Evidence is fragmented
Deadlines, preferences and earlier actions arrive in different shapes and at different times.
Careers operations for further education
Every learner needs a next step. Every adviser needs a workable plan.
Career CoDesk turns fragmented learner signals into a capacity-aware support proposal, keeps evidence separate from AI interpretation, and stops for a human decision.
Interactive walkthrough · choices reset on refresh and never alter the project’s evaluation evidence.
Capacity is finite
A good recommendation is not useful if the team cannot schedule or deliver it.
Consequences need a human
AI can prepare evidence, but an adviser must approve, amend or reject the action.
One inspectable path
From a learner signal to an approved action—and back when it did not help.
The workflow makes the scarce decision visible: who needs which support, what the service can actually deliver, and who takes responsibility for the final choice.
Intake
Allowed fields enter with a source snapshot and simulated identity.
Interpret
Source statements remain separate from provisional AI hypotheses.
Plan capacity
Deterministic rules test slots, alternatives and unmet demand.
Decide
An adviser approves, amends or rejects with a recorded reason.
Deliver
Only the approved action enters the weekly execution package.
Feedback
Helped, unresolved, correction and human-help signals remain visible.
Interactive reviewer walkthrough
Review a fictional case without touching the product database.
Choose a learner, follow the evidence and try a human decision. This facade mirrors the implemented product journey using story-safe labels.
Simulation mode: controls update only this browser view and reset on refresh.
Integration and privacy by boundary
Designed to sit beside a college MIS, not replace it.
The product interface isolates partner-specific field mapping from the decision workflow. A college can adopt the same core model while keeping its own identity, permissions, retention and governance controls.
College systems
MIS / CRM / CSV
- Identity and enrolment
- Existing learner signals
- Institution-owned permissions
Career CoDesk boundary
Adapt, interpret, plan, approve
- Versioned adapter contract
- Allowed-field normalisation
- Evidence / inference separation
- Deterministic capacity checks
- Recorded human decision
Approved output
Execution package
- One approved action
- Versioned local export
- Feedback and reopen signal
MIS adapter boundary: connector not enabled in this public demo.
Public-demo boundary. No real learner names, contact details, dates of birth, addresses or institutional records are used. The design supports data minimisation and traceability; production DPIA, retention, lawful basis, access control and connector acceptance remain the institution’s responsibility.