MSP 2.0 will fail if it just rebuilds the old power imbalance with better branding

Defence is designing the next Major Service Provider model. The RFI (SIAD/RFI/90404/1) asks industry the right questions: forward demand visibility, skill-based workforce planning, standing panels, contract terms, pricing structures, knowledge transfer.

Those are the easy questions. There is a harder one buried underneath them, and it is the one that will determine whether MSP 2.0 actually works.

Who manages the workforce, and who competes for the work it delivers, cannot keep being the same party.

The pattern worth naming

Layered MSP and subcontractor arrangements have a known failure mode. A managing contractor sits between Defence and the delivery subcontractors. It sees the rates, the availability, the forward pipeline. It also bids for adjacent delivery work.

When that happens, subcontractors face rate and contract terms with limited room to negotiate or escalate, on a take it or leave it basis, a pattern Australian Consumer Law already recognises as a marker of an unfair contract term when the imbalance is severe enough. Defence, meanwhile, loses the one thing it actually needs from a workforce model: assurance that resourcing decisions reflect capability, not the managing party's commercial position.

This is not a hypothetical. It is a structural risk that shows up wherever one intermediary controls both the data and the delivery decision. Left alone, it does exactly what you would expect. It compresses subcontractor margins below long-held baselines, sometimes without clear evidentiary link to any genuine Commonwealth-directed change. It discourages investment in Defence-specific capability. It erodes the subcontractor base MSP 2.0 is supposed to be mobilising, not depleting.

What actually fixes it

Three things, and none of them are exotic.

Structural separation, or at minimum transparent and auditable rate-setting, between the entity managing workforce allocation and any entity competing for the delivery work itself. If separation is not achievable, the rate-setting has to be visible enough that nobody has to take it on faith.

Direct, unmediated Defence visibility into subcontractor-level workforce data, rates and capacity, plus an escalation path into Defence's Panel Manager function that does not run through the managing contractor. If every demand signal and every complaint passes through one intermediary, that intermediary shapes what Defence sees. That is not partnership. That is a chokepoint with a service agreement attached.

A shared-risk commercial model where a portion of the managing contractor's margin is tied to independently verified fair treatment of its subcontractor base, not just delivery output. Retention rates. Absence of substantiated disputes over rate changes. Verified by an independent, confidential survey Defence runs itself, not the managing contractor.

Why this matters beyond fairness

This is not only about protecting subcontractors, although it is also that. It is about whether Defence gets what it is actually paying for.

Knowledge transfer that is contractually mandated but delivered by a party incentivised to keep billing gets quietly under-delivered, because doing it properly shortens the engagement. Talent management collapses when the return to industry is squeezed too thin to fund training, clearance sponsorship and career paths, and suppliers default to generalist staff and high turnover instead. Speed to delivery slows the moment capability gets gated by a single intermediary's commercial position rather than genuine availability.

None of this requires naming a party or relitigating the past. It requires designing MSP 2.0 so the incentive structure cannot produce the failure mode again.

The question for the design phase

Defence has an open window right now, before MSP 2.0 is locked in, to decide whether the managing layer and the delivery layer sit inside the same commercial interest or not. Everything else in the RFI, pricing models, contract terms, performance frameworks, works better once that one design choice is made correctly, and works around the edges of a problem if it is not.

The RFI closes 22 July 2026. Worth getting this one right before the model is built around getting it wrong again.