The Future of AI-Powered Design Systems

As we move deeper into 2025, the intersection of artificial intelligence and design systems is creating new opportunities for designers, developers, and product teams.

For me, this topic is not purely theoretical. My thinking around AI-powered design systems began before AI became the center of every product conversation. At The Home Depot, I proposed an early assistant-driven direction inside a CRM experience and created a design roadmap for how structured interface patterns could help surface the right information to associates at the right moment.

At the time, the work was not about generating components with AI. The technology and organizational readiness were not there yet. The work was about identifying which components, patterns, and information structures could support an early form of AI-like governance inside the application.

That distinction matters.

The first phase of the roadmap focused on surfacing important customer information directly inside the associate workflow, reducing the need for service agents to search across the system during a call. This became the foundation for the KnowMe Cards, a product direction that earned recognition and helped establish the value of context-aware service design.

The later roadmap imagined a more assistant-like experience called Homey, a Home Depot-inspired helper that would live inside the CRM. Similar in spirit to Clippy, Homey would help customer service associates by pulling forward relevant information, suggesting context, and supporting service calls without forcing the associate to hunt through disconnected screens.

That future assistant layer continued after I moved on. I was promoted because of the strength and impact of the design work, and another designer later carried forward the roadmap direction.

Problem Statement

At the time this work began, I was a UX designer working across CRM experiences at The Home Depot. I had spent about four years designing service architecture for CRM systems, primarily within Salesforce-based environments.

Across the teams and workflows I supported, I kept seeing the same problem: customer service associates did not have easy access to the information they needed during live customer interactions.

A service call might begin with a simple customer question, but the associate often had to pause the conversation, place the customer on hold, and search through another system that was not connected to the CRM. In some cases, the associate had to contact another support group or contact center because the information lived behind a different operational wall.

This created a frustrating pattern. The associate had to repeat themselves. The customer had to wait. The call lost momentum. The service experience became harder than it needed to be.

The issue was not that associates lacked skill or effort. The issue was that the information architecture did not match the reality of the service interaction.

The organization had effectively walled off important customer and operational information. Associates were given access to the information someone had previously decided they needed, but the actual customer interaction often required more context than the system made available.

That gap became the foundation for the work.

The problem was not simply, “How do we make the CRM easier to use?” The deeper problem was:

How might we surface the right customer information inside the associate’s workflow so they can support the customer without repeatedly leaving the conversation to search disconnected systems?

That question led to the first phase of the roadmap: the KnowMe Cards. The goal was to bring important customer context into the CRM experience at the moment it was needed, instead of forcing associates to hunt for it across disconnected tools and teams.

Research Finding: The Information Already Existed

One of the most important findings from the research was that the customer had often already provided key information before speaking to an associate.

The IVR system was collecting information from the customer before the call ever reached a person. The customer had already answered questions, selected options, and provided context about why they were calling.

However, that information was not being surfaced clearly inside the CRM experience when the associate received the call.

As a result, the associate often had to ask the same questions again. From the customer’s perspective, this created an immediate breakdown in the service experience. They had already given the information to the company, but once they reached a human being, it felt as if the conversation had started over.

From the associate’s perspective, the problem was equally frustrating. They were expected to resolve the customer’s issue, but the system did not provide the context that had already been collected upstream. The associate had to repeat discovery, place the customer on hold, search disconnected systems, or contact another group to locate information the customer believed the company already had.

This changed my understanding of the problem.

The issue was not only that information was walled off across internal systems. It was also that information collected at the beginning of the service journey was not following the customer into the live support experience.

My first design response was practical: collect the relevant IVR information and surface it inside the CRM when the call reached the associate.

That became an early foundation for the KnowMe Cards. The idea was to give associates immediate customer context at the point of service, including information the customer had already provided, so the call could move forward instead of starting over.

The Evolution of Design Systems

Design systems have evolved from simple style guides into comprehensive ecosystems that support entire product families. They now influence how teams design, build, document, test, and scale digital products.

With AI, a new paradigm is emerging. Design systems can become more than static libraries. They can become intelligent frameworks that help teams maintain consistency, improve accessibility, reduce repetitive work, and support better decision-making across product experiences.

In practice, this does not always begin with a chatbot or a fully autonomous assistant. Sometimes it begins with something much more practical: a component that knows what information matters most in a specific workflow.

That was the lesson from the CRM work. Before an assistant can be useful, the system needs to understand what information should be surfaced, where it should appear, and how it should support the user’s decision-making.

The design system was not just a set of visual rules. It was becoming a way to govern how information appeared in a service workflow. It needed to answer questions that were larger than spacing, color, or component variants.

What should an associate see first?

What information has the customer already provided?

What context is most useful during a live call?

What should be visible without forcing the associate to search?

What information should remain controlled, protected, or conditional?

Those are design-system questions. They are also early AI-readiness questions.

From Static Components to Context-Aware Patterns

The KnowMe Cards represented a shift from static interface components toward context-aware service patterns.

A traditional component might display a customer name, account status, phone number, or recent order. A context-aware component asks a better question: what does the associate need to know right now in order to help this customer?

That difference matters.

The associate did not need every piece of customer data. They needed the most useful information for the current service moment. They needed the customer’s known context, relevant history, call reason, and key signals that could reduce repetition and unnecessary searching.

The purpose of KnowMe Cards was not to decorate the CRM. The purpose was to create a better service interaction by bringing the customer’s context forward.

This is where the work began to connect with what we now recognize as AI-powered product thinking.

Before an assistant can recommend anything useful, the product team has to define what “useful” means inside the workflow. Before a system can proactively surface information, the team has to understand what information matters, when it matters, and how it should appear.

KnowMe Cards helped establish that foundation.

They created a pattern for surfacing customer information in context. They reduced the need for associates to hunt across disconnected systems. They gave the service experience a clearer information hierarchy. They also created a path toward a future assistant experience that could live inside the CRM.

That future assistant was Homey.

Homey: The Assistant Layer on Top of the Information Architecture

Homey was the proposed assistant concept that extended the roadmap beyond the first phase.

The idea was inspired by the familiar pattern of a helper inside the software experience. In plain language, Homey was a Home Depot-inspired, Clippy-like assistant that would live inside the CRM and support associates during service calls.

However, Homey was not imagined as a gimmick or a floating character that interrupted the user. The real value was not the personality of the assistant. The value was the information architecture underneath it.

Homey would have been useful because the system had already begun identifying which customer information should be surfaced, how it should be organized, and where it belonged in the associate workflow.

That is the part many AI conversations skip.

A useful assistant is not just a chat window. A useful assistant is a product layer built on top of structured context, clear workflow rules, and a real understanding of user needs.

In this roadmap, Homey would help associates by pulling forward relevant information, proposing useful context, and supporting service calls without requiring the associate to leave the conversation repeatedly. It was the next logical step after KnowMe Cards.

KnowMe Cards surfaced important customer context. Homey would use that foundation to provide more proactive support.

That is why I now see this work as an early example of AI-powered design-system thinking. It was not generative AI in the current sense. It was not automated component creation. It was a roadmap for moving from static CRM screens toward a more intelligent, assistant-supported workflow.

Key Trends

The lessons from this work connect directly to where design systems are headed now.

Context-Aware Interface Patterns

AI-powered product systems need interface patterns that can surface relevant information in context. The interface has to help users see what matters without overwhelming them or forcing them to search through unrelated screens.

In the CRM work, this meant identifying the customer information an associate needed during a service call and bringing it into the workflow at the right moment.

Assistant-Ready Design Systems

Design systems will need to define more than visual components. They will need to define behavioral rules, information hierarchy, workflow context, and approval boundaries for assistant-like experiences.

A design system that supports an assistant cannot stop at buttons, cards, typography, and spacing. It needs to define how the assistant should behave, what information it can surface, when it should intervene, and when the human should remain fully in control.

Intelligent Accessibility and Usability Support

AI can support accessibility and usability by helping teams identify issues, suggest improvements, and apply standards more consistently. However, designers still need to validate whether those suggestions are appropriate for the actual user and workflow.

Accessibility is not only about technical compliance. It is also about reducing unnecessary cognitive load, repetition, and confusion inside real work. When associates had to ask the customer for the same information again, search disconnected systems, and manage the pressure of a live call, the system was creating avoidable friction.

Better information surfacing was a usability improvement. It was also a service-quality improvement.

Predictive and Proactive Service Patterns

Systems can learn from workflows, user behavior, and repeated service needs to suggest what information should appear next. The goal is not to replace the associate or user. The goal is to reduce friction and support better decisions.

In a service environment, proactive support should help the associate stay present in the customer conversation. It should not create more noise, more tabs, or more places to check.

Real-World Application

Before leaving The Home Depot, I proposed an early AI-powered design systems direction and created a roadmap for how the team could begin moving from static CRM screens toward a more intelligent, assistant-supported workflow.

The first phase focused on surfacing critical customer information inside the CRM experience. Instead of making customer service associates search across the application during a service call, the system would bring forward the most important customer context directly in the associate’s workflow.

That first component became the foundation for the KnowMe Cards.

The purpose was simple: give associates the right customer information at the right time so they could understand the customer’s context quickly and support the call more effectively.

The roadmap then extended into Homey, a future assistant concept. Homey was imagined as a Home Depot-inspired, Clippy-like assistant inside the CRM. It would help pull out relevant information, propose useful context, and support the associate during service calls.

The important point is that this was not component generation. It was early AI-readiness through product architecture, component strategy, and workflow governance.

The work explored how the system could:

Surface the most important customer information during service calls

Carry forward information the customer had already provided through the IVR

Reduce the need for associates to ask repeat questions

Reduce the need for associates to search across disconnected systems

Establish reusable interface patterns for customer context

Create a foundation for a future assistant-like experience

Support better service decisions without removing human judgment

Prepare the CRM experience for more intelligent workflows over time

After I moved on, the team began implementing the direction outlined in the roadmap. The assistant layer continued as a future phase under another designer’s ownership, after I had been promoted because of the design work and its impact.

That experience shaped how I think about AI adoption in product design. The strongest AI systems do not begin with automation for its own sake. They begin with a clear understanding of the workflow, the user, the information hierarchy, and the decisions the system should support.

Why the Roadmap Mattered

The roadmap mattered because it created a path from immediate service improvement to future intelligent assistance.

The first phase did not require the organization to leap into a fully autonomous AI assistant. That would have been too much too soon. The smarter move was to begin with the information problem.

What did the associate need?

What had the customer already provided?

What information was trapped somewhere else?

What could be surfaced safely inside the CRM?

What pattern could be reused across similar service workflows?

By answering those questions first, the team could create a practical foundation.

That is how responsible AI adoption should work.

It should not begin with a dramatic promise. It should begin with a specific workflow problem and a clear understanding of the information needed to improve it.

In this case, the problem was not mysterious. Customers were telling the company what they needed before they reached an associate. Associates could not consistently see that information when the call began. The system then forced both people into a worse experience.

The roadmap gave the team a way to fix the immediate pain while also preparing for a more intelligent future state.

That is the bridge from KnowMe Cards to Homey.

Looking Ahead

The future of design systems lies in intelligent, adaptive frameworks that evolve with user needs while maintaining consistency, accessibility, and product standards.

For design teams, this means expanding what a design system is responsible for. It is no longer enough to define buttons, cards, colors, and typography. Future design systems will need to define how information behaves inside a workflow.

They will need to answer questions such as:

What information should the system surface first?

What context does the user need before making a decision?

What should an assistant be allowed to suggest?

What must remain under human control?

What patterns can be reused safely across similar workflows?

What information should travel from one part of the customer journey to another?

What should the system do when information is incomplete, uncertain, or unavailable?

These are design-system questions as much as they are AI questions.

As AI becomes more common inside enterprise products, design systems will need to govern not only what the interface looks like, but how intelligent features behave.

They will need to define the rules for context, timing, escalation, visibility, privacy, and human judgment.

A poorly designed assistant can create more work. It can surface the wrong information, interrupt the user, overstate certainty, or bury the associate in suggestions that do not help the call.

A well-designed assistant should reduce friction. It should make the user’s next best action clearer. It should support the human, not compete with them.

That is why the design system matters.

What This Means for Designers

Designers will continue to shift from manual production toward strategic oversight. That shift includes:

Defining design principles and constraints

Creating assistant-ready patterns

Structuring information for intelligent workflows

Designing approval and escalation boundaries

Validating AI-supported recommendations

Protecting accessibility, usability, and product quality

Understanding how data moves across the service journey

Identifying where human judgment must remain central

The designer’s role does not become smaller in this future. It becomes more strategic.

Designers will need to understand where AI belongs in a workflow, what information it should use, what it should never assume, and how users should remain in control.

They will also need to become better at designing the invisible parts of the experience: the rules, data relationships, system behaviors, and governance models that determine whether an intelligent interface actually works.

That was the lesson I learned from the CRM roadmap.

The visible component mattered, but the real design challenge was underneath it. It was the information architecture. It was the workflow logic. It was the gap between what the customer had already told the company and what the associate could actually see.

Once you understand that gap, the product direction becomes clearer.

The goal is not to make the interface look more intelligent. The goal is to make the system behave in a way that genuinely supports the person doing the work.

Conclusion

AI-powered design systems represent a meaningful shift in how teams approach scalable product design.

The strongest outcomes will not come from simply adding an assistant to an interface. They will come from designing the underlying system that allows an assistant to be useful in the first place.

That system includes components, patterns, information hierarchy, workflow rules, governance, and human review.

My early CRM work at The Home Depot helped shape this point of view. The first step was not an AI-generated interface. It was a practical component strategy for surfacing customer information in context through the KnowMe Cards. That foundation made the future assistant concept, Homey, possible.

The work began with a very human problem: customers were repeating themselves, associates were searching through disconnected systems, and service calls were being slowed down by information that already existed but was not available in the right place.

That is the kind of problem AI should help solve.

Not by replacing the associate.

Not by pretending the system knows everything.

Not by adding a chatbot on top of a broken workflow.

The better path is to design the system so the right information reaches the right person at the right moment.

The key is not choosing between automation and human creativity. The key is designing systems where intelligent support can help people make better decisions, while experienced designers continue to define the standards, constraints, and product direction.