The Chairman's Dashboard
One of five simultaneous client engagements at Lollypop, and the one that became the origin point of the Graduated Autonomy framework.
Jump to outcomes & impactProblem & research background
At Lollypop Design Studio, I wasn't working one project at a time. I was carrying five simultaneous client engagements at once: cybersecurity, digital insurance and health tech, EdTech, POS systems, and an enterprise oversight dashboard built for a Dubai-based chairman. The challenge wasn't any single brief, it was keeping research depth intact across all five without breadth turning into dilution, and the dashboard's urgency meant it pulled a disproportionate share of resources into a short timeframe while the other four kept running.
The starting research base was the chairman's own existing systems: I analyzed the business intelligence reports and reporting tools he already used, and built the first conclusions from that data rather than from a fresh round of interviews. The dashboard needed to pull together a wide spread of information, from CSAT and NPS through investment returns and overall company health, so the chairman could set direction for every unit in the group from one place, and check a unit's reported numbers against the underlying data himself, rather than relying solely on how that unit chose to present them.
My role, process & key decisions
My role & process
Across all five engagements, my process stayed consistent even as the subject matter changed underneath it:
- Research synthesis: combined qualitative and quantitative research, usability testing, and behavioral analytics to find the pain points worth designing around, across every vertical at once.
- Data visualization mapping: built the dashboard's reporting layer in Power BI, using a spreadsheet as the main source feeding into it, to map the group's data onto the requirement we'd identified.
- Design systemization: led design system work from discovery through implementation on multiple client products in parallel, rather than treating each engagement as a one-off.
- Workflow acceleration: applied prompt engineering to structure LLM system prompts for research synthesis, design briefs, and stakeholder communication, cutting time-to-first-draft across the five concurrent engagements.
- Stakeholder validation: rapidly prototyped in Figma and ran structured feedback sessions to reduce misalignment before engineering handoff, on the dashboard as much as anywhere else.
Key design decisions
- Verification over presentation. The dashboard was built to give the chairman a way to check reported metrics against the underlying data himself, instead of depending entirely on how each unit chose to present its own numbers. That distinction, between what's reported and what's underneath the report, shaped most of what the dashboard needed to do.
- Spreadsheet as a starting point, not an end point. The Power BI layer worked, but it sat on top of a spreadsheet as its main data source. Once I saw where that broke down under the volume and pace of a group-wide dashboard, static and manual, it pushed me toward wanting something more dynamic and automatable instead.
That second decision is the one that generalized. Not every piece of information a chairman receives needs the same level of scrutiny or escalation: some of it a system can be trusted to check and surface on its own, some of it still needs a human to sit with it. Designing toward that distinction, and away from a spreadsheet that couldn't scale to it, is what turned into the Graduated Autonomy framework.
Graduated Autonomy: from LLMs to RLMs
This engagement is the confirmed origin point of a framework I've since developed further: how much autonomy a system, human or AI, should be trusted with, and how that trust should scale rather than switch on or off. That project page is still in progress. When it's ready, it'll link back here.
Artifacts & evidence
How this page is styled
This flow is a generic reconstruction built from the case study's own description of what the dashboard needed to do (pull CSAT/NPS/investment returns/company health into one place; let the chairman verify a unit's numbers against underlying data), not a recreation of the real delivered screens.
User Flow: The Chairman's Dashboard
Outcomes & impact
| Metric | Result | Source |
|---|---|---|
| Design iteration time (across 5 engagements) | −30% | Internal tracking |
| Chairman's dashboard outcome | BI-verified | Chairman's existing BI reports and systems |
What's already confirmed is the framework it produced: recognizing where a spreadsheet-backed dashboard stopped scaling, and needed to become something more dynamic and self-checking, is what led directly to Graduated Autonomy, now a central piece of how I think about, and design for, systems that need to earn trust incrementally rather than all at once.