CriticalEyeWorked example
From "improve onboarding" to an approved goal.
A concrete Goal Canvas makes the target, trade-offs, and proof visible before an agent creates a plan.
Approved Goal Canvas - Version 3
New customers get a useful result without assistance.
For new customers who signed up on their own, make the first successful workflow clear enough to complete in one session.
31% of new customers complete a first workflow within 24 hours. Median time to a useful result is 18 minutes.
45% complete within 24 hours for a sustained 30-day measurement window.
Median time to a useful result stays under 10 minutes. Support contacts do not increase by more than 5%.
No broad navigation redesign, pricing change, or automated product tour.
Plan built from the goal
Plan a vertical slice, not a collection of tasks.
Milestone 1
A usable first workflow.
A new customer can choose a template, provide the minimum required inputs, run one workflow, and see a useful result without outside assistance.
This milestone proves a necessary part of the path. It does not, by itself, prove that the complete onboarding goal has been achieved.
- A new customer can complete the guided first workflow end to end.
- The workflow produces a useful result without a blocking error.
- Median time to a useful result is below 10 minutes for the measured group.
Goal evidence
Then check whether the full goal worked.
Tests confirm that the guided workflow behaves as specified. Milestone evidence shows whether this step works in use. Launch checks separately show whether the complete onboarding goal became and stayed true.
- New customers completing a first workflow within 24 hours rises from 31% to 45%.
- The improvement is sustained throughout the 30-day measurement window.
- Support contacts do not increase by more than 5%.
- Quality, cost, reliability, and safety limits remain satisfied.
Launch checks
The goal remains active after the last planned milestone.
Re-evaluate the full Goal when new Evidence arrives: the intended job, quality bar, cost, reliability, time window, and guardrails. A result that worked once may later slip, but CriticalEye does not schedule these checks by itself.
Evidence-driven decision
More users succeeded, but support requests rose too much.
The main result improved, but one of the limits in the goal was missed. CriticalEye shows that the goal is not ready to approve, explains what went wrong, and suggests ways to respond. It never quietly lowers the bar.
Reduce the workflow, improve the guidance, gather more evidence, reopen the plan, or revise the goal.
Final decision
CriticalEye recommends. A person decides whether the goal is done.
The final review puts the approved goal beside the evidence, what each milestone showed, any open concerns, and CriticalEye's recommendation. CriticalEye can say Ready to approve, Not ready, or Not enough evidence. Someone with permission makes the final decision.
Human decision: NOT DONE. Maya Chen reopened the plan on July 28, 2026 because support contacts exceeded the approved limit.
Result slipped after approval
Later Evidence can challenge an earlier approval.
A result that supported the goal when you approved it is no longer holding. CriticalEye has not changed the goal or plan.