What proficiency scale do you use, and can we keep the one we have?

Five levels by default, and you can keep the scale you already run. It configures down to three or up to eight, or maps onto the scale in your HRIS so reporting does not fork. The definitions are yours to edit too: what a 4 means for a given skill, and whether levels are numbered or labelled beginner through expert, sit in a proficiency framework you control. Organisations that arrive with their own competency framework keep it, and we ground it in their task data instead of rebuilding it.

Underneath that one scale, proficiency runs on two separate tracks. Job proficiency is built from tasks: our Task AI breaks a role into its tasks, assigns each task a complexity, and rolls those up into a required level per critical skill. Employee proficiency comes from the same work and HR data behind skill inference (Jira, CRM records, completed courses, project and tenure history), weighted by how recently and how often someone did the work. Every level traces back to its records, so a 4 reads as “level 4, based on 8+ years of experience and 12 workflow records”.

We deliberately do not infer a level for every skill, only for critical, business-relevant ones. We also do not run assessments. We hand back a suggested level, and you decide whether to treat it as a baseline or send it for validation by the employee, a manager, an SME or a formal assessment. Once validated it is flagged as such, and only the validated value feeds matching.

One limit worth knowing before you scope: proficiency inference needs enough underlying data. Where it is thin we will say a skill is required for a job without a confident level on it, or fall back to external market benchmarks until an SME confirms. So settle which systems are connected before you agree which skills get levels. For the task side, read about task intelligence.