Quickbase AI Console
Integration of generative AI into the platform
At Quickbase, I led a project to integrate generative AI into existing product features, working closely with a talented designer on my team. Together, we focused on making it easier and faster for users to accomplish their goals from a centralised place.
Collaborated with
+ 4 PMs and 4 Engineering teams
Q2'24
The Backstory
The idea for the AI Console started with a simple question: how do we make AI accessible across Quickbase without disrupting workflows? Initially, I explored ways to centralize all AI features in a persistent container—something that would always be available when users needed it. My first approach was a Floating Action Button (FAB) that would allow users to quickly summon AI tools no matter where they were in the product.
At the time, the new navigation had not yet been released, which is why the FAB seemed like the best solution—it allowed us to move forward with AI integration without being dependent on the navigation release cycles. However, as the navigation launched, it became clear that introducing a FAB as a core interaction model would have clashed with the new structure.
From a product management perspective, integrating AI directly into the new navigation made more sense. It reinforced the structure we had just introduced, ensured consistency across the product, and helped users adopt the new navigation naturally. This shift also meant AI features felt like a core part of the product rather than an overlay.
Another key challenge was that different AI features were being developed by different scrum teams, each on their own release schedule. This meant we needed a solution that wasn’t just central but also flexible enough to accommodate features at different stages of development. Instead of locking AI into one rigid structure, we took an abstracted approach, designing the AI Console in a way that allowed teams to plug in their features at their own tempo without disrupting the overall experience.
The Problem
Users needed a straightforward way to make sense of their data and streamline their tasks. But the existing workflows were messy—tools were scattered, making them hard to find and even harder to use. For new users, this was overwhelming; for experienced users, it felt limiting.
User Pain Points
Disjointed AI Features: Users had to jump between different areas to get things done, slowing them down.
Limited Control: Power users wanted more flexibility but found the system too rigid.
Steep Learning Curve: The AI tools weren’t clear enough for new users to understand quickly.
The Solution
The AI Console became a single place where users could interact with AI tools without the confusion. We designed it to be simple, approachable, and effective, ensuring it worked for all users, regardless of experience level.
Key Features & Design Decisions
1. Straightforward Navigation
A modular layout puts the most important tools front and center.
Users can quickly access all skills:
"Ask Quickbase AI" for natural language queries.
"Build a pipeline" to automate workflows.
2. Clear User Flows
Instead of vague labels, we used action-based language: “Build a pipeline” and “Get Quick Insights”.
The design guides users step by step, cutting out confusion.
3. Focus on Usability
We kept the layout clean and easy to scan.
Everything adapts well to different screen sizes.
Small details, like icons and micro-interactions, make the experience feel smooth and predictable.
The Process
1. Research & Testing
We conducted user interviews with a wide variety of users to gather insights and ensure the AI Console met diverse needs.
We mapped out existing workflows to see where they got stuck.
We worked closely with product and engineering to define priorities based on user needs.
2. A Phased Approach: Crawl, Walk, Run
To give engineering time to build everything properly—without making users wait too long—we released features in stages:
Crawl Phase
We built the foundation: simple AI queries and clear UI patterns.
Users got an early version, and we collected feedback before adding complexity.
Walk Phase
We introduced more interactive features, like templates and automation.
Based on user input, we refined navigation and interaction to make tasks faster.
We started adding predictive insights to help users work more efficiently.
Run Phase
We expanded the AI tools to include advanced workflow automation and deeper insights.
Power users got more flexibility, without making things harder for beginners.
The final version was fully scalable and ready for future growth.
3. Iteration & Refinement
Michaela and I worked through multiple prototypes, testing them with real users.
We prioritized features based on user feedback, ensuring that we focused on what mattered most while deprioritizing others for later phases.
The final design was simple, practical, and built to solve real problems.
What We Learned
Positive Reception & Adoption
📊 100% of users responded positively to the AI Console, indicating strong approval.
✅ Likelihood of Use: 100%—Users are eager to integrate AI into their workflows.
Usability & Interaction
🎯 60% of users appreciated interactivity, reinforcing the importance of engaging AI tools.
🎨 50% preferred a clean interface, highlighting the need for clarity and simplicity.
Feature Requests & Enhancements
🤖 30% requested proactive prompts—AI should anticipate user needs.
📂 20% wanted advanced filtering, suggesting a demand for refined data control.
Concerns & Challenges
📦 25% worried about clutter, emphasizing the need to balance feature richness with simplicity.
⚡ 20% flagged power-user limitations, signaling a need for more customization.
The Outcome
The AI Console changed how users interact with AI inside Quickbase:
One place for everything—no more bouncing between different tools.
Easy to use, quick to learn—better navigation and clear labels made AI more approachable.
Works for everyone—whether you’re new or experienced, the console adapts to your needs.
Built to last—the phased approach made sure everything was solid, scalable, and actually useful.
Empower '24 Reception
At Empower '24, Quickbase’s annual user conference, we introduced the AI Console to a packed breakout session with thousands of attendees. The response was overwhelming—users were excited and eager to try out the new features. This strong reaction confirmed that we were solving real problems and giving users tools they actually wanted.
Cross-Functional Collaboration
Designing the AI Console wasn’t just about creating an intuitive interface—it required orchestrating efforts across multiple product teams, engineering groups, and leadership stakeholders to ensure AI capabilities were built in a cohesive, scalable way.
To tackle this complexity, we maintained ongoing alignment through multiple collaboration channels:
Leadership & Strategy Alignment – We held weekly meetings with all leads and the CTO, ensuring that product, engineering, and design leadership were aligned on the vision, progress, and roadblocks. Our Director of Design was also involved, keeping the bigger design strategy connected across teams.
PM Alignment Across Teams – Since multiple PMs owned different AI features, we had weekly or biweekly syncs to ensure the AI Console’s structure could support varied product roadmaps, priorities, and release schedules.
Engineering Design Syncs – Together with Michaela, we led regular design critiques with engineers, giving them space to tear apart our designs, ask questions, and surface feasibility concerns before development kicked off.
Scalable, Abstracted Architecture – Because AI features were being built by separate scrum teams, we structured the console as a modular, adaptable system rather than a rigid, one-size-fits-all solution.
This multi-level collaboration approach ensured that AI Console remained strategically aligned, technically feasible, and adaptable for future AI advancements at Quickbase.
Final Thoughts
This project was complex—balancing multiple scrum teams, aligning with PMs, and ensuring a great design experience, all while keeping the AI Console intuitive and user-friendly. We had to make compromises, adjust plans, and rethink certain approaches along the way. But despite these challenges, we created something that works seamlessly, adapts to different needs, and makes AI a natural part of the Quickbase experience. It wasn’t just about building AI features—it was about making sure they fit into the product in a way that felt right for users. Thanks for sticking around 'till the end!