Sierra Nelmes, Principal Design Researcher, and Brent Schneider, Director of Product Design, have spent the past year exploring the role of artificial intelligence (AI) on a recent Veteran healthcare project. This effort is part of VA’s larger AI strategy for adopting high-impact artificial intelligence to improve services for Veterans.
For both Sierra and Brent, the experience of interacting with generative AI as system designers in a product environment has been about steady learning, asking hard questions, weighing tradeoffs, and keeping people at the center of every design choice.
“We find ourselves at a multilayered inflection point where our industry and government partners are being asked to ‘go AI,’ whether that’s building their own in-house custom solution or buying something off the shelf,” Sierra shares. “Then, simultaneously, we are also trying to explore what that means for us as a company. As we are partnering with our clients to understand their needs and the end-user needs, we are also learning how to build and design using AI, and communicate those learnings.”
Sierra and Brent approach the moment with cautious optimism. AI is here to stay and evolving rapidly, but it also presents ethical challenges and real human consequences. At Coforma, our role is to guide government partners through these inflection points, helping them explore what AI can accomplish while prioritizing thoughtfulness, transparency, and human impact.
In this article, Sierra and Brent share what they’ve learned about designing and developing AI systems, including how humility shapes their approach, why small design decisions matter, and how AI can bring human teams closer together.
Leaning in as AI Learners
Sierra and Brent shy away from the term “AI experts.”
“It’s one of those things where the more you learn, the more it feels like the less you know,” Sierra shares. “I don't think there's an ultimate destination where somebody is the go-to expert on AI, because there are so many dimensions to it. I don't think that it's ever going to be static. Just when you think you know enough to be an expert, it’s going to change.”
Instead, she frames her approach through the “empty cup” metaphor drawn from Buddhism. “There’s a kind of expertise that comes with being new to a topic,” she reflects. “When your cup is empty, there’s more space to fill. If you can stay in that learning mode, you absorb more.”
Brent agrees, connecting perceived AI expertise to the Dunning-Kruger effect (the idea that early confidence in a new subject quickly gives way to humility as you realize the true depth of complexity).
“I’ve heard that it takes about 50 hours of study just to move past the beginner stage. At first, you can sound like you know a lot, but then you hit that drop where you realize how much more there is to understand. That’s where we are with AI. We’re learning as we go, and the ground is always shifting.”
For both Sierra and Brent, leaning into the “learner” label allows them to adopt a posture of curiosity, humility, and openness.
The Unseen Design Decisions Behind AI Systems
Earlier this year, headlines told the story of a California family suing OpenAI after their son tragically took his own life. According to reports, the chatbot he’d been using—which was designed to be agreeable—began to confirm his darkest thoughts.
The story reinforces the real-world stakes of AI design and reminds us that even subtle design decisions can have a profound impact on humans interacting with generative AI technology.
“One of the risks with AI is what happens when conversations like this one go on too long,” Sierra shares. “Without lasting safeguards, systems can drift from their intended purpose and begin to reinforce harmful patterns.”
There are decisions users never see—like when and how to reintroduce guidance to the system—that can make a huge difference. “If those guardrails aren’t there, the AI can veer into unsafe territory,” says Sierra. “Companies and teams now have to be able to prioritize and make time for building those guardrails.”
The challenge is that many AI systems are designed to be agreeable. As Brent points out, “I haven’t yet found an AI chatbot that says, ‘No, you’re wrong.’ It’s always, ‘That’s a very good point.’ Many of these AI systems are designed to make humans feel like they’re on to something.”
This design choice, while seemingly positive, can reinforce dangerous thought patterns, creating what he calls another “hidden dark UX pattern of AI design” that isn’t very human-centered in its outcome.
As Coforma continues to explore AI design and development, our team is learning how to navigate these challenges. “We’re learning how to evaluate these parameters in a controlled and safe manner as the technology continues to evolve,” Brent shares.
“There are many variables at play,” Sierra adds. “From how the model is configured to what instructions are re-injected into a conversation, it’s an entire ecosystem of choices that influence the end-user’s experience.”
How, then, do we ensure we’re designing AI systems responsibly? According to Sierra, “The day-to-day processes a team has chosen to curate, implement, and maintain throughout a project are important. This includes prioritizing tools for evaluating and monitoring downstream user impact so we can always trace what’s happening behind the scenes and maintain visibility.”
AI as a Catalyst for Human Innovation and Collaboration
We’re already seeing how developing and using AI technology influences the way we interact with each other in a project environment.
Encouraging Creative Technological Experimentation
Exploring this emerging space of crafting AI products enables us to come together in new ways to design, test, and make informed decisions.
“These are new territories of conversation for our teams,” Sierra says. “When we bring cross-disciplinary groups together to tackle challenges about AI, we’re asking new questions, exploring totally novel topic areas amongst ourselves and with stakeholders.”
Similarly, leveraging AI as part of our own process became a springboard for creative experimentation. “Because we have neurodivergent intelligence on our team, it opened up new ways of seeing and applying AI in the workflows we were designing for end users,” Sierra explains. “That diversity of thinking changed the system we built and how we built it.”
Blurring Boundaries Between Disciplines
AI also has a way of blurring boundaries between disciplines.
“It’s like a watercolor painting,” Sierra says. “The edges between developer, designer, and data scientist roles softened. AI pulled all the disciplines closer together, with more overlap and more interdependency. Not that the boundaries were ever completely rigid, but the shift was more pronounced here.”
When asked if this shift makes specialization less important, Sierra is quick to emphasize that human qualities matter most.
“Skill sets can change quickly,” she says. “What matters more is how teams connect and communicate, how they think together. A side benefit of working on AI is that it strengthens human collaboration. To tackle problems this complex, you really have to put your brains together.”
Breaking Down Language Silos
Brent notes that AI tools help dissolve the language barriers that can separate team roles.
“In the tech industry, there are a lot of single silos, even within one team,” he says. “AI can start to break those down. For example, on our recent project, I was able to use AI to adjust the CSS styling for some of the components and then put that code into a CodePen to share with engineers,” he explained. “Instead of passing design specifications in tooling using my language to define the corner radius and drop shadow, I could show them in their language how the styling was off.”
By translating design intent into something engineers could immediately work with, AI dissolved the silos of language and expertise, creating a faster, clearer path toward shared solutions.
This new way of interacting with technology is exciting, but we’ve observed it’s the creative collaboration—human interacting with humans—that truly unlocks the value of these emerging tools.
Human-Powered, Human-Empowering AI Technology
In many ways, the industry is trying to sell AI as if it were an intelligent entity. However, this project team made the intentional choice not to personify the AI service we designed.
“I believe that the over-personification of AI can lead people to mistakenly perceive it as truly intelligent,” Brent shares. “I’d rather frame AI as what it is: a highly developed algorithmic tool with the ability to process vast amounts of complex data and distill it into clear, understandable insights and execute performative tasks.”
“I don’t think AI alone can be thought of as inherently responsible or human-centered,” Sierra adds. “It all comes down to how humans are using this technology downstream as a tool and the design, technical, and product decisions made by the humans upstream who build it that lead to the AI product experience being human-centered.”
The project team chose to leverage AI’s strength in turning overwhelming information into a manageable resource, empowering human judgment and creativity, not replacing it.
In fact, one of the most impactful takeaways from Sierra’s recent work on designing and building AI has been how people-centered it turned out to be. “When I started the project, I was totally new to using AI. I didn’t even know about ChatGPT, so I really did start from scratch,” she recalls. “What I found most fascinating was realizing how much this work of crafting AI experiences for users comes back to human qualities like staying open, being self-aware, observing without ego.”
Sierra says it’s not technical skills that win the day in this environment of breakneck evolution, it’s qualities that shape how people connect and collaborate.
“You can learn skills fast. Tools will keep changing. But what doesn’t change is humans being human,” she explains. “It’s the same with the people we serve. Our ability to observe, empathize, understand, and communicate about lived human experiences doesn’t go away. That’s the constant.”
AI will keep evolving. The methods and interfaces will look different year after year. But the heart of responsible AI implementation remains humans understanding humans, and designing technology that supports, not replaces, that connection.
“Even with AI in the loop,” Brent says, “what matters most is that we don’t lose sight of people helping people.”
About the Authors
Brent Schneider, Director of Product Design at Coforma, is a seasoned leader with 25+ years of experience. He’s focused on guiding teams to deliver human-centered products and services that create lasting impact for government agencies.
Sierra Nelmes is a Principal Design Researcher at Coforma with 15+ years of experience spanning startups, Fortune 500s, and civic technology. She specializes in crafting product experiences and systems that honor nuance, build trust, and respond to real-world complexity.
