Verified readiness: proof a new hire can do the job, not a completion tick

A completion record says the modules were opened, the slides were clicked through and a quiz was passed, however many tries it took. It does not say whether a bank teller will recognise a suspicious cash pattern at the counter, or whether a support engineer knows when a customer's data request has to go to the privacy team.

Most HR and learning software uses AI to write training faster. That makes more content, and the same completion tick at the end of it. Ascend uses AI as the examiner: it checks what each person actually understood, records it as a skill level against their job role, and refuses to call anyone ready on completion alone.

• The company brain. Your policies, SOPs and job descriptions are loaded into a knowledge base scoped to your organization. Everything below is drafted from it.
• Drafted, role-specific training and onboarding. AI drafting agents write a training program and a 30/60/90-day onboarding plan for a job role, citing the document passages they drew on. By default a person approves every draft before anyone is assigned it.
• The Understanding Check. Each module ends with a question the person answers in their own words. The question and its rubric are generated in the same run as the module, from the same sources, so the test cannot drift from what was taught. The answer is scored criterion by criterion.
• A skill level against the job role's bar. Each job role lists the skills it needs and the level required, on a 1 to 5 scale. A passed check records the level its module verifies, with the submission as the evidence. A pass never lowers a level someone already holds.
• The readiness gate. Every onboarding plan carries at least one gate, and a gate closes only on strong evidence: a manager's sign-off, a passed Understanding Check, work a manager reviewed and accepted, or a passed certification conversation. Ticking a box or finishing a program never closes one.

A failed Understanding Check returns what the answer missed, criterion by criterion, and a short re-explanation of that concept written from the module's own approved text. The person gets three attempts. If grading itself fails for a technical reason, the attempt is not counted against them.

After a third failed attempt the person can move on so they are not stuck, but the record says so: the check is marked force-unlocked, it confers no skill level, and it cannot close a readiness gate. A pass that took more than one attempt, or left a criterion unmet, schedules two refresher questions, three and ten days later.

For roles where the job is a conversation, a gate can close on a certification conversation: the AI plays the customer, the person handles the situation, and the transcript is scored against a rubric written with the scenario. A pass is recorded at once, and the person's manager is notified, can read the full transcript, and can overrule the verdict.

• Readiness is worked out from the evidence every time it is read, never typed in: not started, in progress, at risk, waiting on a sign-off, or ready.
• Each plan has a printable evidence pack: the person, the role, the readiness state, every piece of evidence with its kind and date, and anything still outstanding.
• Every Understanding Check verdict, force-unlocks included, is written to the audit log.
• No onboarding template can be published without at least one gate. That rule is part of the product, not a setting an administrator can switch off.

Does a person have to sign every readiness decision?

Every plan needs at least one gate closed by strong evidence. Where that gate is a manager sign-off, the manager signs it and nobody can sign their own. A passed Understanding Check or certification conversation can close a gate by itself, and a certification pass goes to the manager, who can overrule it.

Isn't an AI-graded answer just a harder quiz?

The person writes the answer in their own words, and it is scored against named criteria from the module it tests. A failed answer comes back with what was missing and a new explanation, not a retry of the same questions.

Where does the job role's bar come from?

You set it. An administrator builds each job role with its skills and required levels, and decides which modules verify which skill at which level.

What happens when someone moves to a new role?

The new plan is drafted from the gap between the skill levels they hold and the ones the new role requires. Skills they have already proven appear as already demonstrated, with the evidence that proved them.