Discord
AI Server Creation Feature

Building smarter servers without disrupting what users already love

Discord is used by over 100 million people to communicate, game, and build communities. While creating a small friend group server is relatively simple, setting up a large, well-organized community server is time-consuming, overwhelming, and often incomplete without prior planning. I designed two interconnected AI features — an AI server generator and an AI bot fetcher — that integrate natively into Discord's existing interface without breaking its familiar feel.

The core challenge was not just solving a real user problem, but doing so in a way that felt like it belonged in Discord all along.

Server creation is powerful but painfully slow

Communities are at the heart of what sets Discord apart from competitors like TeamSpeak and Mumble. But the process of building one from scratch is where the platform falls short. Without prior planning, most servers are left in a barebone state, lacking the channels, bots, and structure that make a community actually function.

From user frustration to a feature that feels built-in

This project followed a full UX design process, moving from competitive analysis and user interviews through multiple wireframe stages and into a tested, iterated hi-fi prototype.

Understanding where Discord's competitors fall short

Before interviewing users, I conducted a SWOT analysis of Discord's three primary competitors — TeamSpeak, Mumble, and Revolt — to identify the gaps Discord could exploit with an AI feature, and the strengths it needed to protect.

Talking to the people who build and live in Discord servers

I conducted 5 user interviews with Discord users ranging from casual chatters to experienced server admins. Sessions ran approximately 12 to 15 minutes each and were recorded with permission via Otter.ai. Participants included Brian, Edgar, Gustavo, Dannel, and Angel — each with varying levels of experience in server creation and AI tools.

• 1
  Most servers are small and informal The majority of users who have created servers did so for close friends or family, not large communities. Those who found setup simple had very limited scope. Complexity scales sharply with ambition.
• 2
  Bots elevate the experience but discovery is broken Users know bots exist and value them, but most are unfamiliar with how to actually find and add them. Bot functions in larger servers can be unclear, and the process of evaluating one bot over another lacks guidance.
• 3
  AI is welcomed, with some caution Most users envision AI making server creation and bot management significantly easier. A small number expressed uncertainty about AI's role, which was addressed by framing it around specific, practical use cases during the design process.
• 4
  Familiarity drives loyalty Users stay on Discord because it's familiar and feature-rich. Any new feature needs to feel like it belongs natively to the platform, not like an external tool bolted on.
• 5
  Admin complexity is a genuine pain point Managing permissions, roles, and channels for a large server is overwhelming. Users without prior admin experience often don't know where to start, and Discord offers no guided assistance.

Four insight clusters from affinity mapping

After interviews, I organized observations into an affinity map, grouping notes across all 5 participants into four meaningful themes that shaped the direction of the feature design.

Two users, two distinct relationships with server creation

Research produced two primary personas — one representing an aspiring community builder who is overwhelmed by the process, and one representing a social user who wants deeper customization without the learning curve.

Sketching natively before opening Figma

Lo-fi wireframes were hand-drawn to map out the two core interaction surfaces: the "Generate with AI" entry point in the server creation modal, and the "Fetch Bot" feature accessible from the server dropdown menu. Keeping sketches rough deliberately prevented premature attachment to visual details.

• ~
  Entry point placement needed validation The initial lo-fi placed the AI generation button outside the main content card as a separate element. User feedback during early review suggested it felt disconnected from the rest of the interface.
• ~
  Dual input methods planned from the start Lo-fi sketches explored both keyword selection and free-text input for server description, giving users flexibility regardless of how they prefer to articulate their ideas.
• ✓
  Bot management needed a dedicated home A new "Bots" tab within Server Settings was identified in lo-fi as the right location for managing added bots, extending an already-familiar settings architecture rather than creating a new one.

Two parallel AI-powered paths through Discord

The user flow mapped two distinct task paths that could be completed independently or sequentially. The first covers AI-assisted server creation from the home screen through to a fully generated server. The second covers bot fetching from within an existing server's settings.

Both paths were designed to be short and recoverable, with clear back navigation at every step so users never felt trapped mid-flow.

Grounding the feature inside Discord's actual UI

Mid-fidelity wireframes translated lo-fi concepts into Discord's real interface layout, using the platform's dark theme, typography, and component patterns. This stage was critical for validating that the new features felt native rather than foreign.

Eight screens were produced at mid-fi, covering the server creation modal, the AI description step with keyword selection, the server confirmation state, the fetch bot flow, and the bot results overlay. Discord's existing modal and dropdown patterns were preserved throughout.

A feature that looks and feels like it shipped with Discord

The hi-fi prototype fully adopted Discord's dark UI, typography, iconography, and component patterns. The goal was for a user to encounter "Generate with AI" and "Fetch Bots" and feel that they had always been there — not that a designer had inserted something new.

The final prototype covered 15 screens across two complete user flows, including hover states, active keyword states, server confirmation, bot results with descriptions and ratings, and the bot-added confirmation. Both flows were fully interactive and prototype-linked for usability testing.

Three meaningful changes driven by user feedback

Usability testing surfaced three clear areas for improvement. Each iteration was grounded in direct user observations rather than designer preference, and each brought the feature closer to feeling intuitive and complete.

Both features validated with strong completion rates

Usability testing was conducted with 6 participants across two tasks — creating a server using the AI generation feature, and fetching a music bot to add to an existing server. Testing was moderated and sessions were recorded for review.

• ✓
  The design felt native to Discord Users described the features as feeling "familiar" and said nothing felt out of place. Maintaining Discord's existing visual language was the single most important design decision in building that confidence.
• ✓
  Dual input method was well received Users appreciated having the option to either type freely or select from keywords when describing their server. Each participant naturally gravitated toward one method without hesitation.
• ✓
  AI was universally welcomed in this context Every participant who was initially cautious about AI in Discord became more open once the specific use case, solving a real setup frustration, was clear. Framing mattered.
• ~
  Multiple navigation paths created minor inconsistency Each user had a preferred way of accessing settings, some via right-click on the server icon, some via the dropdown menu. Both paths were supported, but future iterations could add context-aware shortcuts.

What it means to design for a product you didn't build

The biggest lesson from this project was how strongly users resist change to something they've used for years. Moving the "Generate with AI" button from outside the content card to inside it wasn't a dramatic change visually, but it made all the difference in how natural the feature felt to users. That's the kind of micro-decision that you can only arrive at through testing — not intuition.

Finding a way to make AI useful on a platform that is already feature-rich was the central creative challenge. The key was framing AI not as a new layer on top of Discord, but as a time-saving assistant operating within the existing flow. It doesn't replace the user's decisions — it just makes the setup faster and smarter.

I also encountered a small number of users who were cautious about AI implementation. Rather than avoiding that concern, I addressed it directly by designing for a specific, low-stakes use case and letting the prototype speak for itself during testing. By the end, every participant had accepted the feature as genuinely useful.

Where the feature goes from here

This project established a solid foundation for AI-assisted server creation and bot discovery. Several enhancements were identified during testing that fell outside the current scope but represent a clear roadmap for a next iteration.

How this project was built