Design Work With AI
By Dan Harris, Head of Product & Design at Dogdrop
March 2025
The Path We’ve Been On
I’ve been designing for 10 years. Over the past 6 months, the tools that allow designers and product creators to bring ideas to life have changed with greater velocity than at any point in my career. I know that sounds hyperbolic. However, this sentiment is echoed by peers across roles—engineers, CEOs, designers, and CTOs. How we work, design, and collaborate is changing more rapidly than ever before.
In general, I think the recent innovations that have created Artificial Intelligence (AI) tools that allow us to move from idea to execution faster are mostly positive. But, with rapid change there will be some feeling of chaos. My focus for this essay will be on the rapid change that is taking place, and not on antidotes to the chaotic feeling.
For people willing to embrace change, this moment represents an inflection point where AI tools are eliminating traditional barriers, raising both the quality ceiling and accessibility floor of design work, and creating opportunities for those who can orchestrate these new capabilities to create exceptional products.
Note: I am viewing these tools through the lens of product & design for startups and small teams, where engineering resources are often the constraint.
From Idea to Creation
What You Create Is Important. The Tools Are Not As Important.
The undercurrent for this essay will be about the act of creation specifically for designing and thinking through products. Ideally we want to move from idea to a working product as seamlessly as possible. Sometimes we will want to do this quickly, other times we will want to focus on quality and not optimize for speed alone. The tool, or toolset at our disposal, is one important constraint on how much friction is involved to move from idea to finished creation. The ideal is as little friction as necessary in the tools to get us to the final result.
So I will cover specific tools and processes in this essay, but it will not be for the purpose of talking about “the best tool for x” or the “optimal framework for y”. The tools and processes that I’ll discuss are of importance because they allow us to making something useful and beautiful. Additionally, by following the evolution of these tools, and how they fit into our work, we can perhaps envision where the future of this type of work is moving.
Landscape of End to End Tools
For the first time in the majority of my workflow touches some AI Tool. The entire toolset is increasing quality and depth of thought, speeding up the workflow, and allowing entirely new possibilities for me as a designer. This section will cover briefly how these new tools are interacting in each phase of my workflow, along with why the sum is allowing my teams to have better results. We will also briefly summarize what “better results” means.
Usually most of my work moves from left to right on the visual above. Starting with a problem or opportunity that is laid out with the product team, then usually design engages next to figure out what the user should experience, and how they want to execute on the opportunity. Then once the broader team agrees on the direction, engineering begins to build it. It is usually far more organic and less linear than I just described it, but that is at least a useful framework in the direction of the workflow.
Product Thinking
On product teams, tools like Deep Research from Perplexity, Gemini, or ChatGPT are massively helpful for finding information on customer demographics, general market financials, research studies, and more. Instead of spending days clicking through many Google links and then aggregating the product research, this can now be done with a single action. Furthermore, if I want to interrogate the data further or start to question something specific I can do so.
Previously if I hit a dead end with data that existed in a specific research source, I couldn’t ask it to go deeper down the rabbit hole with me. Less time clicking through a bunch of links with unknown quality or depth. More time to think through how to solve the problem, or if there is a better problem to solve in the first place.
Once we have all the information that we think we will be useful to start the project, that is where tools like Google’s Notebook LM come into the picture. With Notebook LM you can drop in dozens of files, and start to contextualize your thoughts using all the sources. Previously tools like ChatGPT or Claude had a limited context window size for uploads.
For product and design when you are trying to create a solution given a vast array of information, each with varying types of information and structure, Notebook LM is very helpful. One example might be the marketing writing used on a specific webpage or view of an application. You might upload Google keyword data, competitor research, and customer interviews and use all of these sources to decide on writing that is better informed with a holistic understanding about the customer.
Reasoning models like o1 from ChatGPT I use as a combative argument partner. Sometimes your team may be too busy to try to strengthen a hypothesis or find weak points in your thinking, so the LLMs with the focus on improved reasoning I found are good theoretical sparring partners.
Design
That covers most of what I’m using for product thinking (though not all), so let’s move on to design. For designers the the gap between idea and an early version of software that we can have users experiment with is quickly being eliminated. The design tools and processes actually haven’t changed a great deal yet, though I suspect Figma is working on something big. I have reasons why I think the design part of the workflow hasn’t changed a great deal, but that should probably be it’s own separate essay. In the Further Reading section you can watch the CEO of Figma explain his hypothesis on this.
The two design areas of work that have changed over the past 6 months have been image generation from a visual identity standpoint and design handoff to engineering. Below you can see an example of how Midjourney’s personalization tools now allow brands to highly personalize their output using personalization ratings, styles, and moodboards. The design handoff is now being smoothed out between design and engineering with tools like Builder.io and TeleportHQ which generate code from a Figma artboard for engineering teams. Bolt.new also just recently released (between the day I presented and time of writing this) a Figma integration to more accurately generate code from a Figma design.
Engineering
After the previous two stages of the workflow are complete, designers (or product managers for that matter) if they wish can now give a much more complete artifact to engineers using code generation tools so the engineers can focus on more technical and complex tasks like automated testing, architecture, and the like. Designers like myself, have teams that are usually engineering constrained, so we are using tools like Cursor and Bolt to give working prototypes to engineering teams that exactly match their desired specifications, interactions and animations.
Better Results
When designing and creating products, it is extremely rare to nail the perfect execution on the first try. A large part of the design process is iteration. A single sign in page could have 20 blank canvases in Figma, and you could explore 20 new ideas for that single view of the app to land on the best one. In product, it might take 11 versions of a feature to truly get it to a place where users love it. So the more we can iterate both internally and with people using the product, the more likely we are to have better results. All of the tools in the workflow reduce friction on the path to better ideas, better execution, and more iteration.
Technology Demo
The live Demo that we ran during the Tech Forum presentation was focused on chat-centric code generation tools with Claude and Bolt. Roughly the middle of the spectrum above. We used Claude to code the front-end of a feature for a web application given a product requirement document and Figma exports. Then moved to integrate this output into an existing Bolt project.
Given my context, I’m not that technical on the engineering side so the place I started was the middle. If you are and engineer, thing like Windsurf and Cursor on the right side of the spectrum may be better for you. If all of these tools are new to you, I would recommend creating something small and simple with the simple chat interfaces of ChatGPT or Claude (or some other favorite).
For the short-term I don’t think these are replacements for existing tools on the far left or the far right. There is still a place for Framer/Webflow for example. Tools should allow you to have maximum creativity and uphold quality. So for me, Figma and Framer still have a very useful place in iteration that the current generative tools do not replace. Mainly because the type of iteration I can do in generative AI tools is still very limited.
The Next Versions of Tools
For now there seem to be two major problems with AI tooling, 1 - the tools seem to be stuck in single player mode and 2 - context and personalization are both still highly limited.
If we look at the last generation of workplace tools - Slack, Dropbox, Figma they were highly collaborative. The current AI tools are not. We each seem isolated in our own AI experiences. I do think this is preferred for deep thought. You want silence and focus for that. But I am excited to see how some of the knowledge, images, and thoughts created in chat or code generation with AI tools can be more seamlessly integrated with the rest of my team’s work.
We are now (March 2025) only beginning to see early personalization settings and integration across tools. Claude just released an update to allow Github connections. And ChatGPT still only has minimal integration with the other tools you use for your work. The next version of tools with have deep orchestration of the suite of tools that companies use, each sharing context among them.
The future versions of these tools will also reduce the fragmentation that currently exists with collaboration and allow far more customization to fit each user and company.
AI tooling is still very poor at orchestration among tools, orchestrating actions, and taking in many various types of sources that exist within a company already (existing knowledge base). Modified from a visual that Mainframe created to describe their AI agent orchestration
The Path Ahead
It gets easier to build (more of the same)
The new generation of tools focused on AI will allow us to create companies and products more seamlessly and allow us to do it with less resources. If we think about the trajectory we have been on over the past 100 years, this is actually part of a larger trend of tools that allow us to build and innovate more fluidly. And by allowing people to create products more efficiently, companies have become better and better over time.
Paul Graham, Co-founder of Y Combinator (slightly compressed and modified)
Every decade it gets easier to start a startup. The main reason it’s easier to start a startup now is that it’s cheaper. Technology has driven down the cost of both building products and acquiring customers.
The decreasing cost of starting a startup increases the number of people that start them. Those that start them can also raise money on better terms.
There’s a third factor at work: the companies themselves are more valuable, because newly founded companies grow faster than they used to. Technology hasn’t just made it cheaper to build, but faster too.
This trend has been running for a long time. IBM, founded in 1896, took 45 years to reach a billion (2020) dollars in revenue. HP, founded in 1939, took 25 years. Microsoft founded in 1975, took 13 years. In 2021 the fastest growing startups, the norm for fast-growing companies was 7 or 8 years. [Now, Cursor may achieve this in 2 or 3 years. - my addition]
Though Paul is talking about Revenue here. He is making a more holistic point as well. That over time it becomes easier and easier to create new products that power a company's success. New tools allow us to create products more efficiently, which allows us to create more effective companies.
More creators. Quality bar gets raised.
Along with tools that allow us to be more creative and efficient, more people will have the ability to do so by lowering the barrier to entry in tooling.
In the Further Reading below, the Figma team—which ten or so years ago allowed even more people to become designers with more simple tools—predicts that the floor to get into design or engineering will be lowered even further. We will be able to create using natural language, and more natural inputs. At the same time, because the tools allow us to work more efficiently, the quality ceiling will continue to get raised. Meaning the best teams, engineerings, and designers can now create even higher quality work. Watch the Figma’s team talk on the future of design to learn more.
Modified visual from the Figma design team
Closing thought and Further Reading
So let’s bring it back to where the essay started. All of these AI tools will allow us to create better products. And the tools to create and build will come and go. Just as they always have. But it will continue become easier to move from idea to manifestation as the tools evolve. As just one example of many over the recent months, a great designer by the name of Sam used Vercel’s v0 to create a web app (see below) from a series of Figma exports without any technical background. The only minimal coding that was needed was about 12 hours of work up from an iOS engineer over the span of 72 hours to get it approved in the iOS App store properly.
Further reading/watching
Mainframe and the new approach to computing
Figma design team on the future of Design with AI
Y Combinator CEO interviews Figma’s CEO on Future of Design
About the RC Tech Forum
The RC Tech Forum: Building a Community of Innovators
The RC Tech Forum was founded with a vision to create a dynamic and inclusive space for technology enthusiasts, innovators, entrepreneurs, and students. Based in the Inland Empire, a region brimming with untapped potential, the forum serves as a centralized platform where individuals can come together, gain valuable experience presenting their ideas, and learn from seasoned speakers. Our mission is to amplify technology growth and foster entrepreneurship within the Inland Empire and beyond.
With over 400 members and growing, the RC Tech Forum is the premier hub for tech-minded individuals from across Southern California. Our community thrives on the hunger for knowledge, collaboration, and innovation. Each month, we host engaging meetups that feature diverse topics ranging from blockchain and artificial intelligence to business strategies for startups and beyond. By adapting our format regularly, we ensure our events remain fresh, relevant, and tailored to the interests of our members.
At RC Tech Forum, we value feedback from our community to guide the topics we explore. Whether you're delving into highly technical subjects or engaging with non-technical content, our events promise something valuable for everyone. Together, we aim to elevate education, spark meaningful discussions, and tackle challenges that drive the tech ecosystem forward.
