As the human services sector responds to workforce pressure, providers need a clearer way to measure whether AI reduces administrative burden and improves return on investment. Measuring AI's impact on the workforce is still emerging, but it should be a priority for organisations seeking to address turnover, burnout and service capacity.
Where the sector sits
National Disability Services (NDS) publishes an annual workforce census with the University of Sydney. Its 2025 report, covering the 2024 calendar year, found the following.
| Measure | Result |
|---|---|
| Permanent staff turnover | 16 per cent, the highest since reporting began in 2015 |
| Casual staff turnover | 26 per cent, up from 24 per cent the year prior |
| People who left their job | More than 15,300, across 264 organisations |
| Estimated onboarding cost to the sector | $32 million to $50 million |
The onboarding estimate is built from an Australian Services Union (ASU) figure of $2,130 to $3,320 per new starter, which covers advertising, screening and induction. It does not count the vacancy itself, the ramp to full productivity, supervision time, or for a therapy role, the hours that go unbilled while the position sits empty.
That omission is the reason the sector consistently understates what turnover costs it.
High turnover is not a fixed feature of human services work. A consistent body of research points to administrative load as one of its main drivers, which matters because administrative load is something a provider can change.
- 43 per cent of NDIS workers report feeling burned out at least half the time. (NDIS Review, 2023)
- More than a quarter of aged care workers in Australia and the United Kingdom spend less than a year in care roles. Early departure, rather than gradual attrition, is the dominant pattern. (Thwaites and colleagues, 2023)
- Documentation can consume up to 40 per cent of a nursing shift at the upper end. Time spent on records is time not spent with the people being supported, which is the reason most staff entered the sector. (UW Health trial protocol, 2026)
Churn is expensive, its main drivers are known, and at least administration burden sits inside the provider's control rather than in funding policy or the labour market.
What one departure costs
The cost of losing a support worker and the cost of losing an allied health practitioner are different in kind, not just in size. A vacant shift can be backfilled at a premium. A vacant therapy role stops a billing line and grows a wait list.
| Cost component | Support worker | Allied health practitioner |
|---|---|---|
| Recruitment, advertising and screening | $2,100 to $3,300 | $3,000 to $20,000 |
| Induction, training and buddy shifts | $1,500 to $3,000 | Included in ramp below |
| Cover during the vacancy | $3,000 to $6,000 backfill | $58,000 to $78,000 in lost revenue |
| Ramp to full productivity | Included above | About $29,000 in lost revenue |
| Supervision and management time | $1,500 to $3,000 | About $10,000 |
| Exit admin, final pay and entitlements | $500 to $1,000 | $500 to $1,000 |
| Total per departure | $8,600 to $16,300 | $100,500 to $138,000 |
The assumptions behind these figures are a $70,000 package for a support worker and a $100,000 package for a practitioner. Time to fill averages 8 to 12 weeks for a support worker and 12 to 16 weeks for a practitioner, longer in regional areas. The therapy figures assume 25 billable hours a week at $193.99, the current price limit for occupational therapy and speech pathology.
The NDS census also reported on what providers spend on retention. Across all providers the average is $61,000 a year on burnout and fatigue prevention programmes. This cost needs to be added to the cost of retention to understand the full cost to organisations.
There is a further benefit to reducing turnover that sits outside the cost line. Avoiding 2.5 therapy departures returns roughly 35 weeks of clinical capacity to participants, about $170,000 in billings that would otherwise never be earned. This is counted separately and is not added to the figures above, to avoid double counting.
What is not counted
Every item below would increase the figures. None are included.
- Overtime and agency premiums paid while roles sit open, beyond the backfill already costed.
- Participants who move to another provider during a vacancy, and the revenue that leaves with them.
- Wait list growth, and the referrals declined because there is no capacity to take them.
- The effect of an unstable team on audit outcomes, complaints and quality indicators.
- Manager and executive time spent on recruitment rather than on service delivery.
What the evidence says about reducing administrative burden with AI
If administrative load drives churn, the question becomes whether reducing that load measurably improves retention. The evidence is encouraging on burnout but lighter on turnover itself. Each part matters, and early signs in research are positive.
| Study | Population | Result |
|---|---|---|
| Artificial intelligence scribe across six United States health systems, 30 days (Olson and colleagues, 2025) | 263 clinicians | Burnout fell from 51.9 per cent to 38.8 per cent, a fall of 13 points |
| Mass General Brigham with Emory Healthcare (2025) | 1,430 clinicians | Burnout fell from 52.6 per cent to 30.7 per cent, a fall of 22 points |
| Cleveland Clinic user research (2026) | Artificial intelligence scribe users | 60 per cent said it made them more likely to stay in practice |
Conclusions and recommendations
We are in the early stages of understanding the impact of AI on the human services workforce. Early research clearly points to administrative burden as a key factor in burnout, and to burnout leading to long periods of leave or resignation, so the impact of AI on workforce retention needs to be more fully explored.
The stronger position for organisations is to treat this as measurable rather than proven. Record baseline turnover by workforce before any change, then measure at six and twelve months.
Workforce churn is a measurable indicator, and so is the cost of recruitment, but they may not sit in the same operational budget. When reviewing return on investment for AI, organisations may need to take a bird's-eye view, looking across budget lines to understand the impact of AI on their business.
Sources
- NDS Workforce Census Report 2025, National Disability Services with the Centre for Disability Research and Policy at the University of Sydney.
- NDIS Review final report 2023, Part 2 Markets and Support Systems.
- NDIS Pricing Arrangements and Price Limits 2026-27, Schedule 4 Therapy and Professional Supports, effective 1 July 2026.
- Healthcare Australia 2026 Salary Guide.
- NDIS Review, Building a more responsive and supportive workforce (2023).
- Thwaites C and colleagues, Factors impacting retention of aged care workers: a systematic review, Healthcare 2023;11(23):3008.
- Olson KD and colleagues, JAMA Network Open 2025;8(10):e2534976.
- Mass General Brigham and Emory Healthcare study, JAMA Network Open (2025).
- Cleveland Clinic research, reported in MedCity News (August 2026).
- Scoping review of artificial intelligence documentation and wellbeing, PubMed 42421723 (2026).
- Digital health technology: factors affecting implementation in nursing homes, literature review of 46 articles (2024).
- UW Health ambient artificial intelligence nursing trial protocol, ClinicalTrials.gov NCT07456241 (2026).