The horse and the rider: Steering AI toward human-centric design
Why speed alone doesn’t deliver value
While GenAI has revolutionized the pace of design, from initial research to final implementation, we must be careful not to mistake velocity for value. AI can simulate plausibility, but it cannot simulate humanity. It lacks the lived experience and emotional intelligence required to create truly empathetic solutions.
In traditional Human-Centered Design (HCD), the Double Diamond serves as a vital safeguard. It separates the act of properly framing a problem from the act of generating solutions. AI threatens to collapse this process entirely, galloping from brief to execution in a single leap. When we allow this, we create a dysfunctional relationship where the “how” replaces the “why.” The challenge for modern designers is clear: How do we bridge the raw speed of AI with the intentionality of the Double Diamond?
We’ve been exploring a metaphor to guide us: AI is the horse; the designer is the rider. The horse is a marvel of strength and agility, capable of feats no human could achieve alone. But a horse is also prone to charging toward the wrong horizon if left to its own devices. As riders, we provide the intent, the ethics, and the direction. To reach a truly human-centric goal, we must stay firmly in the saddle.
Here’s how we are currently partnering with our horses across our design process to ensure every output is high-quality, high-context, and human-led.
Discover
AI can map the territory, but it cannot navigate the specific stakes of a project.
At the start of a project, the objective is to ramp up on a complex new topic at lightning speed. AI can ingest thousands of pages of secondary research in seconds, offering a massive advantage in breadth. However, because it lacks the wherewithal to understand a client’s specific business stakes, relying on AI summaries alone is like building on sand.
The Horse
We use tools such as NotebookLM and Claude (configured with specific research “skills”) as expert clerks to accelerate the initial ramp-up. They aggregate sources, scrape competitor data, and identify industry vernacular at a pace no human can match. We typically use these notebooks throughout the project to continually have a central repository to search and query.
The Rider
We manually read the foundational documents for a project to build our own knowledge before relying on an LLM. Without this manual grounding, it’s very hard to discern what is truly important or if the AI is hallucinating industry logic. For example, for one project that relied on psychology terminology, we needed to read articles about the topic before we could discern whether a scientific term that an LLM spit out was creative phrasing or industry standard. We use the horse to cover the ground, but the rider ensures we aren’t being steered off-course.
Define
Synthesis is currently AI’s strongest suit, but a pattern is not a strategy.
As we move into synthesis, AI provides an incredible resource for querying specific questions and finding patterns instantly. But while the horse is great at seeing correlations, it might not be able to assign meaning or strategic potential. Humans are required to determine which data points actually move the needle for a business.
The Horse
We use Otter.ai to query hundreds of hours of transcripts for specific themes and pain points. This turns a mountain of messy data into a searchable library, saving days of manual tagging and memory-retrieval. No longer does a researcher have to pose to the project team “remember when that person said that thing?”
The Rider
Sometimes the quotes that AI transcripts surface feel too perfect for the scenario, and sometimes they are! We manually verify the quotes used to anchor a strategy to ensure it hasn’t been “Frankensteined” into something that sounds clean but lacks the user’s true intent or voice. We use AI to handle the labor of the data so we can spend our time on the relevance and story of the strategy.
Develop
Design is based on the lived experience of what could be, not the probability of what already exists.
Once we begin solutioning, AI shifts from a research tool to a visualizing tool. It offers the advantage of massive creative volume, generating dozens of visual variations in minutes. AI feels least strong as an ideation partner. Because AI is a prediction machine, it will always gravitate toward the most probable answer. The designer’s value lies in the improbable; the creative leap the machine would never suggest.
The Horse
We use image generators to add creative friction and stress test visual styles. While working with a global tech company, we combined hand sketches and mood boards with Midjourney prompts to establish a visual language and explore dozens of directions, so that a designer could jump into visualization faster and with more clarity and foresight.
The Rider
The human mind can push provocative “How Might We” prompts and selectively ignore certain constraints that a machine would naturally filter out. We don’t rely on AI for any finished solution, but use its output as a sparring partner to help us find the edges of our own imagination.
Deliver
AI can polish a product to the finish line, but it cannot sign the guarantee.
As we move toward launch, AI becomes a high-performance technical partner. It excels at that last mile of refinement, such as checking systems for errors and writing code. But the machine lacks a moral compass; it doesn’t care about accessibility standards or manufacturing ethics unless explicitly prompted.
The Horse
AI is a world-class troubleshooter. We use LLMs for technical heavy lifting such as writing code snippets, checking design systems for inconsistencies, and simulating edge-case user flows to see where a product might break.
The Rider
We humans perform the final audit to ensure intellectual property and ethical standards are met. We delegate the refinement to the machine, but we never delegate the responsibility. The human rider is the only one who holds the reins of accountability.
Wrangling AI isn't about working faster; it's about working deeper
By letting the “horse” handle the heavy lifting of data organization and iterative exploration, we as designers are freed to do what they do best: keeping the human at the center. The horse can run the race, but only the rider knows why the race is worth winning.