AI-Assisted Cover Design and Enterprise Resource Discovery
Exploring how AI could support creative authoring, enterprise resource discovery, research organization, and mobile app prototyping while keeping users in control.
This project brings together several AI-related explorations across professional and personal work.
The work was not about adding AI for its own sake. It was about looking at where AI might help users create, find, understand, or act on information, and where the product still needed strong UX judgment, clear constraints, readable outputs, and human review.
The explorations included a concept for helping authors create cover-art materials for book submissions, a chatbot concept for connecting resources across related enterprise tools, the use of ChatGPT to help organize user research notes, and a personal mobile app experiment built with AI-assisted coding tools.
I explored these concepts as a UX and product designer, with a focus on how AI would fit into real workflows.
- Exploring AI-supported creative workflows for authors
- Thinking through how authors might choose fonts, text styles, image direction, and cover backgrounds
- Considering readability, contrast, and the relationship between generated imagery and cover text
- Exploring where AI-generated patterns or images might support book-cover material creation
- Mapping where an AI chatbot might fit into an enterprise training and resource system
- Thinking through how a chatbot could provide links, summaries, and guidance across related tools
- Using ChatGPT to help organize user research notes and feedback categories
- Reviewing AI output critically for accuracy, consistency, and fabricated details
- Experimenting with AI-assisted coding tools for a personal mobile app concept
- Evaluating where these tools helped and where they still required human product judgment
The central challenge was deciding where AI was useful, not just where it was possible.
AI can generate images, patterns, text, summaries, links, code, and interface ideas quickly. But speed is not the same as product quality. The more important questions were what the user was trying to accomplish, where AI reduced friction, where it created risk, how the user would evaluate the output, what needed to stay editable, what needed review or confirmation, and what happened when the AI was wrong.
Across these explorations, the design problem was not simply to add AI. It was to understand how AI could support the workflow without taking away control, creating unreadable results, fabricating information, or making users trust something they should verify.
For one exploratory project, I looked at how AI could support authors creating cover-art materials for book submissions. The concept included ways authors might choose fonts, text styles, backgrounds, generated patterns, and AI-generated images. A key part of the exploration was readability. Cover text still had to work over the image, and the system needed to help users avoid choices that made titles, subtitles, or author names hard to read. The most interesting part was not just image generation. It was the product structure around image generation: choosing styles, combining type and image, evaluating legibility, and helping non-designers make better visual decisions.
In a separate enterprise project, I explored where a simple AI chatbot feature might fit into a larger training and resource workflow. The idea was to think through how a chatbot could help users find information, get links to relevant resources, and understand connections between the main project and related tools. This required thinking through what the chatbot should know, where it should appear, and how it might move users between different parts of a larger system. The value was not in making the chatbot feel impressive. The value would come from helping users find the right resource faster, understand what tool or section they needed, and get clear links or guidance without searching through multiple systems.
I also used ChatGPT to help organize research notes and user feedback during an enterprise project. I did not treat the output as automatically trustworthy. Because I had taken the notes myself and understood the project context, I could review the AI's organization, correct fabricated or inaccurate details, and make sure the synthesis matched what users had really said. Used carefully, it helped group related feedback, identify themes, and collate user needs more quickly. The useful part was not replacing judgment. It was speeding up organization while keeping a human review layer in place.
I have also been experimenting with AI-assisted coding tools for a personal mobile app project. I started with Amazon Kiro while preparing for an Amazon role interview, partly to understand the tool itself. Over a weekend, I was able to get a full app structure and navigation system working. The larger project is more ambitious and not fully functional yet, but the experiment gave me a clearer sense of where vibe coding can help and where it still falls short. The next step is to compare tools such as Claude Code and others against the specific needs of the project. I am interested in how far these tools can go for structure, navigation, interface behavior, and implementation support, while still relying on my own product direction and design judgment.
AI is more useful when it appears at a point where the user has a real need: generating options, organizing information, finding resources, or getting unstuck.
For creative tools, users need to adjust style, text, image, layout, and output quality. For enterprise tools, users need to verify sources, follow links, and understand why a result is being shown.
AI output needs evaluation. For cover art, that might mean readability and composition. For enterprise tools, that might mean source links, permissions, confidence, and traceability.
In research and enterprise contexts, fabricated or unsupported information can cause real problems. The interface needs to make uncertainty visible and avoid presenting guesses as facts.
The more open-ended the AI feature is, the more likely the user may get interesting but unusable results. Structure, constraints, and clear interaction design help make AI more useful.
These explorations helped me think more clearly about AI as a product material rather than a magic feature.
For creative workflows, the strongest opportunity was not just generating images. It was helping users make better design decisions around type, readability, style, and output quality.
For enterprise workflows, the strongest opportunity was not a general chatbot. It was a focused assistant that could connect users to known resources, explain relationships between tools, and provide clearer paths through complex information.
For research work, AI was useful for organization and theme-finding, but only with careful human review from someone who understood the project. For personal prototyping, AI-assisted coding showed real promise for getting structure and navigation in place quickly, while still requiring product judgment, design direction, and technical follow-through.
These explorations reinforced that AI still needs design. Users need to know what they can change, what they can trust, what needs review, and what action to take next. The interface around the AI matters as much as the AI output itself.
They also reinforced that AI tools are most useful when paired with someone who understands the domain, the users, and the risks. I do not fully trust AI output by default, but I do think it can be useful when I can review, correct, structure, and apply it to a real product problem.