The compounding logic
Skilled Education is a rollup vehicle. SFU is acquisition one; more are expected. That changes the unit of value. You do not build a learning companion for SFU. You build the companion and the orchestration layer once, on SFU's real content, then each new institution reuses that same substrate by configuration, not by rebuild.
The work that is institution-specific (course material, branding, language, faculty workflows) is data fed into a shared engine. The work that is hard (the grounded tutor, the retrieval pipeline, the outcome harness, the self-improving loop) is built once and amortised across the portfolio. That reusability is the rollup multiplier: the second institution costs a fraction of the first, the third less again, and the asset you own gets more valuable with every acquisition rather than more expensive.
Per-institution cost curve
ASSUMPTIONS / illustrative only. The figures below are round directional numbers to frame the conversation. They are not a quote, a proposal, or a commitment.
| Institution # | What is reused | Net new build | Relative cost |
|---|---|---|---|
| 1 (SFU) | Nothing yet - this is where the substrate is built | Full companion + retrieval + outcome harness + faculty tooling | 100% (the reference build) |
| 2 | Companion engine, retrieval pipeline, outcome harness, admin shell | Content ingestion, branding, faculty config, language pass | ~40% |
| 3 | Everything from #2 plus the multi-tenant patterns learned | Mostly content ingestion + light config | ~25% |
| 4 and beyond | The whole substrate, now battle-tested across 3 institutions | Onboarding + configuration only | ~20% and trending down |
The shape is the point. The first institution carries the build. Every one after it rides the asset. By institution four, adding a university to the group's AI layer is an onboarding exercise, not an engineering project.
Why this is the investor case
For Oakley, a reusable AI asset across the portfolio is not a cost line - it is a value-creation lever. It raises measurable outcomes at every institution it touches: retention, study-time efficiency, student satisfaction. Those are the metrics that move enterprise value in private higher education, and they compound as the group grows.
There is in-portfolio proof this works. IU Group's Syntea companion - an Oakley sister asset - runs across 80,000+ students and cut study time by 27% per IU's own data. Oakley has already watched an AI learning layer produce a quantified outcome inside its own holdings. The Skilled Education case is to take that proven shape, apply it to SFU and then to the group, and own it as a reusable asset that lifts every acquisition's outcomes and shows up directly in the value of the platform at exit.
Honest scope
The single-institution companion (Branch B) is buildable now. The hard pieces already exist on Tony's one.ie substrate:
| Capability | State | What it means |
|---|---|---|
| Tutor chat grounded in course material | Scaffolded Landed | Core retrieval + chat in active use; ready to apply to SFU content |
| Quiz / summary / podcast generation | Scaffolded | Structure exists, not yet activated on SFU material |
| Single-tenant companion (Branch B) | Scaffolded Landed | The pitch we prove first, on one real SFU module |
| Full multi-tenant group platform (Branch C) | Designed Not Built | The architecture and roadmap - the vision, not yet shipped |
So the proposal is direct: pitch the group platform as the vision, and prove it with a working SFU companion first. The single institution is buildable today. The group layer is real architecture and a roadmap, labelled as such, validated by the institution-one build that proves we ship.
All figures on this page are illustrative assumptions to frame the conversation. They are not a proposal, a quote, or a commitment of price or scope.