Discy.ai
AI-Powered Qualitative Research Platform
Discy AI is an insight platform that turns interviews, notes, and documents into defendable, cited insights.
As the Product Designer at Discy AI, I led the redesign of its flagship Insight Studio. My focus was to shift it from a manual tagging tool into a proactive, AI-driven discovery space where suggested insights, structured narratives, and rich data-visualizations sit at the core.
Turning research into a credible, shareable story is exhausting.
The old Insight Studio produced one static report fed by scattered inputs and rigid views. AI stopped at classification, so teams still had to re-author content, copy-paste between tools, and chase drifting versions.
This created slow cycles, inconsistent outputs, and a competitive gap — which prompted the founder to push for a step-change beyond classification toward credible, adaptable insight delivery.
AI-assisted canvas replaces static reporting
Canvas Workspace
- Centralise: one place for a project with multiple reports/canvases.
- Multi-tab & compare: switch or compare views/canvases.
- Templates / Resume: start from canvas templates or previous projects.
NEW INSIGHT CARD , THE BUILDING BLOCK
- Build with cards: drop modular cards onto the canvas.
- Rich forms: pick forms from a gallery (Framework / Narrative / Graph) that inherits the card’s data/filters.
AI NARRATIVE CARD
- Generate inline: create Summary and Bullet points directly on the canvas.
- Reuse & refine: apply prompt presets, one,click Regenerate by tone or focus.
User and market evidence
We ran a moderated, task-based in-product walkthrough with 7 experienced users to validate pain points and surface needs. In parallel, a competitive scan benchmarked entry model, multi-form views, AI controllability/traceability, and export templates.
We ran 30-minute in-product walkthroughs with 7 Discy users. Participants thought aloud as they tagged in Data Studio and assembled in Insight Studio (Chart, Table, Ask AI, Saved), describing where they got stuck or what felt slow or unclear. We ended each session with a “magic-wand” question: what they’d want if anything were possible to surface needs and trust blockers.
We benchmarked 7 current competitors in the qualitative,research software space. We focused on capabilities that drive speed and trust. The scan showed a gap: ops tools (Asana/Linear) and evidence tools (Dovetail/Atlas.ti) don’t produce customizable, cited reports in one place; Notion/copy.ai offer flexible docs and rewrites but lack provenance, confirming white space for an AI,assisted canvas with reproducible, shareable outputs.
Walkthroughs show the story splinters across tabs: filters aren’t preserved, Ask AI is one,shot and uncited, side,by,side report comparison is impossible, and people export to slides to finish.
The competitive scan confirms a gap: no tool combines a canvas workspace, in,place view switching, steerable + cited AI, and reproducible saved views.
So we asked:
How might we centralize analysis on an AI,assisted canvas that adapts to audiences?
Sketching and Shaping
From the HMW and walkthrough findings, we explored three seeds: blank canvas, a templated dashboard, and an AI-enabled narrative card. Afterwards, we refined them through two rounds.
We started exploration from the HMW and walkthrough findings, anchoring on three early ideas.
We sketched multiple variations of these three screens.
Design walkthroughs
We ran mid-fi wireframe walkthroughs with 6 internal stakeholders. Sessions captured think-aloud reactions and we iterated again to close the loop.
We ran 30-minute moderated walkthroughs with 7 internal stakeholders. Participants thought aloud while creating a canvas, browsing templates, adding graph/narrative cards, and trying A. Overall, canvas entry and templates landed well, while first-step clarity, filter and AI placement controls between global and single cards, and unclear icons need further improvement.
1. Clearer Canvas Entry
- Add templates next to + New Insight, making the first step obvious and cut time to action.
- Clear template categories, My, Discy, and context,recommended, that helps people pick the right pattern instantly.
2. Blank goes straight to canvas
- Choosing Blank Insight Board now opens an empty board immediately, not another chooser.
- A lightweight New Insight Card picker appears on-canvas only to help place the first block, and can be closed anytime.
3. Icons that explain themselves
- Hover tooltips now appear above icons, with clearer status indication.
4. Faster narrative: prompt history + A/B drafts
- Recently used prompts surface inline for one-tap reuse and consistent tone.
- New questions return two draft options side-by-side, review, edit, and choose.
- Regenerations keep filters and citations, then insert the selected draft to the canvas.
Video demo to pitch the AI-assisted Insight Studio
After refinement, I delivered a detailed Figma prototype to developers and created a 90-second video demo introducing Discy’s new Insight Studio and key features. The demo was used for stakeholder pitch and new-customer intro, while the prototype unblocks build planning and implementation.
Results in 4 weeks
Insight Studio was designed, prototyped in Figma, and developed into a working release. After the new features shipped, usage increased and teams completed reports faster. The video demo and prototype also helped unlock broader stakeholder buy-in and opened pilots with new customers.
- Won buy-in fast: the demo unlocked 3 new pilots and became the pitch opener.
- Real adoption: Weekly active users +34%.
My Key Learnings
- We had tight time and budget, so I shipped in small, fast loops and talked with the founder and devs a lot; that rhythm kept scope real and decisions moving.
- We couldn’t run a full study, but quick internal walkthroughs still paid off, they revealed blind spots I hadn’t noticed.
- A short video demo worked way better than slides; it helped people “get it” quickly and sped up buy-in and sales chats.
Open to next role Have a nice project?
Hi, my name is . You can answer me on this email . I am looking for help with a .
Scan on another device
