Weeks of programme feasibility research, delivered in minutes, scored dimensions, every claim traceable to source, as a stakeholder-ready PDF. Built on Azure with a production-grade multi-agent pipeline.
Before a university commits to a new course or qualification, someone has to answer “is this actually worth launching?”, weighing student demand, job-market outlook, competitors, cost and accreditation. Today that takes weeks of manual research and often ends in inconsistent, hard-to-defend conclusions.
An executive submits a proposal and, in minutes, gets a scored feasibility analysis across every strategic dimension, every claim traceable to its source, as a stakeholder-ready PDF, plus a discovery engine that surfaces new programmes worth exploring.
Runs on Azure as a production-grade, multi-agent pipeline: Entra SSO sign-in, two-layer input guardrails, and a deterministic scoring engine, with dimensions, sources and prompts your own experts configure, no code, no redeploys.
From proposal to a scored, fully-cited feasibility report, in one continuous walkthrough.
Program feasibility shapes strategy and revenue, but the research behind it is slow, subjective and hard to audit, creating real strategic risk.
A React portal where an executive submits a proposal, watches a multi-agent AI analysis run live, and downloads a scored, fully-cited PDF, across multiple strategic dimensions.
Hierarchical agents, Guardrails → Orchestrator → Dimension Agents → Research Teams → Synthesis, run in parallel on Azure OpenAI. Running all dimensions concurrently is what turns weeks into minutes.
Entra SSO sign-in, then two-layer guardrails check the query before any cost is incurred.
Config loads from SQL; agents research all dimensions concurrently with hybrid RAG search.
Each dimension is scored 0–100 with reasoning traces the AI must show, not fabricate.
A final recommendation is synthesised and delivered as a stakeholder-ready, fully-cited PDF.
The AI reads the evidence; a deterministic engine turns it into the number. Same evidence in, same score out, every time. Explainable, auditable, and tunable by your own experts without touching code.
The synthesis agent labels each sub-criterion, strong, moderate, weak, none or negative, with a cited evidence summary.
Labels map to values and combine as a weighted mean across each dimension’s sub-criteria.
Cap, penalty and red-flag rules apply governance ceilings and flags, caps only ever lower the score.
Soft adjustments, a 1–100 clamp, and a confidence + viability tag. Dimension scores then roll up with configurable weights.
| strong, clear positive evidence90 |
| moderate, partial, some gaps70 |
| weak, thin or ambiguous50 |
| none, no evidence found40 |
| negative, actively unfavourable20 |
| Sub-criterion | Wt | Label | Score | Contrib. |
| Employment outlook | 0.35 | none | 40 | 14.0 |
| Enrolment demand | 0.30 | none | 40 | 12.0 |
| Job-market activity | 0.20 | none | 40 | 8.0 |
| Competitive intensity | 0.15 | strong | 90 | 13.5 |
Given identical evidence, the engine returns an identical score, every time. All variance comes from the AI’s reading of the evidence, never the maths. The AI can explain a score but cannot fabricate or nudge it, and every step is fully traceable. Your SMEs retune it from the admin portal, no engineering release.
48,000+ enriched chunks across 17 hybrid search indexes, answers grounded and citable.
Up to ~30 agents run in parallel per feasibility, Guardrails, Orchestrator, Dimension & Research agents.
Admin-defined logic the AI explains but cannot fabricate or nudge.
Deterministic checks in <10ms plus a semantic model, with a token-budget circuit breaker.
A source behind every claim, with every query, decision and config change logged.
Real-time SignalR dashboard as each dimension completes, then a one-click cited PDF.
A focused React experience so academic executives can drive an AI-heavy workflow without training, run a scored analysis, discover new programmes, and configure dimensions and prompts, all in one place.
Adequate resources with some investment required.
High prospective student interest and enrolment potential.
Well-aligned with institutional strengths and resources.
Existing partnerships with clear room to expand.
Positive ROI projected within four years.
Full compliance with accreditation standards.
A national critical skills shortage; analyst postings up ~50% in six months. A Graduate Certificate could launch in 8 months using 60% existing curriculum.
↗ Run feasibility: Graduate Certificate in CybersecurityCross-referencing past analyses reveals an untapped interdisciplinary opportunity, health data science is projected to add 8,000 new roles by 2030.
↗ Run feasibility: Bachelor of Health Data ScienceEvaluates alignment with institutional strengths and academic resources.
A production-grade stack, every component defined in Bicep, deployed via GitHub Actions with OIDC, scoped to Managed Identity. Zero secrets in code.
Azure OpenAI
Azure AI Search
Document Intelligence
Azure FunctionsDurable Functions
Azure SQL
Microsoft Entra ID
Azure SignalR
Bicep + GitHub ActionsRuns entirely within your Azure tenant with role-based access and encryption. Every research run shows its reasoning and confidence, fully logged and reviewable, and a person always has the final say. Aligned with the eight core principles of the Australian AI Ethics Framework.
We'll run the platform on a real proposal and show you the scored, cited result in under a minute, and we can deploy it into your own Azure environment in a matter of weeks.
Get in touch →