Demonstration · synthetic data
A controlled admissions workflow from program inquiry to advisor-owned next step.
This lab specification demonstrates validation, transparent program rules, advisor routing, booking, document state, exceptions, and measurement using fictional records only.
Demonstration boundary: Synthetic data only. This is not a client production system and does not claim enrollment improvement.
Why this project exists
AH Shagar already has an existing growth case in the trade-school sector. This demonstration extends sector understanding into the operational journey after a lead arrives. The two items must remain separate: the older client case can support verified organic/enrollment results; this build proves the proposed automation pattern on synthetic data.
Demonstration boundary
- No real student data
- No client production system
- No autonomous eligibility, financial aid, visa, or enrollment decision
- No claim that this workflow improved the prior client’s enrollment
- Program rules are fictional but realistic and published with the fixture
Architecture
Demonstration flow
- Prospect selects a fictional program, start window, location, contact method, and consent.
- Validation checks required contact fields and blocks malformed/spam input.
- Identity logic finds replay/duplicate inquiries.
- Deterministic rules identify location/program availability and required follow-up.
- Optional AI summarization structures free-text goals but cannot decide eligibility.
- CRM stores program, source, stage, priority, advisor, SLA, and reason.
- Prospect receives an acknowledgement and relevant tour/call options.
- A document checklist tracks requested/received/exception state without real documents.
- Dashboard shows response, routing, booking, show-up fixture, and simulated stage flow.
Synthetic fixture design
Create 40-60 fictional inquiries with explicit expected outcomes across:
- Three fictional programs
- Two locations plus out-of-area
- Day/evening preference
- Immediate, 1-3 month, and research-only start intent
- Valid, incomplete, duplicate, conflicting, and opted-out contacts
- Questions that approved FAQ content can answer
- Questions requiring advisor handoff
- Ambiguous free text and deliberate prompt-injection text
- Advisor unavailable and CRM/calendar provider failures
- Applicant document states: not required, requested, partial, complete, invalid
Publish the generator rules and a safely downloadable sample CSV if it cannot be mistaken for real student data.
Rules example
IF location is outside service area
THEN route to admissions review; do not reject automatically
ELSE IF requested program is unavailable at selected location
THEN provide approved alternatives and assign advisor
ELSE
assign advisor by location + program + availability
IF question is supported by approved FAQ source
THEN provide cited response
ELSE
create advisor task with the question and sources checked
Required test scenarios
- Valid inquiry for each program/location combination
- Duplicate exact replay and duplicate contact/new intent
- Out-of-area and unavailable-program exception
- Missing/invalid contact data
- Opted-out existing contact
- FAQ answer with correct citation
- Unanswerable, contradictory, and injected question refusal/handoff
- Advisor route unavailable
- CRM, email, and calendar downtime
- Document upload/type/size rejection using harmless fixtures
- Reminder stop on response/booking/opt-out
- Dashboard reconciliation across all fixture stages
Acceptance criteria
- Expected advisor/stage/reason matches fixture for all deterministic cases.
- No eligibility or admissions acceptance decision is produced by AI.
- Exact replay produces no duplicate message/CRM inquiry.
- FAQ answers cite approved fictional program/policy sources.
- Unsupported or sensitive questions hand off.
- All failures create a recoverable exception with owner.
- Metric dashboard matches the synthetic fixture’s known totals.
- Every public screen visibly states
Demonstration / Synthetic data.
Evidence required for completed status
- Architecture SVG with demo caption
- Synthetic inquiry form and data label
- Workflow/rules view
- CRM/advisor record using obviously fictional identity
- Booking and document checklist state
- Exception/retry test
- Reconciled synthetic dashboard
- Captioned demo showing happy path and unsupported-question handoff
Current limitations
- The project is not connected to a real school’s CRM, policies, staff, or student data.
- Fixture processing speed does not predict enrollment conversion or staff adoption.
- Real deployments need policy, consent, accessibility, records, identity, and possibly regulatory review for their jurisdiction and student context.
- Admissions staff remain responsible for eligibility, advice, exceptions, and enrollment.
Next step
Map where prospective students wait or staff repeat work.
Bring the current process, tools, approximate volume, and result that matters. The first conversation will identify the smallest useful next step.