National Audit Office NAO

The Access to Work scheme

2026 UK2026accessToWork — Categorised against INTOSAI ICS (GuidICS)
SCALE
  • 74,200 people received Access to Work funding in 2024-25, up 97% from 37,700 in 2018-19
  • 157,000 applications for support in 2024-25, more than double the 76,100 made in 2018-19
  • GBP 321 million spent on provision in 2024-25, a 96% real-terms increase since 2018-19
  • DWP forecasts spending could reach GBP 517 million by 2029-30
  • 62,100 applications awaiting a decision at 31 March 2025, against 21,700 three years earlier
  • average processing time of 109 working days in November 2025, against a 25-day target
COMPLIANCE
  • Employment and Training Act 1973, section 2 - the power under which grants are made
  • Equality Act 2010 - the separate duty of employers to make reasonable adjustments
  • five internal DWP principles governing awards, which are not statutory rules
  • the DWP service target of deciding applications within 25 working days
ECONOMY
  • spending against the Access to Work budget
  • the cost of staff administering the scheme
EFFICIENCY
  • staff numbers assigned to the scheme
  • evidence-reduction measures for travel-to-work, special aids and support-worker applications
  • cases processed per case manager per day against the DWP standard
EFFECTIVENESS
  • outstanding applications and payments against staff numbers
  • whether DWP assesses the scheme against its objectives
  • complaints about the scheme and what they concern
1. DWP cannot show whether the scheme delivers value for money Results review Monitoring

While the scheme clearly has benefits, DWP "does not know whether the scheme provides value for money"; it has little evidence on the difference the funding has made and does not check whether the support was of benefit (Summary para 8, p.7; paras 1.10-1.12, pp.15-16). There is no business case or underpinning legislation providing a reference point for assessing value for money, and DWP has undertaken no quantitative assessment of the scheme's effectiveness.

  • Domains: Persons with disabilities, Labour
  • Function: monitoring, Governance
  • Quality: Functioning oversight and governance
2. Loosely defined guidance has let award decisions drift from policy intent Guidance Regulatory_system

DWP "has concerns that, because it has not designed the scheme in a prescriptive way or defined the rules tightly, awards have not always aligned with the policy intent and funding decisions have been inconsistent" (Summary para 9, p.8). An internal audit in August 2024 found DWP's guidance for Access to Work "high-level" and limiting what case managers can do to challenge applicants (para 1.13, p.16); a 2025 review of 200 cases found some 40% of funded support should have been provided by the employer, employee or NHS, and that employers had failed to make reasonable adjustments in 18% of cases reviewed (para 1.16, p.17).

  • Domains: Persons with disabilities, Labour
  • Function: Regulations, Governance
  • Quality: Clear objectives and goal-setting
3. Processing times and backlogs have grown far beyond DWP's own targets, with no early relief expected Goal-setting Monitoring

Average processing time "more than doubled, from 28 working days in 2021-22 to 66 working days in 2024-25" and reached 109 days by November 2025, against DWP's 25-day target (para 3.2, p.29); the backlog of applications awaiting a decision "almost trebled, from 21,700 at 31 March 2022 to 62,100 at 31 March 2025" (para 14, p.10). DWP "does not expect the backlog to fall significantly in the short term" without policy change, additional budget and further productivity improvements (para 18, p.11).

  • Domains: Persons with disabilities, Labour
  • Function: monitoring, Work process design
  • Quality: Responsiveness to user needs and feedback
4. Case data are not robust enough to diagnose where the process breaks down Documentation Tools

The NAO "were unable to use DWP's data for more detailed analysis, for example to identify where in the process cases are most commonly held up", because the data "are not sufficiently robust as the manual input of large volumes of data by case managers increases the risk of inaccuracies", and DWP "does not record information on all parts of the process" in a form susceptible to analysis (para 3.4, p.31).

  • Domains: Persons with disabilities, Labour
  • Function: documentation, IT
  • Quality: Reliable, integrated information base
5. Efficiency measures have knowingly traded away part of the scheme's control framework Procedures Regulatory_system

Having removed evidence requirements for travel-to-work, special-aids and support-worker applications to speed up decisions (para 3.11, p.35), the NAO notes that "while the measures may have reduced the administrative workload... reducing the amount of supporting evidence that customers have to supply will also weaken DWP's control framework and increase the risk that grant awards are not aligned with the policy intent of the scheme" (para 3.12, p.36).

  • Domains: Persons with disabilities, Labour
  • Function: Work process design, Regulations
  • Quality: Sound risk management
6. A fast-growing volume of reconsiderations is consuming capacity meant for the backlog Monitoring Responsibility

Requests for reconsideration of awards rose "from 194 in 2021-22 to 385 in 2023-24, with a more significant increase to 1,575 in 2024-25"; case managers spend on average 241 minutes on a reconsideration against 115 minutes for a new application, so "the more reconsiderations there are, the fewer resources there are available to tackle the backlog of outstanding applications" (para 3.13, p.36).

  • Domains: Persons with disabilities, Labour
  • Function: Work process design, monitoring
  • Quality: Streamlined, standardized processes
7. More than doubled staff and spending have not lifted case-manager productivity to DWP's own standard Human_capital Training Goal-setting

Scheme staff rose from an average of 247 full-time equivalents in 2021-22 to 588 in 2024-25, with administration costs rising in real terms from GBP 12.6 million to GBP 24.4 million over the same period (para 3.8, p.34), yet "since May 2023, case managers have processed fewer cases per day on average than" DWP's own standard of 2.8, which DWP no longer considers achievable (para 3.10, p.34); induction for new case managers takes around six months, diverting experienced staff from casework to training (para 3.14, p.37).

  • Domains: Persons with disabilities, Labour
  • Function: Human resources, training
  • Quality: Adequate resources and competences
Control focus
ICS phaseControl functionCases
Completion of processResults review1. DWP cannot show whether the scheme delivers value for money
Work processesMonitoring1. DWP cannot show whether the scheme delivers value for money<br/>3. Processing times and backlogs have grown far beyond DWP's own targets, with no early relief expected<br/>6. A fast-growing volume of reconsiderations is consuming capacity meant for the backlog
Tools4. Case data are not robust enough to diagnose where the process breaks down
Procedures5. Efficiency measures have knowingly traded away part of the scheme's control framework
Initial phaseGuidance2. Loosely defined guidance has let award decisions drift from policy intent
Goal-setting3. Processing times and backlogs have grown far beyond DWP's own targets, with no early relief expected<br/>7. More than doubled staff and spending have not lifted case-manager productivity to DWP's own standard
Responsibility6. A fast-growing volume of reconsiderations is consuming capacity meant for the backlog
Training7. More than doubled staff and spending have not lifted case-manager productivity to DWP's own standard
UncategorisedRegulatory_system2. Loosely defined guidance has let award decisions drift from policy intent<br/>5. Efficiency measures have knowingly traded away part of the scheme's control framework
Human_capital7. More than doubled staff and spending have not lifted case-manager productivity to DWP's own standard
Functions applied to all stagesDocumentation4. Case data are not robust enough to diagnose where the process breaks down
This page is part of CUBE, a knowledge-sharing initiative of the EUROSAI IT Working Group. Its purpose is to make what supreme audit institutions find easier to search, compare and reuse — by auditors, and by the wider public who rarely reach these reports in their original form. It presents an analysis prepared, with AI assistance, by Paweł Banaś (NIK — Najwyższa Izba Kontroli, Poland) on the basis of the publicly available report of National Audit Office, categorised against the internal-control terminology of INTOSAI's Guidance on Auditing Internal Control (ICS), drafted by the Internal Control Standards Subcommittee, which NIK (Poland) chairs. The categorisation and the case selection are ours, not the audit institution's, and so is any error in them. Readers are warmly encouraged to go to the original report, linked above; this page is a way in, never a substitute. Underlying data.