# Designing Mantium: Enterprise AI Chat UX (Case Study)

By David Skelton - 2026-08-18 - https://skeltonmedia.com/blog/case-study-mantium-secure-rag-chat.html

**Mantium was a generative AI platform that let enterprises build secure, AI-powered chat
applications on their own data, with no code required. Signing up took five minutes and eight clicks.
A new enterprise customer's first real deployment averaged over two hours. Getting that under
thirty minutes was the work.**

Designed for teams with strict compliance and security requirements, the platform used
Retrieval-Augmented Generation to deliver intelligent, context-aware responses. As the product
designer, I led end-to-end design across feature planning, prototyping, research, and dev
collaboration, delivering a human-first interface in a fast-paced, AI-driven environment.

## The challenge

Enterprises wanted to deploy custom AI chat systems using their internal data, but most solutions
lacked security controls, lacked intuitive design, or required engineering resources to build.
Mantium's challenge was to deliver enterprise-grade functionality without complexity. A design
that could bridge the technical sophistication of RAG with an approachable no-code UI.

The gap showed up in the numbers. Getting *in* was easy: under five minutes, eight clicks. Getting
something *useful* out was not. What people were actually doing was transforming their enterprise
data and creating a knowledge base for their company to use internally, all without technical
knowledge. Four things had to happen before anyone else could see a result: connecting the data
sources, transforming and preparing the data for deployment, assigning permissions at the row
level, and inviting other people in to view and test the application. For a first-time enterprise
customer that averaged over two hours, and ran longer with more data, since the transformation
itself cost tokens and time.

## Research and conceptualization

We conducted in-depth research targeting enterprise customers seeking secure, AI-enhanced
communication platforms. Competitive analysis of leading platforms, including Anthropic, Google, and
OpenAI, informed our design strategy, and shaped the roadmap around user trust, data privacy,
and ease of use.

## My role and approach

**End-to-end product design.** Led all stages of the product lifecycle, from gathering
requirements to shipping high-fidelity designs. Created user flows, wireframes, and clickable
prototypes tailored for enterprise-level use cases.

**AI research and strategy.** Deep research into enterprise AI adoption needs and expectations.
Benchmarked against Anthropic, OpenAI, and Google to define differentiators.

**Rapid iteration and startup agility.** Adapted designs weekly in response to fast-changing
technical specs and user feedback. Kept close communication with leadership and engineering to
balance feasibility with UX quality.

**Developer collaboration.** Used GitLab, Loom, and Slack for async design reviews. Shared Figma
prototypes and specs that supported tight weekly release cycles across a growing international
team. 

**Human-centric interaction design.** Focused on natural, intuitive interactions so a technically
complex product felt lightweight. Designed conversational interfaces that emphasized clarity,
context, and user trust.

## What changed

- Average time to first deployment, new enterprise customers: **2+ hours → under 30 minutes.**
- Signup stayed what it always was: **under five minutes, eight clicks.**
- **Three major releases** shipped, through product-market-fit exercises, rapid prototyping, and
  user research.
- Production-ready designs delivered for the critical workflows: chat setup, knowledge ingestion,
  and prompt configuration.
- Mantium positioned as a secure, no-code alternative to the major AI platforms.

Separately, and for our own team: I built a centralized AI knowledge base in Notion integrated
with Slack: a living documentation system across processes, HR, and product updates. It cut
new-hire onboarding time by roughly 40% and made async-first collaboration workable. That became
the template for how I think about AI knowledge systems today.

## What this project proves

Building secure, enterprise AI tools doesn't have to mean complex UX. With clear communication,
smart research, and rapid prototyping, even the most technical products can be made accessible.
This reinforced how GenAI design workflows and cross-functional alignment can move fast without
breaking trust or usability, especially in mission-critical enterprise environments.

## Selected screens

![Mantium chat interface with templates](assets/case-studies/mantium-secure-chat/chat-templates.webp "Chat interface: templates for common enterprise tasks")

![File upload flow](assets/case-studies/mantium-secure-chat/file-upload.webp "Knowledge ingestion: file upload")

![Supporting content panel expanded](assets/case-studies/mantium-secure-chat/supporting-content.webp "Supporting content panel: sources behind every answer")

![App builder](assets/case-studies/mantium-secure-chat/app-builder.webp "App builder: the no-code deployment flow")

![Human in the loop agent design](assets/case-studies/mantium-secure-chat/human-loop.webp "Agents with a human in the loop")

![User management](assets/case-studies/mantium-secure-chat/user-management.webp "Admin: user management and row-level permissions")

![Mantium design system](assets/case-studies/mantium-secure-chat/brand.webp "The Mantium design system")