JJachAI

How it works

One engine, two products, an honest lifecycle.

JachAI's central promise is a trustworthy assessment lifecycle: create, publish, invite, attempt, score, review, report and verify. AI assists people within that lifecycle; it does not control official scores or publish unreviewed questions.

Sign in or create a workspace Compare plans

The assessment lifecycle

Every attempt follows the same path, whether it is an Academy batch or a Talent campaign.

  1. 1

    Draft questions

    Instructors and recruiters write single-choice, multiple-choice or manually-graded short-answer questions. AI-generated questions remain drafts until a human reviews them.

  2. 2

    Publish an immutable version

    An assessment is configured and published as an immutable version. Editing later creates a new version; published versions are never changed in place.

  3. 3

    Invite or assign

    A learner or candidate receives an assignment or invitation. The server validates availability, membership or invite, attempt limit and module entitlement.

  4. 4

    Start a server-timed attempt

    The server starts the attempt and records the deadline. The client timer is visual only; the server decides whether an answer is on time.

  5. 5

    Save answers and submit

    Answers are saved with validation and an attempt state check. Submitting is idempotent and cannot change a finalized result.

  6. 6

    Deterministic scoring

    Objective questions are scored by deterministic rules against the published version. Manual-review questions enter a review queue. AI may suggest feedback but cannot silently set official scores.

  7. 7

    Release and verify

    Result release follows organization policy. Topic-level breakdown and weak areas use the versioned question/topic data. Academy certificates are issued only after the required pass/review conditions, and verify publicly.

Instructors & recruiters

Build the question bank, publish assessments, manage campaigns, review results and issue decisions.

Learners & candidates

Take assigned assessments and see released results. Academy learners can use the AI Tutor after release and receive verifiable certificates when eligible; Talent review decisions remain staff-only.

Organization admins

Manage workspace, memberships, roles, billing and result-release defaults. The server re-checks authorization on every request.

Integrity signals, honestly framed

  • Tab visibility, optional fullscreen exit, copy/paste and client-reported answer-save timing can be recorded for an active campaign attempt.
  • They are client-reported, unverifiable and not proof of misconduct. JachAI makes no automatic cheating verdict.
  • No IP address, user agent, clipboard contents or keystrokes are stored, and the data expires on a TTL.
  • Signals are evidence for human review, not a block on the attempt.

AI, where it helps

  • Instructors can ask for AI-drafted questions; they always arrive as drafts for human review.
  • After a released result, learners can use the AI Tutor to ask questions about their performance.
  • AI never sets or changes an official score, and never publishes unreviewed questions.
  • If no provider is configured, the UI says so honestly instead of faking a response.

Security and trust

The same rules apply to every tenant.

Tenant isolation

Every organization-owned document carries an organizationId. Services include it in every read, update and delete.

Server-side authorization

Roles are scoped to an organization. A client-supplied organization ID never grants access by itself.

Append-only audit

Role changes, assessment publishing, score overrides, certificate actions and billing transitions are recorded with actor, target and timestamp.

Curious what a plan can do? Compare FREE and TEAM, or create a workspace to try the engine.