Practical AI workflows
Use Claude Whiteboards to make research decisions easier to inspect
Group observations, preserve participant context and connect a design change to a small evidence table.

In this guide
How can these tools help with a practical task?
Which usability issue deserves investigation in the next design iteration?
A neat research board can hide weak reasoning if observations, interpretations and proposed solutions look the same. Claude Whiteboards is useful when it helps a designer keep those layers visible. Pair the board with a small Sheet that preserves the denominator and the task being observed.
What Anthropic announced
On November 10, 2026, Anthropic introduced Claude Sheets and Whiteboards in beta and made Claude Data, formerly Dashboards, generally available on paid plans. Sheets supports editable formulas and Excel/CSV export; Whiteboards supports shared visual planning. Data works with connected data sources and adds visual stories, including narrated videos. Certified queries and live sharing remain beta features.
Interactive connectors are also expanding to full app views where supported, including Adobe and Atlassian. Availability depends on rollout and organizational settings. Claude Meetings is described as coming soon in the overview; the Whiteboards help page documents Google Meet beta support on Team and Enterprise. Check your account before planning around it; Zoom and Microsoft Teams are not supported in that documented beta.
Before you start
Open Claude and confirm that your paid account has the Sheets and Whiteboards templates in Artifacts. Team plans enable them by default; an Enterprise owner must enable them. If a template is missing, check your plan and organization settings before following this exercise. No live meeting or connected business account is required for the sample below.
Start a chat, paste the labeled sample data and ask for a Sheet. In a separate request, ask for a Whiteboard using the reviewed findings. Copy the relevant context explicitly: the Whiteboards help page says connected-app context is not currently available to the board. These are suggested prompts, not a transcript of a hands-on test in Claude.
Your four-step workflow
- Prepare safe notes. Use permissioned, anonymized observations and preserve task and participant labels.
- Separate the layers. Ask for a Whiteboard with Observations, Interpretations and Design questions.
- Check the counts. Build a Sheet showing attempts and unassisted completions for each task.
- Test a design change. Choose one issue, explain the rationale and define the next usability check.
Worked example: Which usability issue deserves investigation in the next design iteration?
Fictional practice example. These inputs and expected outputs demonstrate the method. They are not client results, portfolio metrics or evidence that this task was run in Claude.
| Task | Participants attempting | Completed without help |
|---|---|---|
| Find delivery price | 5 | 3 |
| Choose a plan | 5 | 4 |
| Find cancellation terms | 5 | 2 |
Paste this prompt into Claude
Create a Claude Sheet using only the following SYNTHETIC PRACTICE DATA. Do not connect accounts or invent missing values. Keep raw inputs separate from calculations. Include visible formulas and an Assumptions tab. The decision is: Which usability issue deserves investigation in the next design iteration? Task,Participants attempting,Completed without help Find delivery price,5,3 Choose a plan,5,4 Find cancellation terms,5,2 Use these formulas, with the header in row 1: D2 = B2-C2 → participants needing help or not completing E2 = IF(B2=0,"",C2/B2) → observed completion rate Explain the calculation and flag missing inputs. Keep zero denominators blank and identify why the result is unavailable. Do not claim these are real business results.
Check the expected result
In this invented five-person study, unassisted completion is 60%, 80% and 40% respectively. Cancellation terms have three participants needing help or not completing.
Five participants do not establish a population conversion rate. Keep severity, observed behavior and participant context beside the counts. A person who finishes after a serious misunderstanding may reveal more than a simple success label. Review the original notes before treating a theme generated by Claude as a finding.
Turn the reviewed result into a board
Create a Whiteboard for this reviewed sample. Make four sections: Evidence, Assumptions, Open questions, and Next action. Evidence: In this invented five-person study, unassisted completion is 60%, 80% and 40% respectively. Cancellation terms have three participants needing help or not completing. Interpretation to review: Five participants do not establish a population conversion rate. Keep severity, observed behavior and participant context beside the counts. A person who finishes after a serious misunderstanding may reveal more than a simple success label. Review the original notes before treating a theme generated by Claude as a finding. Add one owner and review-date placeholder; do not invent names or commitments. Keep the sample labeled fictional. Do not send messages or change connected apps.
Make the handoff useful
Do not have Claude invent quotes or fill a missing session with a plausible story. For the demonstration, use synthetic notes and label them. In client work, check the approved research tools and data-sharing arrangements. The designer remains responsible for whether an observation actually supports a proposed change.
Before sharing, check one ordinary row, one missing value and one zero-denominator case. Confirm that totals use the intended rows. Open the exported Excel file and inspect formulas and formatting; a CSV export does not preserve a workbook’s tabs and presentation. Share only the reviewed version with the intended people.
If you later use Claude Data, inspect the data source, metric definition and refresh time. Ask for the uncertainty to appear in any data story. If Meetings is available to your team, follow your organization’s participant-notice and recording practices; otherwise use notes you are permitted to share. The worksheet exercise works without a meeting assistant.
Try it on one real task
Replace the sample with data you are authorized to use, keep the same review question and record what needed correction. A useful first result is a checked artifact and a decision someone can act on.
Official sources
Feature information checked November 11, 2026. Examples, prompts and interpretation are original editorial guidance.
Sources and further reading
- Anthropic: Sheets, Whiteboards and Data — November 10, 2026
- Claude Help: getting started with Sheets
- Claude Help: Whiteboards, including Meetings availability and limitations
- Anthropic: interactive connectors and full apps
- Anthropic: the wider workplace announcement
Sources checked November 11, 2026. Review the linked publisher information for current details.
Put one idea to work
Do not have Claude invent quotes or fill a missing session with a plausible story. For the demonstration, use synthetic notes and label them. In client work, check the approved research tools and data-sharing arrangements. The designer remains responsible for whether an observation actually supports a proposed change.
Explore design research worksheets →Explore more from UX Design CoachGet design practice in your inbox →Keep practicing
Agent-Friendly Website UX: A Practical Audit Checklist
A practical UX guide to agent-friendly websites: accessible controls, clear states, llms.txt, task testing and a worked booking-flow example.
November 11, 2026 · 8 min read
AI browser workflowsUse ChatGPT and three browser tabs for a more useful UX review
Compare visible interface patterns, keep observations separate from assumptions, and turn the review into one design question. Includes five illustrated steps and an example evidence log.
November 11, 2026 · 8 min read