Choose a perspective that matters to your product
Your team knows where the important button is. Someone opening the product for the first time may not. A persona gives Flock a particular perspective to bring to that experience: familiarity with the domain, patience with friction, priorities, and expectations.
Start with the audience and task you care about. A novice perspective can help you examine onboarding. A skeptical evaluator can help you inspect how clearly pricing and commitments are explained. A busy operator can bring attention to the effort required to finish a routine task.
These are AI reviewers, not recruited customers. Use their observations to find questions worth investigating and improvements worth testing.
Select personas for a review
In the dashboard, choose New Review, keep Guided selected, and choose the kind of review you need. Single Page Run reviews one page. Goal lets you define a task and choose whether to stay on that page or follow the flow.
In Target, enter the URL to review. Then, in Personas, choose from the catalog available to the selected workspace. It includes library personas and published workspace personas. Start with a small set of distinct, relevant perspectives; adding more of the same perspective makes the review larger without necessarily making it more useful.
For example, review a pricing page with an available novice perspective and a skeptical evaluator. Inspect what each understood about the plans, commitment, and next step. When they disagree, open their individual findings and screenshots rather than relying only on an average score.
Complete Auth as needed, then inspect the setup and estimated usage in Review before choosing Launch Review. Creating a review currently requires a desktop-sized browser; you can read existing reviews on mobile. The getting-started guide walks through the full setup.
A persona is a perspective. A goal is a task.
Keep reusable characteristics in the persona and the task for this particular review in the goal.
- Persona: a first-time workspace administrator with limited familiarity, moderate patience, and a preference for clear guidance.
- Goal: create a workspace and reach its welcome page.
- Completion criteria: the expected welcome message or page is visible.
- Boundaries: stop before sending invitations or entering payment details.
That same persona can review signup, settings, and invitations with a different goal each time. A profile’s goal orientation describes its approach to work; it does not replace the review’s task or completion criteria.
Create or adapt a workspace persona
Open Personas in the dashboard. If persona authoring is available for your workspace and role, choose Create Persona or Create from attributes. To adapt a library profile, use Create Workspace Fork. A fork gives your workspace its own version without changing the library persona.
Give the persona a useful name and describe the context that matters. “First-time workspace admin” is more helpful than a biography full of details that have no bearing on the review. Define how familiar they are, what they prioritize, and what guidance they expect.
Save and validate the draft, then publish it when it is ready to use. A saved draft is distinct from a published version you can select for reviews. Workspace personas belong to the workspace where you created them; changing workspaces can change the catalog you see.
For a personal workspace, the owner can author personas. In a shared workspace, owners and admins can author them. Feature availability and limits also depend on the workspace. Use the controls and limits shown in your dashboard rather than assuming every member has editing access.
Make traits and guidelines useful
The editor supports preferences such as clarity, trust, speed, friction tolerance, and experience. Numeric traits use a 1–10 scale to express preference strength. They guide the model’s review; they are not calibrated weights that produce a predictable score change.
Use guidelines for concrete context. For a first-time administrator, you might add: “Look for a clear explanation of what inviting a teammate will do.” Keep the actual invitation task and its success criteria in the review setup.
Profile details alone do not establish a tested capability. For example, describing an accessibility need does not mean a screen reader was used. Inspect the evidence and the type of review that ran.
Reuse the perspective across your workflow
Select published personas when creating a dashboard review. In the GitHub App, choose default personas in review preferences, with separate selections for a target where needed. Keeping personas consistent helps your team compare the same journey over time.
With MCP, your coding agent can recommend a review using the personas available to its authorized workspace. Persona creation and editing happen in the dashboard; the onboarding tools select existing personas.
For example, ask your agent: “Recommend a small review of this signup flow using personas available in this workspace. Explain why they fit, show the scope and estimate, and wait for approval.”
Customer perspectives and specialist reviews
The selector can also offer Design Reviewer and technical auditors. These answer different questions from a synthetic customer’s impression of an experience. A technical audit can identify a browser or implementation issue without supplying all six experience ratings.
Some configurations also support generated populations. A population creates a set of synthetic perspectives from configured distributions. It can broaden a review, but its size is not a sample of real customers or a measure of market prevalence. Check the available configuration, workspace limits, and estimate before running it.
Read what this persona encountered
After the review, open the persona’s run. Start with its task outcome, then inspect the findings and screenshots. Use review scores to understand its impression and findings and evidence to understand which specific claims have support.
When comparing a change, keep the persona and version, goal, starting page, sign-in, and flow limits consistent where practical. The useful question is whether the same difficulty remains under comparable conditions.