Zoom Sheets: from first task to lasting results.

After bringing in data, people need to express relationships, establish rules, keep adding information and work with others. I organized core capabilities and AI assistance around this work.

This task framework connects the four features: charts express data, conditional formatting marks business conditions, onboarding starts real work, and data validation supports ongoing input and maintenance.

People first need to understand which data the chart represents, then adjust its presentation. Both the data and the intended message continue to change after creation.

Users and goals

One framework, different data relationships

Both panels share Data / Format. Their internal controls follow each chart’s data relationships.

Repeated series names and the tooltip are part of the design example.

Explore Series details and delivery decisions

Check the structure with real content

Long cross-sheet references, localized labels and varied backgrounds affect legibility. I refined display, hints and width with engineers and checked ECharts support. Selecting a chart object to locate its editor remained exploratory.

Color may be visible without its cause being clear. Maintaining an existing spreadsheet requires a path from the result back to the rule.

Users and goals

While organizing feedback and bugs, I found that users applied their usual fill-color behavior without getting the expected change. Removing the toolbar entry also lengthened the path to editing. I grouped the issues around rule awareness, editing paths and capability gaps.

Prioritize creation and maintenance for editors

Separate this work from table styles and focus on conditional formatting. Connect creation and management within existing entry constraints. The team jointly prioritized editing scenarios.

Use candidates for clear conditions, AI for complex intent

I proposed two layers: deterministic candidates for clear conditions and optional AI recommendations for complex expressions, while retaining native creation and editing.

Make explanations available without disrupting reading

Tips beside cells were close to the result but interfered with continuous reading, so they were dropped. This tradeoff kept explanation and editing needs grounded in the workflow.

Historical structural draft with placeholder copy.

Some people arrive with a file, some want to explore, and others follow a shared link to collaborate. The first-use experience needs to support these different starting points.

Users and goals

Later proposal: make demonstrations relevant to the task

Occupation provides context; behavior expresses intent

Let people select the work they do, then organize relevant tasks. Demonstrations are opt-in and use a separate sheet, preserving the real file. Occupation and proficiency weights are design hypotheses, not validated user segments.

From managing rules in the current sheet to resolving selection and application scope, make it clear which rule changes and which cells are affected.

Users and goals

Management: make cell-level rules discoverable and editable

Group the current sheet’s rules in a sidebar, identified by type, condition and range. Separate creation from management; confirm bulk removal and clarify that cell values remain.

Identify the rule

Keep type, condition and range together.

Define the operation

Edit a complete rule from the panel; create for the current selection from the grid.

Control bulk actions

Confirm removing all rules while preserving cell values.

Scope: handle the relationship between a selection and existing rules

Setting validation may create a rule, override part of one, or edit a complete rule. I separated paths by entry point and selection state to avoid arbitrary prefilled rules or unintended changes outside the selection.

No existing rule

Create a rule for the selected range.

A subset of one rule

Distinguish changes to the selection from changes to the full rule, including ranges narrowed inside the dialog.

Multiple rules or a mixed selection

Start without arbitrary rule values; apply the new condition to the chosen range and handle the overwritten portions.

Edit from rule management

The complete rule is the editing target; handle cells added, removed or overlapping other rules.

Entry usage did not change for charts, conditional formatting or data validation. After entering configuration, success rates improved, and average operation time fell for charts and formatting.

My work spanned versions, the overall experience, critical features and shared design foundations. Product scale and business performance came from the team’s collective work.

This experience sharpened my focus on what comes after generation: whether people can inspect sources, understand rules, edit the result and hand work to the next collaborator.

The difficulty of converting meeting activity directly into Sheets led me to explore task-specific entry points and reviewable structured outputs. A2A and RSU are independent design and implementation projects.

Shimo AI: early exploration of input, output and expression

I led all brand and product experience design at Shimo with a team of three to four designers. AI explorations began with information types and input/output frameworks, keyword assistance and result flow, informing my later focus on tasks, context and maintaining results.

View the early framework and designs

From collaborative design in a complex product to independent implementation, I keep asking whether people can understand what the system has done and continue their work.