Joined Qamaq, an enterprise AI platform, to redesign it so it works for every skill level.
The platform assumed technical proficiency: dashboard overload, and a creation flow too complex for beginners.
Adaptive, skill-level-based dashboards and a dual-path "build a solution" flow, guided or blank canvas.
69% of beginners who couldn't complete an agent without help can now do it on their own.
Overview
One platform, every skill level.
Qamaq is an enterprise AI platform that lets teams consult internal documents, automate workflows, and build custom solutions all-in-one.
I joined Qamaq as an AI Product Designer to redesign a platform that had been built by the CEO and the dev team but was only really working for technical users. The challenge was clear: make it adaptive for every type of user, regardless of their skill level.
Working directly with the CEO, I led the full redesign of the platform's core flows. This included designing the new level-based dashboard and completely rethinking the "build a solution" experience.
The problem
Built for technical users, by default.
The platform was solid, but it was built assuming everyone already knew how to use enterprise AI. Two things kept coming up.
The goal
Keep the power. Lose the overload.
The goal was to keep Qamaq's full enterprise power while making it work for non-technical users. Two things had to happen.
User personas
Three skill levels, three sets of needs.
To kick off, I started by mapping users into three skill levels and used them to drive every design decision from that point on: Beginner, Intermediate, and Advanced, each with distinct needs, expectations, and comfort levels with enterprise AI tools.
Sarah Lee the beginner
Lifestyle
- Focuses on business efficiency, not technology.
- Relies heavily on standard SaaS tools like Slack, Workday, Notion.
Goals
- Automate simple, repetitive tasks without asking anyone for help.
- Rely on guided experiences and pre-built templates for guaranteed success.
Pain points
- Paralyzed by an open "blank canvas," like a standard ChatGPT prompt box.
- Intimidated by technical jargon and complex dashboards.
Mark Tyler the intermediate
Lifestyle
- Tech-savvy, though not a formally trained software engineer.
- Familiar with low-code automation tools, like n8n.
Goals
- Have enough freedom to build and modify workflows without writing a single line of code.
- Quick access to what's already running, picking up where he left off instead of starting from scratch.
Pain points
- Standard beginner templates are way too rigid and limiting for his needs.
- Full developer environments are too complex and require too much technical knowledge.
James Miles the advanced
Lifestyle
- Deep technical expertise and understanding of AI behaviors.
- Comfortable writing code and building complex data systems.
Goals
- Build complex, custom AI agents and multi-step workflows from scratch.
- See the health of all his solutions and spot issues the moment he logs in.
Pain points
- Overly simplified interfaces preventing him from doing his job.
- No visibility into how his automations are actually performing.
Competitive audit
Don't reinvent the pattern people already know.
To map all necessary components to include in the dashboard, I ran a competitive audit, identifying core features and user flows across top competitors.
The best dashboards build on patterns people already recognize, not ask them to learn new ones. So instead of designing from scratch, I leaned into established AI UI patterns: initial CTA, example gallery, AI suggestions. The result was a dashboard that felt intuitive from the first click.
Dashboard design
A dashboard that adapts to who's looking at it.
Dashboard for beginners
Guided CTA with AI suggestions. Shows users what they can do before they even start typing. No blank box, no confusion.
Clear section headlines. "Start building from scratch" and "Or use an existing solution" tell users exactly what their options are without making them think about it.
Prominent main actions. The core creation paths are big, colorful, and impossible to miss.
Example gallery. Instead of an empty platform, users land on a page that already shows what's possible, passive onboarding before they've done anything.
Key decision. The founder resisted simplifying the interface, worried it would strip the platform's power. I argued that simplicity doesn't mean less capable, it means less overwhelming. The adaptive approach was the compromise: full power for advanced users, a guided entry point for everyone else.
Dashboard for intermediate users
Active solutions up top. They already have things running, so that's the first thing they see.
To Dos front and center. AI generated and manual tasks sit right there on the dashboard.
Recent documents. The files they've already been working with surface automatically.
Simplified actions. The big CTAs shrink to compact buttons because they already know what they want to do.
Smarter suggestions inside the chat. Personalized based on their actual history.
Dashboard for advanced users
Performance metrics up top. Agents active, success rate, time saved, open issues.
Solutions as a table. No visual cards, just status indicators, success rates, and flagged issues they can scan in seconds.
Build a solution flow
Guided for beginners, blank canvas for experts.
The next challenge was what happened when users tried to build something. I redesigned a flow that actually guided beginners through the process while letting advanced users skip straight to what they needed. Intermediate users could pick either path: the guided tutorial if they wanted the extra support, or the blank canvas if they already knew what they were doing.
Path A: Beginners
Guided tutorial. Instead of a blank canvas, the platform walks them through it, asking simple planning questions and generating a visual plan before it builds anything.
Path B: Advanced users
Blank canvas. No hand-holding, no extra steps. They land directly in the builder with full control from the start.
Guided tutorial for beginners
Progress indicator. Beginners can see exactly where they are in the process. It feels achievable instead of overwhelming.
Step-by-step instructions. Shows exactly what to do at each phase, so the user always knows what to expect next.
Guiding microcopy. The chat tells users exactly what kind of input is expected at each step.
Solution builder redesign
Cleaner navigation. Removed extra settings and version controls from the top bar.
Prominent CTA. Replaced the easy-to-miss "Spec Document" link with a high contrast button.
Consolidated sidebar. Merged the middle column into the sidebar, adapting to the current phase.
Phase dividers in the chat. A visual marker appears when a step is done, keeping the conversation easy to follow.
Adaptive guidance. The AI adjusts how it talks based on who's using it: step by step for beginners, straight to the point for advanced users.
Results
From zero to deployed, without support.
Previously, non-technical users couldn't build an agent without support. Now, a first-time user can go from zero to deployed, on their own.
By combining established AI UX patterns with adaptive guidance and clear microcopy, we removed the technical barrier that made the original platform inaccessible to most of its intended audience.
69%
Of beginners who couldn't complete an agent without help can now do it on their own.
* Data from user testing conducted after the redesign.
Key takeaways
What this taught me.
Beginners shouldn't have to learn how to talk to an AI
Instead of leaving them with a blank box, we added suggestions, examples, and microcopy to point them in the right direction.
The UI should adapt
Beginners get guided steps, fewer options, and simpler AI responses, while experts get direct access and more technical feedback.
