guidelines into Cursor rules etc.

We do not provide any specific guidance yet. Mostly because our main audience isn't using it at all and is very hesitant about it. Any sort of automation is still treated skeptically even when we encourage it.

Yes but not for a product that used the design system.

On OpenStudio, a Discord-based storytelling event, I led design for a genAI co-creation experience. Participants used AI to help build a webcomic—but most were unfamiliar with prompting or AI workflows.
The website to explain the project had a design system, but there was nothing for the AI tool. It lacked patterns for guiding prompts, handling failure, or helping users iterate. I had to create everything from scratch:
* A UI for reviewing multiple generations
* Visual cues to separate AI-generated content from human input
* Microcopy to explain the unpredictable nature of generative AI
That project showed me how most design systems still focus on visuals, not behavior. They don’t support the kinds of interactions AI tools require—like navigating uncertainty or building user trust.
This gap reminded me that in 2019, I tested and redesigned an AI-driven mock interview tool for students. Users were hesitant and confused. The interface used a blue-dot avatar, offered no way to skip questions, and made assumptions about traditional career paths. For example, it couldn’t support a computer science student applying to a biology internship.
Even now, I see similar issues. I recently applied for a job where the company was using an AI application tool. It asked 3 questions—2 were nearly identical. Does having 2 real questions hurt my chances since I didn’t get to discuss the full range of my skills? The flow was linear, impersonal, and hard to engage with.
In contrast, when I designed a generative AI interface for a state agency last year, we made flexibility a core feature. We used a custom GPT in a chat format, so users could steer the conversation, shift direction midstream, and build confidence at their own pace.
Design systems need to evolve. AI demands new patterns—ones that support uncertainty, adaptability, trust, and real dialogue. Even lightweight frameworks with clear examples could help teams design more thoughtful and inclusive AI experiences

Yes

While individuals may be using AI to assist their work we do not have standards or practice established for team or division AI use.

All AI integrations prioritized by the company are for internal employee use, not customer facing. As such, these initiatives do not require using the design system and so the design system team has not had to address a use case.

Yes

We don't have anything to share on this at the moment, but hope to soon!

We haven't supported AI integrations

AI integrations, yes, meaningful? Maybe!

Yes

At this point it's basically a set of examples of what AI integrations are possible apart from chatbots.

There had been experiments but I don't think we have anything live now.

We haven't really thought about it other than providing lego bricks/components for the front of the AI integrations.

Company is behind in AI integrations

The bank I work for has been very cautious (slow) around integrating AI features in order to be absolutely sure our customers' data is protected.

We've been looking into how we can support, but are just too pressed for time and budget.

Nothing more than the usual requests - new icons and components.

We currently are not supporting any AI integrations in our design system I would be keen to understand how AI can be leveraged for DS Support, contribution and gov ernance

We primarily work in the insurance industry, where regulations make it hard to start bringing in AI integrations, so this is not something we have had to consider yet.

I've used Cursor AI to build entire test suites in our React library but am having trouble getting our development team to experiment with these tools in a meaningful way. I have not found a use case for Figma or design tools.

Haven't had the time to look into any integrations, but we did create a custom gpt for a token rewrite initiative. Essentially the instructions were things that we valued, standards we want to enforce, desired outcomes, and other project details that the group agreed on. Still early in testing to see if it pays off or note.

i'm not sure how much has shipped, but there is a big push around AI and more roles aligned to AI

i don't have any stories, but we have not at all supported AI integrations. there are a lot of controls around AI w/in a Bank setting. But, there is still momentum behind integrating AI and our DS team has not leaned in at all.

Yes

Our system outlines a standardized style for AI powered components, how to communicate AI was being utilized and provide system status to users. These standard have since gone through a re-brand via an internal pod who took over ownership of AI features. We will be adopting their new guidelines and branding to the system.

A teammate at my current company wants to spin up an AI assistant for our design system based on documentation, but we aren't yet in a place where we have good documentation to feed it.

There is an internal AI product being developed but the design system team has not been involved in it.

Our organization has not implemented any broad AI integrations. We recently received licensing for Microsoft Copilot and will be exploring this tool and identifying how we might pull in AI to websites. There is a university hosted website that speaks to approved tools and guidelines and shows events and workshops related to AI discussions.

Spending such a long time in an old healthcare organization, doing anything with AI is painstaking. There was sub-par infrastructure, few approvals, and little prioritization from leadership.

Some product teams were playing around with AI chatbots, but since this team was an engineering strategy team, there was no considerations in consuming the design system until far too late.

In healthcare, observations have noted a bias toward using AI recommendations if presented before a provider documents their recommendation. We're thinking about how does that work in a billable, time-focused environment.

The design systems I've built have basically had an input field and chatbox, but never any AI guideslines beyond that. I can't wait to hear what others have done.

Teams that consume our design system support a lot of AI initiatives, but don't use the DS for said initiatives (because we don't support AI yet)

We have not yet launched our Alpha so we are very young in our maturity stage. But many of our teams have AI driven initiatives so we will start to consider how to support them in the future.

In the works for a beta later this year

The product design teams are the ones leading this work with minimal collaboration from the DS standpoint. So far, support has primarily focused on UI and visual aspects, aiming to establish a standardized chat and interaction pattern across teams working on AI assistants.

Not there yet in regards to a system to include AI

AI (at my former company) was its own entity with it's own sub-system.

There was a AI design system that was separate from the core design system. Visually, the UI elements have a distinct look from the core system/product.

The system has it's own set of guidelines.

Yes

At first, the Design system team was trying to document all of the AI guidance teams would need. But this became impossible as teams forged ahead to deliver AI features, and the guidelines on the design system site quickly became outdated and impossible to keep up.

We recently switched to a new team called Growth or Future or something like that, that is responsible for documenting the AI guidelines separately from the design system. The design system team hopes that at some point in the future, the AI guidelines which are now hosted on a separate website called an AI Hub will be rolled into the design system site.

(sorry for any typos or grammar issues… I’m speaking my responses to my phone)

Yes

Some of the design system primitives have been used in more local (but still reusable) patterns.

In progress

Our AI features have been built in pages that cannot consume our design systems, and our design system has provided *visual guidance* only on aligning styling for AI features so that they still look like our DS components, but it has not provided guidance on interaction patterns or best practices for building AI experiences.

Working on it, planned for a July release.

Yes

Starting with a contribution from one product team and understanding the direction several other product teams were moving towards it became clear the need for the design system team needed to gather information quick and put together initial guidance and continue to pivot and refine as the landscape of AI is constantly changing as well as product team testing with different user groups / roles across our enterprise digital experiences.

Unlike UI patterns we've had on the web for years, the difficulty is that AI is an emerging pattern. It's challenging to suggest systems to support something we're just learning about ourselves. I think many of us are also unprepared with new, mostly non-visual medium. I think we should consider pivoting to inclusive systems, where the system conforms to the user's preferences over the brand.

Our team is trying to figure out ways to integrate and leverage AI, but the majority of leading edge stuff comes from product designers and marketing.

The team started to release but then pulled it back because the product team leading the AI efforts keep moving too fast and iterated beyond what the system had defined.

As the chat UI was getting applied to several products we asked our design system team to improve the product team's introduction of new components. But the product team that was making the request did not pause long enough to consume what the design system team was documenting and publishing and kept iterating on patterns based on user feedback and stakeholder requests.

So the work that the design system team was eventually pulled back from our design system and we communicated to products to work with the AI team directly and for designers to rely on documentation and Figma components from our AI library.

We are attempting to collaborate across the product design team on how to standardize other AI features.

The one example of AI being incorporated into the design system components is our rich text editor where we embedded AI features to generate/modify text. The design system adopted the design interaction from the AI team and polished it as they typically do a great job refining everything they work on.

We did a support chat bot, but that was only the easy idea. We're looking at more ways of introducing AI functions.

We're still trying to understand AI and how it might fit into our product.

Yes

I have only worked on a chat bot shared library/design system.

I am trying to learn as much as I can about design systems. I would like to attend along with my friend Laura Graham.

Yes

I left the company before our AI features were integrated into our design systems, but I heard about a special "AI library" having been set up.

We are working on integrating AI into our internal processes (ie. using it for research, building a knowledge base for support questions), but we haven’t yet had a use case where we needed to support ai integrations for a consuming team.

Unfortunately, I don’t have any experience discussing AI in the context of a design system, but I’m very curious to hear about your experiences, guys!

We're currently doing early planning around support for chatbots and conversational design UI.

Yes

We are just now starting to explore how we can integrate AI into our system space.

Our design system has not yet support specific documentation, yet. We are first looking into Figma AI access on features and go from there. To be honest, we feel a little bit behind from a design system AI support perspective.

They did not know how to build components that had to have that extra information that the content is being generated by AI

Yes

AI tooling is SO early-days in terms of what it even means to have AI integrations. Are they in-house, built solutions? Vended 3rd party solutions? A chat bot? A behind-the-scenes integration?

Organizations seem to not have a good handle on which level of AI they ultimately want, aside from "yes, we need to focus on AI...". Some are doing this very well, but many are still flying blind, in a certain sense (based on talking with other peers).

We've taken the approach, like any other: when there's semi-validated but REPEATABLE patterns, flows, aspects of the UI, interactions, etc, then we (the system team) would be happy to step in and help standardize; until then, we'll keep tabs on the work as it progresses.

we're currently on a journey of creating a design system (mostly done) and also redefining how we build software products with AI. So with that as context, we're looking at AI and Design Systems as more symbiotic partners in the process used to create software products...

Yes

Fluent does provide guidelines for Copilot features for other teams to integrate within their product verticals for both appearances and high level interaction patterns. However, with hundreds of concurrent experiments run by different teams and with new capabilities that are constantly launching, guidelines tend to get outdated rather quickly. So we're now focused on giving guidance to cosmetic elements in order for the teams to look and feel coherent.

100