Sludge audits and how they backfire

As agents strip administrative friction out of a service unilaterally, they can clear a rationing step before anyone has audited whether it was doing a job. A friction that only excluded people should go, but a friction that was rationing scarce capacity or verifying eligibility, without that function ever being documented, breaks something when it disappears with no replacement.

Telling those apart before the friction is gone is the work: auditing each point of friction not just for what it costs, but for what it was holding in place.

Sunstein's sludge audit. Sunstein introduced the sludge concept in 'Sludge and Ordeals' and set out the audit itself in 'Sludge Audits': institutions should regularly catalog the costs of sludge and decide when and how to reduce it. Neither paper carries the SNAP figures often attributed to them, which appear in the later book reporting a study where the average application took more than five hours, including two trips to a local office and about USD 10.31 in out-of-pocket costs.

USDA research behind the figures. An Economic Research Service study of the food stamp program carries the same figures, summarizing a 1999 study, and puts the out-of-pocket cost at 6 percent of an average monthly benefit, with each re-certification taking about half as much time again. USDA puts SNAP participation at 88 percent of eligible people in fiscal 2022, its highest on record. More than one eligible person in ten still receives nothing.

OECD International Sludge Academy. The OECD, with the Government of New South Wales, ran an academy in which sixteen government organizations from 14 countries completed sludge audits, out of 22 governments taking part. The audit method is the NSW Government's, and the resulting principles cover behavioral journey mapping, experience scoring, public engagement, and evaluation by randomized trial, A/B test, or before-and-after comparison. Equity, psychological costs, and temporal costs are measured explicitly.

Herd and Moynihan on administrative burden. The book provides the framework underlying sludge audits, identifying learning costs, compliance costs, and psychological costs. It demonstrates that administrative burdens are 'the result of deliberate policy choices'. Where a burden is an intentional rationing tool, removing it is a policy change and not a usability fix.

myGov and the Data and Digital Government Strategy (Australia). The platform carries more than 20 million linked active accounts and serves as the digital front door, and the Data and Digital Government Strategy commits to stopping agencies asking for information another agency already holds. The same strategy reports a survey in which around two-thirds of people said they often had to give the same information to different government services more than once. The Government acknowledges a need to balance usability against friction that safeguards security and fraud risks.

The Consumer Data Right sorts friction by direction. CDR Rule 4.24 bars a data holder from adding 'any requirements to the authorisation process beyond those specified in the data standards and these rules', from collecting extra information, from offering other services, or from referring to other documents. On withdrawal, the CX Guidelines require the consumer be advised to review the consequences before sharing stops, and endorse deliberate friction such as a two-step withdrawal. Friction is prohibited where it helps the institution keep the relationship, and prescribed where its absence would cost the person a service.

The sludge audit framework is highly transferable and already being adopted internationally. The critical adaptation for AI-agent contexts requires a new analytical step, the sludge inversion question. For each friction point identified in an audit, ask:

1. Is this friction purely exclusionary (sludge)? If so, remove it, and note that AI agents will remove it regardless.
2. Does this friction perform a legitimate rationing or verification function? If so, replace it with a purpose-built mechanism (rate limit, structured intake, identity verification) before AI agents render it moot.
3. Does this friction create a beneficial "pause for reflection" (cooling-off periods, mandatory consideration periods)? If so, preserve the pause but redesign it for an agent-mediated context.