Artificial intelligence (AI) has firmly arrived in government and civic tech, gaining traction as a tool that can positively transform how federal, state, and local agencies serve the public. At Coforma, while we are actively adopting and designing AI technology, we’re not interested in taking a reckless “move fast and break things” approach.
Instead, we’re taking considered, intentional steps when it comes to AI adoption. We’re leaning into exploration and innovation, while also pausing before implementation to ask ourselves not just, “Can we?” but, “Should we?”
Being human-centered to our core, we want to fully understand what introducing a technology with such potential to amplify human impact means for us, our partners, and the communities they support.
In 2024, we formed an internal AI working group to explore this theme. This year, we rolled out Gemini for Coforma employees to integrate secure AI tools in their day-to-day practices, and we’ve been actively working on AI projects with state and federal partners.
We also recently welcomed our new Chief Innovation Officer, Victor Garcia, who will be piloting new ways Coforma can leverage this technology in product development and more.
“One of the things Coforma is really good at is doing human-centered design,” he shares. “I think we have an opportunity to be leaders in the space of human-centered AI.
“I’ve seen the conversation shift from the hype that was AI a year or two ago to how it’s actually being applied in meaningful ways,” he continues. “Learning what that looks like for us, and how we navigate it responsibly and with integrity, comes from building, trying things out, and adapting as the technology evolves.”
Victor is careful not to equate innovation with AI, saying, “I like to focus on what problem we are solving versus starting with the solution, and AI tools are just one potential solution for how we might solve things.”
In the coming months, Coforma thought leaders will share their experiences interacting with these new systems and tools. But before we put pen to paper, we realized we needed to define our terms. To do this transparently and collaboratively, we decided to run internal focus groups to gather perspectives on Coforma’s relationship with AI.
Here’s what we learned.
Defining What We Mean By AI
In our focus groups, we discussed AI broadly, with our scope including traditional machine learning models, natural language processing, and the large language models that are rapidly gaining traction today.
Participants were quick to point out the need for further specificity when we write or speak about AI in context.
Here’s a short glossary of terms from Gemini and Speaking of AI Design. Developed by Coforma’s Principal Design Researcher, Sierra Nelmes, it’s a community-powered glossary for designers and researchers working on AI product development. We’ll refer back to these terms throughout our upcoming content series.
Term
Definition
Artificial Intelligence (AI)
Computer systems designed to perform tasks that typically require human intelligence, such as learning, reasoning, problem-solving, and creativity.
Machine Learning (ML)
AI subset that uses algorithms to learn patterns and make predictions from data without being explicitly programmed for the task.
Generative AI
Type of machine learning that creates new content by learning the patterns and structure of existing data.
Large Language Models (LLMs)
Models trained on vast datasets to understand and generate human language by statistically predicting the next word in a sequence.
Hallucination
When a model generates plausible but false information.
Human in the Loop
Design pattern wherein humans review or approve AI outputs before they're finalized.
Defining Our Approach to AI
Plain language principles call on us to communicate clearly, with intention and rigorous attention to what we mean and how others interpret our message. Knowing this, when we set out to talk about the way Coforma interacts with and designs AI technology and systems, we wanted to be precise.
To home in on a more meaningful term, we assembled Cofolx from across disciplines, gathering perspectives on how Coforma defines its relationship with AI.
Five internal facilitators led 22 participants in three, one-hour workshops. Our goal was to explore what we really mean when we talk about using this technology, from implementing AI tools in our internal practices to designing new AI systems for our partners.
Cofolx offered insights and opinions, not as AI experts, but as technologists learning as they go.
We began by brainstorming the most appropriate words to describe, at a high level, how Coforma as a company should approach AI. We then upvoted these words to find which terms resonated most with the group.
The words that gained consensus were: responsible, human-centered, conscientious, principled.
We then asked participants to examine the meaning of these words more deeply in the context of Coforma’s AI use, including any potential negative perceptions associated with them.
This drove further discussion about what “responsible,” “human-centered,” and “principled” AI use look like at Coforma, with “responsible” gaining the most traction.
It also prompted discussions around what responsible AI use doesn’t look like.
Other Emergent Themes
Interesting throughlines surfaced across our conversations. Here are some of the most prominent examples:
- The need to be informed about what AI truly is, its capabilities, ripple effects, and inherent limitations, such as bias and hallucinations
- The decision to use AI should be a proactive, considered choice based on its clear advantages for a specific problem
- The importance of having humans in the AI loop vs. AI in the human loop
- The need for accountability and transparency
- The value of establishing principles that will lay the foundation for our processes
Coforma’s Responsible AI Principles
VP of Design, Melissa Casburn, has been leading responsible AI implementation at Coforma over the past year. She formed our internal AI Working Group, led the rollout of AI tools, and helped pen our principles.
“I’m a big fan of guiding principles,” she shares. “They help us translate our corporate values into actionable rules for decision-making. They give us a way to answer the question, ‘Does this decision align with who we are and where we want to go?’”
Melissa and others synthesized information from our workshops, evaluated existing responsible AI frameworks, and iterated with employees and leadership to arrive at Coforma’s guiding principles for responsible AI use. Those are:
- Stewardship: We view AI not as a goal in itself, but as a human-powered, human-empowering tool requiring deliberate and continuous governance.
- Intentionality: We will explore and experiment with AI thoughtfully, guided by purpose and measurable impact. When moving from exploration to production use, we will evaluate alignment with our mission and the outcomes we aim to achieve.
- Human Agency: We commit to using AI as a tool that supports and amplifies human capabilities, rather than a replacement for human intelligence and judgment.
- Impact: We will consider the impact and ripple effects of AI use, and weigh them against our goals of creating meaningful, positive outcomes that advance equity, access, and efficiency for the people and organizations we support.
- Accountability: We will actively seek to identify common AI flaws such as bias, drift, and hallucinations, and design solutions that address them.
- Privacy and Security: We will design and operate AI systems with a strong commitment to security and privacy, and adopt practices that safeguard personal information and maintain trust.
These principles are meant to institute guardrails against irresponsible AI use, but they’re also meant to guide us toward new possibilities, expand our capabilities, and help us innovate for our partners and the communities they serve.
As we’ve seen, AI advances at an exponential pace. Our principles and processes constitute a working definition that will be regularly evaluated, adapted, and updated as the technology, our understanding of it, and its social impact evolve.
Defining these principles is not a final iteration but a minimum viable product (MVP) that represents our progress toward building a culture of AI innovation.
“To me, responsible AI doesn’t mean avoiding risk,” says Victor. “It means managing it thoughtfully. It’s about treating these tools like any new craft. Something we learn by doing and get better at over time. Responsible AI isn’t about slowing down, it’s about steering with intention.”
About the Author
Amanda Harr is a curious, creative storyteller who weaves narratives that breathe life into brands. She believes in the transformative power of language and has a knack for synthesizing complex ideas into messaging that resonates with diverse audiences. A graduate of the Plan II Honors program at UT Austin and an award-winning creative writer, she’s on a mission to move hearts and minds with well-placed words.
