Field Digital Agenda Draft 0.1
← All priorities

Priority 02

Redesign services and reduce burdens with digital tools and AI

Simplify whole services, redesign human processes, and use digital tools and AI where they reduce burden and improve outcomes.

This priority advances
Right-sized procedure Workforce

The case

Change the work, not just the tool.

An AI tool can make a task faster without making a service better. If staff still chase the same documents, move information between disconnected systems, or wait for the same approvals, the bottleneck may simply move. The larger opportunity is to ask which work is necessary, who should do it, and what a good outcome requires.

Suggest an edit

Right-sizing procedure does not mean abandoning public protections. It means distinguishing a legal obligation from the inherited routine used to meet it, then designing a less burdensome way to achieve the underlying purpose. Frontline staff, program leaders, lawyers, technologists, and the people using a service need to make those choices together.

Suggest an edit

What the agenda asks

Remove the burden. Redesign the service.

  1. 01

    Start with the burden, not the AI use case

    Map a real service from the resident’s need to its resolution. Identify repeated information requests, handoffs, reports, and approvals; distinguish what can be removed now from what needs a policy change. Make simplification a funded part of modernization.

    Suggest an edit
  2. 02

    Redesign and test the whole workflow

    Give a multidisciplinary team authority to change roles, procedures, and tools together. Compare AI-assisted work with a simpler non-AI approach, and measure total time to resolution, staff effort, errors, and resident experience, including work pushed downstream. Plan for new pressures too: AI-generated submissions can increase the volume and complexity of work government must assess. Test how the whole service handles that demand.

    Suggest an edit
  3. 03

    Make protection part of how the service works

    Build appropriate human judgment, accessible alternatives, privacy safeguards, and ways to correct mistakes into the workflow. Test with affected staff and residents, monitor performance after launch, and stop or change tools that do not improve outcomes. In fraud prevention and program integrity, measure both detection and errors that delay or deny legitimate access, and make review and correction usable.

    Suggest an edit

Across the field

Selected examples

These illustrate the priority, not a census. They include proposals, pilots, current resources, and completed work with lessons to reuse. Sources were reviewed on the dates shown; status may have changed. Inclusion implies neither endorsement of the Agenda nor a funding recommendation. Use “Suggest an edit” on a card to correct its status or description.

01
Stanford RegLab

Research and policy development

Using AI to find reporting requirements worth retiring

Stanford RegLab is working with New York, California, and Maryland to inventory reporting requirements and examine what agencies produce, what it costs, and whether it is useful. The work uses AI to make accumulated procedure visible so officials can decide what to simplify or remove.

Why selected: Makes inherited reporting burdens visible before deciding what to remove.

Activity: A cross-state effort to examine the work required, not just automate report writing

The effort examines reporting work; the signal does not quantify burden reduced.

Suggest an edit
Right-sized procedure

Learn more

RegLab: How AI is helping states cut through decades of red tape

Sources reviewed

02
Nava Labs

Piloting and evaluation

Open-source AI tools built around caseworker workflows

Nava Labs has released its Caseworker Empowerment Toolkit after testing benefit-navigation tools with government and community partners. Its tools support policy lookup, referrals, document processing, and form filling, with caseworkers reviewing information before acting. Public code and documentation give other teams a starting point for adaptation.

Why selected: Offers reusable code to test against real caseworker workflows.

Activity: Toolkit released in April 2026, with public demos, code, and pilot results

Pilot results are linked in the sources. Wider adoption and durable service improvements need separate assessment.

Suggest an edit
Right-sized procedurePurpose-fit digital infrastructure

Learn more

Nava: Toolkit release and pilot background Caseworker Empowerment Toolkit

Sources reviewed

03
USDR Arizona DES

Piloting and evaluation

A human-reviewed AI translation workflow for public benefits

U.S. Digital Response and Arizona’s Department of Economic Security co-designed an AI-assisted translation workflow using staff input, resident research, a SNAP glossary, reusable prompts, and evaluation. The pilot combines automation with human review and prepares the agency to maintain the approach itself.

Why selected: Pairs AI assistance with staff review and agency stewardship in a benefits workflow.

Activity: A SNAP translation pilot that pairs reusable tools with staff review

The pilot demonstrates an approach to test; broader rollout and service outcomes are not established here.

The source describes a pilot; a planned rollout was not confirmed in this review.

Suggest an edit
Right-sized procedureWorkforce

Learn more

USDR: Arizona language-access pilot

Sources reviewed

04
Public Policy Lab Civic Futures

In development

Making benefit notices easier to produce and understand

Public Policy Lab is developing AI-assisted SNAP notices with New Mexico and Oregon, supported by Center for Civic Futures’ Public Benefit Innovation Fund. The project targets the manual work of updating program notices, with clearer, more actionable communications as the intended result.

Why selected: Tests whether redesign can reduce notice-production work and improve communication.

Activity: Two-year development award announced in December 2025

Development funding is an input. Notice comprehension and benefit access are outcomes to assess.

The award documents development, not delivered notices or measured results.

Suggest an edit
Right-sized procedure

Learn more

Center for Civic Futures: Projects receiving two-year awards

Sources reviewed

05
FAS

Research and policy development

Redesigning permitting work alongside new technology

The Federation of American Scientists is translating permitting practitioners’ experience into proposals for stronger interdisciplinary teams and better workflows. Its 2026 Permitting Corps proposal connects digital and AI skills with project management and scientific expertise, and calls for sharing effective tools and processes across agencies.

Why selected: Connects permitting tools with the people and processes needed to use them.

Activity: An August 2026 proposal informed by permitting practitioners’ experience

This is a proposed approach; implementation and service effects have yet to be established in this example.

Suggest an edit
Right-sized procedureWorkforce

Learn more

FAS: We Need a U.S. Permitting Corps

Sources reviewed

What could happen next

Possible next moves

These illustrate ways to advance the priority, including development, adoption, and sustained operation. They invite discussion and support; they are not an announced grant program.

Adoption and sustained operation

Readiness: Exploring

Help governments adopt a durable product operating model

Work with a small number of government teams on a service they choose, improving delivery while establishing public product ownership, decision rights, mixed career and temporary roles, live-work coaching, and targeted procurement changes. Document what remains after outside support and what another government would need to adapt the approach.

What would help

Find a few government teams with a live service problem, then design time-bound support with existing providers and practitioners that leaves ownership and evidence inside government. Explore this opportunity

Development and testing

Readiness: Scoping

Prove a reusable public system through child welfare

Explore a working open-source Comprehensive Child Welfare Information System (CCWIS) with a state partner as a test of reusable public systems. Pair software with changes to agency practice and policy, then publish the code, decisions, costs, and adoption playbook for a second state to test what truly transfers.

What would help

Bring together child welfare agencies, families and caseworkers, delivery teams, federal partners, and funders to test demand, scope a first deployment, and plan long-term stewardship and a second adoption. Explore this opportunity

Development and testing

Readiness: Exploring

Build hands-on AI capacity inside congressional committees

Explore embedded technical practitioners working with congressional committee staff on actual legislative and oversight tasks. Pair model access with institutional context, independent evaluation, and a plan for staff to retain the ability to judge and use these tools after a placement ends.

What would help

Find a committee-defined task and bring staff, embedded practitioners, and independent support together to design and evaluate a first placement. Explore this opportunity

Development and testing

Readiness: Exploring

Make privacy rules executable

Explore a bounded privacy-as-code demonstration in one real government data-sharing workflow. Translate human-approved rules on authority, purpose, access, minimization, and retention into testable controls, then publish safe tests and evidence of how they work. Encoding a rule also exposes legal and policy interpretations that people must resolve.

What would help

Find an agency data-sharing use case and bring together legal, privacy, program, engineering, and oversight expertise to design and test a first implementation. Explore this opportunity

Help move this priority

See a gap or a way forward?

Tell us about a need this priority overlooks, even if you do not have a project to propose. Or add work already underway, with a public source. Editors review suggestions before publication.

Tell us what’s missing

Add a project