What a 2026 DCWP Review Actually Checks

Summary: The December 2025 State Comptroller audit did not just criticize DCWP's enforcement volume, it named the specific tools DCWP had been failing to use: its own Enforcement Workbook, and the technical expertise of the city's Office of Technology and Innovation. DCWP has committed to using both going forward. That commitment translates into a materially more specific set of things a 2026 review will actually check, beyond simply asking whether an audit exists.

Why this is worth reading closely

Most compliance guidance treats a DCWP review as a binary question: do you have a bias audit, yes or no. The Comptroller's audit revealed that this was, in effect, the level DCWP itself had been operating at, and that it produced a large miss rate. Of 32 employer and vendor bias audit disclosures DCWP reviewed and found largely compliant, the Comptroller's own review of the same 32 identified at least 17 potential non-compliance issues. The gap between those two numbers is the gap between a surface check and a real one.

What DCWP committed to changing

Three specific commitments came out of the audit response, and each implies a different thing a reviewer will now be equipped to check.

Using the Enforcement Workbook. DCWP had not been using its own internal workbook when evaluating whether tools qualified as AEDTs. A workbook implies a structured set of criteria applied consistently, not an ad hoc judgment call. Employers should expect classification decisions to be tested against a defined standard rather than accepted at face value.

Consulting the Office of Technology and Innovation. DCWP staff lacked the technical expertise to evaluate AEDT functionality and had not been consulting OTI's resources. A reviewer with access to technical expertise can meaningfully question how a scoring or ranking feature actually works, not just whether a document labeled "bias audit" exists.

Proactive, research-driven enforcement. DCWP committed to identifying potential non-compliance through research and leads, not only complaints. This changes where scrutiny is likely to originate. A remote job posting, a LinkedIn description of a company's screening stack, or a vendor's own marketing about its AI features are all now plausible starting points for a review, not just a candidate complaint.

What to have ready

Translating those three commitments into a practical checklist:

1. Documented reasoning for every tool classification, not just a conclusion. If DCWP is now applying a structured workbook, "we decided it wasn't an AEDT" needs to survive being tested against actual criteria, in writing.
2. A bias audit methodology a technically literate reviewer could interrogate. Selection rates and impact ratios calculated correctly, across the right categories, by an auditor whose independence can withstand scrutiny.
3. A publicly posted audit summary that is actually current and actually findable, since the Comptroller specifically flagged inadequate review of exactly this kind of disclosure.
4. Notice delivery you can evidence, not just describe. A policy stating notice is sent is different from being able to show it was sent, to whom, and when.
5. An honest inventory of scoring or ranking features embedded in third-party platforms, since a technically resourced reviewer is more likely to spot a feature an employer did not think to flag.

The shift in posture, summarized

Through 2023 to 2025, the operative question was largely "does a bias audit exist." Going into 2026, the operative question is closer to "would this audit, this classification, and this notice process survive being checked by someone with a workbook and technical backup." That is a materially higher bar, and it is the bar the Comptroller's audit was specifically designed to raise.

Frequently asked questions

Does this mean DCWP is auditing every employer now? No. Enforcement remains bounded by DCWP's resources. What has changed is the sophistication and proactivity of reviews that do happen, not necessarily their volume, at least initially.

What is the Enforcement Workbook? An internal DCWP tool for evaluating whether a system qualifies as an AEDT and whether an employer's compliance documentation is adequate. The Comptroller's audit found DCWP had not been using it consistently.

Does OTI involvement change what counts as an AEDT? It does not change the legal definition. It changes DCWP's ability to correctly apply that definition to a specific tool's actual functionality, rather than taking an employer's or vendor's characterization at face value.

Should we expect more complaints-driven enforcement, or more DCWP-initiated reviews? DCWP has specifically committed to identifying potential non-compliance through research and leads, in addition to complaints, which points toward more DCWP-initiated activity than before.

Is there a way to know if our documentation would pass? The clearest test is whether your classification reasoning, audit methodology, and notice evidence are written down in enough detail that someone outside your company, with technical background, could follow and verify them.