Real Benefits of AI
Preliminary Research
I incorporate AI into my early research process to automatically cast a wide net in terms of exploring what is out there and setting a direction for a project.
Below is an example of an N8N workflow that sources a task from Google Sheets, filters out keywords, and runs automatic research on them while writing the findings back into the sheet. This way, I can get some early research done automatically just by documenting upcoming tasks into Sheets.
Human In The Loop
Since LLMs are prone to hallucinations and stating untrue statements with high confidence, it is crucial to use the AI research as a starting point and double-check the output. Never rely on it completely or use it for emulating user responses.
Light-Speed Ideation
AI writes code and sets up design systems faster than any human can. This can be leveraged heavily in the exploration and ideation phases.
Instead of manual, time-consuming look-dev, we can move through a large volume of ideas quickly and discard them as needed, without attachment bias.
Systemic Prompting
This approach will burn through AI credits like no tomorrow. A systemic prompting method, with controlled keyword variations targeting design or structure is essential for strong price-to-performance results.
Component Granularity
Perhaps the biggest unlock for day-to-day design workflows is the ability to target specific flows or components in Figma and infuse them with AI through an MCP server in Cursor, allowing for fast iteration at any point in the process.
Instant Enterprise Value
When building cross-platform SaaS products, AI is often ineffective because of limited contextual understanding and an inability to interconnect and build on complex concepts.
With this method, we eliminate the need for LLMs to understand the broader landscape and instead focus their power on specific chunks.
Extending Horizons
Using AI agents to help write code marks a monumental shift in extending the reach of a talented designer.
We already have the underlying principles of structure and aesthetics established, which enables us to explore conditioning logic, backend solutions, and specific agents and channel them into polished and intuitive products.
Digital products are solved in pixels, and now a high-performing designer can go further to deliver the final feel of a shipped product early in the process.
Sobering Limitations
Usage Limit Hell
Sooner or later, we run into usage limits for prompting. Some companies are even reducing available quotas in real time to maximize profit margins.
The crucial thing is to gain as much experience as possible now, while generous free tiers are still available. Another hack is Figma Make, which does not yet enforce limits, making it an excellent learning resource.
Mid Output
AI will not turn mediocre designers into top performers. The initial output is often painfully mid, comparable to a junior or intern-level designer, and requires a large amount of re-prompting to reach high-quality results.
Iterate & Refine
AI agents often stumble on accessibility, content layering, spacing, hierarchy, and consistency, issues that to humans appear obvious at a glance.
Take the Initial output as a starting point and systemically work on elevating the design level step by step. This is where deep craft and taste of a designer take centre stage.
The Workflows
Full AI Vibe Ride
Undoubtedly, the industry benefiting most from the AI boom is the entrepreneurial sector.
Taking digital products from idea to shipping has never been easier. The obvious question is how we ensure that actual value is delivered to users and the correct products are built. That is the part AI cannot help with.
Solid Basics are Key
In these circumstances, where AI brings you to a good-enough starting point, solid principles of user-centric design are crucial.
To avoid producing AI slop nobody actually needs, it is more important than ever to understand your user base and market fit.
AI can be leveraged here by going to market fast, testing MVPs early and pivoting on the spot as needed, with the flexibility of quick re-builds.
Control Workflow
This flow focuses on delivering instant value to targeted bottlenecks in the digital product design process.
Sub-loops of this workflow can be deployed in enterprise pipelines today to facilitate innovation, requirement shaping, design system prototyping, early feature flow exploration, and internal MVPs.
Innovation Phase
Vibe coders perform much better with visual references. A major advantage of using AI in the initial stages is the ability to quickly achieve a high-quality visual look and feel of a product through Midjourney.
It is also easier than ever to explore alternative directions and gather inspiration before committing to a specific path.
Feature Prototypes & MVPs
Digital products breathe in pixels, which is why I place high value on high-fidelity screen design.
With vibe-coded prototype flows or components, we can quickly achieve the real product feel and gather genuine feedback from live interactions.
Design Systems
Using AI within an existing design system can be tricky. The key is having impeccable structure in auto layouts, variables, components, and spacing to provide the LLM with a solid reference.
Once the basics are established, AI can be leveraged for rapid iteration of components based on specific parameters.
Final Results
Presentation Website
A fully functional, one-page presentation designed to catch the curiosity and set the stage in terms of characters, story, and basic mechanics.
Character Generator
The best way to get user buy-in is by offering a hands-on experience, which is exactly what this fantasy character generator provides. It is empowered with AI to deliver real-time visuals and descriptions.
Developed Game Idea
This project explored the conceptualization of a tabletop game and assessing the limitations of AI in terms of fully sjipping an idea from scratch.
Deliverables included an initial sketch of the concept and basic game systems, the development of characters and the storyline behind them, and the exploration of world-building.
Part of the concept also examined the conversion of digital visuals into real-world, physical tabletop elements.
