Russell Buzby

Learn more

Functional Capacity, Functional Trust: The NDIS reform’s implementation window

When Mark Butler addressed the National Press Club on 22 April 2026, he announced the biggest changes to the National Disability Insurance Scheme since it began in 2013. Over four years, participant numbers will fall from roughly 760,000 to 600,000. Eligibility will move from diagnostic categories to functional capacity assessments using the I-CAN tool. Average plan funding will drop by AU$5,000 in the next two years. Annual cost growth, which has been running near 10 per cent, is being aimed at around 2 per cent. The Minister also acknowledged a AU$13 billion blowout since December 2025.

On paper, the case for reform is sound. The 2023 Independent Review chaired by Bruce Bonyhady and Lisa Paul had warned that the scheme had drifted from its original intent of supporting people with permanent and significant disability. Cost growth had outpaced every projection. The funding ratio between Commonwealth and states had fractured. NDIS Minister Jenny McAllister has been explicit that the policy goal is to return the scheme to its founding promise of choice and control, with dignity and independence, for the people the original architects had in mind. That is a coherent argument rather than a rhetorical one.

The harder question is what happens next. Funding changes for social and community participation begin on 1 July 2026. Providers and participants will feel the impact within weeks. Eligibility changes will not take effect until at least 2028. That is a gap of close to two years between announcing the criteria reset and operationalising it, during which 760,000 people will be wondering whether they remain in the scheme. For a population that has just lived through Robodebt and the Disability Royal Commission’s findings on violence and abuse, the lag is consequential. Trust is not a recoverable asset on the same timetable as a procurement.

This is the live reputational risk. The Conversation, on 21 April 2026, ran a piece drawing the Robodebt parallel directly and asking how computers were still deciding who got care. The article connected automated welfare decision-making to the current NDIS and Support at Home models. The instrument itself is defensible. The I-CAN Support Needs Assessment is a clinically validated tool developed at the University of New South Wales and used in disability research for over a decade. What is at issue is the population context into which it is being introduced. The Australian Commonwealth’s last major experiment with automated assessment of welfare entitlements produced a Royal Commission. Catherine Holmes’s findings on Robodebt were direct about the institutional behaviours that caused the harm, and they remain the operating reference point for any vulnerable Australian who hears the phrase “automated assessment” attached to their support.

The distinction between the tool and the environment around the tool is where most of the implementation work has to happen. Thousands of Local Area Coordinators and planners will need to be retrained to apply functional capacity assessments consistently across geographies and disability types. Provider contracts written under diagnostic categories will need to be redrafted. Appeals processes designed for binary eligibility decisions will need to handle a more graduated, contestable assessment logic. Communications will need to reach people whose first language is not English, who do not use online channels, or whose disability includes a cognitive or communication element that makes navigating reform notices a serious barrier. The scheme is currently running at roughly AU$48 billion a year. The change management envelope has to scale to that, and is not currently sized to do so.

The federalism dimension is the one most likely to be undercooked. The whole-of-system reset depends on intergovernmental agreement on what stays inside NDIS scope and what falls into “foundational supports” delivered by states and territories. NSW Premier Chris Minns has been part of the public framing. The history of joint Commonwealth-state social policy reforms suggests that the announcement is the easy bit. The actual stand-up of foundational supports across eight jurisdictions is a multi-year coordination problem. Consistent eligibility and funding rules across all of them have previously broken on differing state priorities and treasury negotiations. The history of schools and hospitals funding, alongside the original NDIS rollout, shows how the COAG-era version of this work tended to run.

There is a sequencing question here that I have written about separately in the context of the APS reform program. Majed Alsuhaimi’s 2026 study of large-scale public sector change found that change readiness, defined as workforce capability and communication infrastructure alongside operational tooling and supporting information systems, is the single strongest predictor of whether governance reforms produce their intended outcomes. The NDIS reform is being announced before the change readiness work is visible. That pattern is familiar in Australian Commonwealth practice, but it carries a specific cost in this case. The longer the gap between the announcement and the implementation infrastructure being demonstrably in place, the more room there is for advocacy groups and legal challenges to define the reform as a cuts package rather than a sustainability package, and the harder the media environment becomes for the Minister to recover.

There is also an automation governance question that deserves to be in scope. Functional capacity assessment tools produce a numerical score that translates into an eligibility band. Even where the tool is administered by a clinician, the score is the artefact that the system uses. The history of Australian Commonwealth automated decision making, from the 2015-2019 income compliance program through to the Administrative Appeals Tribunal review of NDIS planning decisions, suggests that the gap between “the tool informs a human decision” and “the score is the decision” tends to close over time as caseloads grow and processing pressures increase. Robodebt’s specific failure was the absence of meaningful human judgement and external review at the points where the algorithm produced incorrect outcomes for individuals. The algorithm was the visible artefact of the program; the actual breakdown sat in the appeals architecture and the burden-of-proof structure that the rest of the system layered on top of it. Building functional capacity assessment without an accessible and well-resourced review pathway is the design choice that determines whether the reform is judged in 2030 as the moment the scheme was saved or the moment it was hollowed out.

The substance of the reform may well be correct. The fiscal pressure is real, the original scope creep is real, and the case for returning the scheme to people with permanent and significant disability is defensible. The implementation challenge, however, is the actual policy. Whether the change management infrastructure and federal-state coordination get built, and whether the appeals pathway and the trust environment around the I-CAN tool get built to the standard the population has earned, will be the only thing that matters by 2028. The reform’s credibility was set at the National Press Club. Its outcome will be set in the next 24 months, in the systems nobody is watching yet.

References

Leave a comment