Over the past two years, I’ve been building tools with AI and for AI (agents, automation, machine learning, computer vision, data labeling for model training)—sometimes for companies, sometimes for internal products, sometimes for experiments that are never meant to see the light of day. And somewhere along the way, my entire approach to product design quietly shifted.
The Old Flow Wasn’t Broken — But It Was Slow
The design thinking framework still works. Workshops still work. Deep research still works. But in today’s software space—whether it’s an AI-powered tool or a traditional software or app—speed matters more than ever. AI has made things cheaper and easier to build, and it’s made the custom software market incredibly competitive, with companies able to do more on their own. Ideas age fast, and assumptions get validated (or crushed) in days, not months.
I didn’t even realize it at first—it happened slowly. We started consolidating, collaborating more closely with stakeholders, and going directly from assumptions to prototypes. That gradual shift changed the nature of our role—Software Design Strategists can no longer design in the same slow, front-loaded way. Instead, we have to move in tighter loops, with design, development, and product strategy happening almost simultaneously.
The New Product Design Flow (My AI-Era Version)
1. Capture Needs Up Front—Fast
Instead of multi-week discovery phases, I run short, targeted kickoff sessions with stakeholders and execs. In just a few hours, we:
- Document the problems we’re solving
- Identify users or proto-personas
- Capture business goals, impact, and KPIs
- Whiteboard existing or desired user flows
- Map out core challenges and brainstorm potential solutions.
It’s not about being exhaustive—it’s about getting just enough to start making.
2. Become a Subject Matter Expert—Quickly
AI has become my research partner. I use LLMs like ChatGPT or Claude to:
- Digest mountains of client documents in minutes
- Summarize user feedback
- Scan the competitive landscape
- Learn more about the product space / client industry. I built a SME Custom GPT to help me get up to speed faster.
- Synthesize specific types of information into structured formats
Within days, I can talk about my client’s industry like I’ve been in it for years. I’m not a true subject matter expert—but I can get close enough to hold my own in strategic conversations, maybe even closer than before. That kind of rapid understanding is critical when you’re making design decisions this early.
3. Distill It All into a PRD
I’ve learned to value a concise, transparent Product Requirements Document. It’s not a formality—it’s the anchor that keeps everyone aligned while we move fast. Business goals, user needs, constraints—they’re all in one place.
- Delivery, Design, and Development teams collaborate closely to craft it.
- We often use CustomGPTs or ChatPrd to structure and format the document.
- These tools can also help break down high-level user stories into actionable items for design or implementation.
4. Prototype Before the Workshop
Here’s the biggest change: we skip the little “d” design entirely at first. We’re not polishing UI—we’re already prototyping directly in code.
Why? Stakeholders want to see value delivered faster.
The pros:
- We can explore multiple directions early
- We avoid falling in love with a single idea too soon
- Tools like Cursor, Figma Make, Vercel V0, and Replit make this possible
- In AI products, functionality often shapes the design, not the other way around
5. Bring in Big “D” Design
This is ideally where we lean more into big “D” Design.
- Testing, getting feedback, iterating
- Refining the product direction—or decisively selecting the path forward
- Bringing in little “d” design strategically, updating the UI where it improves usability, accessibility, and functionality
This balance ensures we’re shaping the overall product vision while still making thoughtful, high-impact design decisions.
Why This Works for AI-Era Products
- Faster Feedback Loops: Real conversations happen when people can see and use something, not when they’re looking at sticky notes.
- More Critical Thinking: We’re forced to consider what AI or agents can handle, where AI can be brought in strategically, what tasks stay with the user, and how multimodal experiences change the interface.
- Idea Diversity: We can test multiple approaches without committing to one too early.
The Risks I’ve Learned to Watch For
Moving this fast isn’t without pitfalls:
- Core UX Principles Can Slip: Accessibility, usability, and delight need deliberate attention.
- Long-Term Costs Increase: If we release something clunky, fixing it later costs more.
- Products Lack Empathy: Why the “The design is very human” meme exists.
- Sameness: With AI-generated components, products can start looking alike—sometimes good for usability, sometimes not for brand or market differentiation.
A Word Of Warning
More companies are skipping the step of bringing in a dedicated product team or design expertise altogether. While it might feel faster or cheaper at first, it often leads to products that miss the mark on usability, market fit, and long-term scalability.
We’ve been here before. In the early days of the web, companies rushed to launch tools and websites without thoughtful design. Many of them failed—poor usability, unclear value, inconsistent experiences. That pain sparked the rise of design thinking and human-centered design. We’re entering a similar cycle with AI. For designers, this means that eventually there will be a renewed wave—a renaissance of human-centered design in the AI space. Those who are ready for it will lead the next chapter.
It’s Not About Throwing Out the Old Ways
I haven’t abandoned empathy or research—I’ve just rearranged them. They’re now threaded throughout the process instead of being front-loaded.
The truth is, this shift from little d design (UI polish, workshops, deliverables) to big D Design (systems, functionality, product shaping) is where the real impact happens—especially in AI-driven products.
Two years in, I can’t imagine going back—though if I’m being authentic, it’s challenging. I’m learning new things and I’m grateful to be working with AI, but I’m constantly afraid I’m going to be left behind. There are way too many tools to keep up with, and some days it’s honestly a struggle. Still, I know that everyone in the design space is trying to figure it out at the same time, and that shared uncertainty keeps me moving forward.