Can The Federal Government Help States Finally Build Child Welfare IT That Actually Works?
Wonk PremiumBrief:
Can The Federal Government Help States Finally Build Child Welfare IT That Actually Works?
ACF's new Child Welfare Technology Incubator is the latest attempt. Here's what it would take to succeed.
By Kurt Heisler, Phd
In February 2026, the Administration for Children and Families announced the Child Welfare Technology Incubator, a new initiative to help states build modern child welfare IT systems, with Iowa as its first test case and Google Public Sector providing technical expertise.
The announcement was thin on specifics.
No budget. No staffing. No explanation of how this differs from what ACF was already doing before it eliminated the unit responsible for that work.
That unit — the Division of State Systems, which oversaw states' implementation of the Comprehensive Child Welfare Information System (CCWIS) — was cut in the spring 2025 HHS reorganization.
Whether its institutional knowledge was transferred, reorganized into capable hands, or simply lost remains unclear.
The Incubator may answer that question.
Whether it's a meaningful one depends entirely on whether it's designed to do something different in kind rather than degree from prior efforts.
WHAT BROUGHT US HERE
The introduction of CCWIS in 2016 was a genuine step forward. Its predecessor, the Statewide Automated Child Welfare Information System (SACWIS), incentivized monolithic builds by defining the system as a single comprehensive application with little room to flex.
CCWIS gave states more flexibility in how they could build, encouraged modular approaches, and signaled that the federal government understood something about modern software development.
And yet, a decade later, not a single state has a fully operational CCWIS. While CCWIS was a step forward in terms of frameworks, it may not have been a step forward in terms of outcomes.
California’s replacement project, once celebrated as a national model for agile, modular procurement, has become a $2 billion cautionary tale, stripped of its CCWIS designation by ACF in 2024 for missing milestones.
Michigan spent $231 million on a system a federal court later ordered independently assessed — an assessment I conducted, with a separate independent review subsequently reaching the same conclusions — which recommended building an entirely new system.
North Carolina spent nearly a decade and $120 million before pausing and starting over.
Zero fully operational systems. Billions spent. Caseworkers still working with systems that fall short of what they truly need.
I published my 2014 dissertation on the challenges of building an effective SACWIS. Twelve years later, very little has changed.
A FAMILIAR MODEL, WITH A KNOWN CEILING
In late 2015, GSA’s 18F, a unit that embedded technologists inside government to help agencies procure and build better software, partnered with California to do what the Incubator might try to replicate.
They helped California abandon a 1,500-page monolithic RFP in favor of a ten-page modular approach, pioneering a way of buying and building software in child welfare that favored small, iterative contracts over massive multi-year builds.
It was genuinely innovative, and the field watched closely.
Ten years later, California has retreated from that model entirely, reinstating Deloitte as systems integrator — the single-vendor dependency the whole effort was designed to avoid.
18F itself was eliminated by DOGE in March 2025.
What went wrong wasn’t the quality of 18F’s advice.
What went wrong is that planning and technical assistance, however well delivered, have a ceiling when the underlying structure doesn’t change.
The annual legislative budget cycle is incompatible with iterative development. Federal oversight, governed by the Advanced Planning Document (APD) process, grades states on schedule adherence, which directly penalizes the flexibility that modern software development requires.
California lost its CCWIS designation not because it stopped building, but because it missed milestones.
The procurement rules weren't the barrier — 18F had demonstrated they already permitted modular contracting.
The barrier was everything around them: the budget cycle, the oversight structure, and the institutional muscle memory that reasserted itself the moment external coaching ended.
The state was coached on how to be a better systems integrator without being given the sustained support needed to become one.
When the 18F team moved on, little of what they had modeled endured.
The Incubator arrives in 18F’s absence, at a moment of reduced federal capacity, into the same structural environment.
The question is whether it will be content to repeat the coaching model, or whether it will pursue the structural reforms that would make the coaching stick.
IOWA AS EXPERIMENT, NOT BLUEPRINT
Iowa is the first test of whether the Incubator is serious about any of this.
The state reportedly spent roughly two years defining system requirements and workflows before selecting a technology vendor. That is faster than typical, and a departure from how most states have approached procurement.
But Iowa’s planning success is a proof of concept, not yet a replicable model. It should be treated as such until the approach has been tested, documented, and shown to work elsewhere.
Iowa’s approach, if it holds, could demonstrate something the field has needed for a long time: that deliberate, structured planning before procurement produces better systems than the reverse.
But that is precisely what 18F tried to demonstrate with California, and California’s project ultimately collapsed under the weight of the structural conditions 18F never had the authority to change.
The underlying tension is that this was never about leadership commitment to creative approaches or vendor capacity to be innovative. Nor is it about whether Iowa's planning is sound.
It's whether anything about this arrangement — the partnership model, the federal posture, the state's own commitments — addresses the structural conditions that derailed California.
WHAT WOULD ACTUALLY MOVE THE NEEDLE
Three structural shifts would distinguish a meaningful initiative from a rebranded APD review process.
These are the signals to watch for.
MEASURE SUCCESS BY OUTCOMES, NOT FEATURES.
The federal APD framework currently measures progress through schedule adherence.
That is structurally incompatible with modern software development, where flexibility is a feature and plans are expected to evolve.
The Centers for Medicare and Medicaid Services (CMS) has moved its Medicaid IT oversight toward outcome-based metrics: is the system processing applications accurately, can it adapt when policy changes without requiring a full re-procurement, is it generating reliable data?
For child welfare, the same logic would mean things like measuring whether caseworkers can complete core workflows without workarounds, whether the system can measure what needs to be measured, and whether it can incorporate new program requirements without triggering a protracted vendor negotiation.
Redefining what progress means, in terms of outcomes rather than feature milestones, would do more for CCWIS modernization than any advisory service.
STATE CAPACITY AS A REQUIREMENT, NOT AN ASPIRATION.
California received some of the best planning help available and still couldn’t execute, because planning capacity and execution capacity are different things.
Methodology alone cannot fix a broken procurement and oversight environment.
What state capacity would look like is internal product owners and engineers who have the time, bandwidth, and authority to manage vendors and hold them accountable — people who understand what they are buying and can push back when delivery falls short.
Advisory help that arrives for the planning phase and then leaves has a known failure mode.
The goal should be an Incubator that makes itself unnecessary in every state it touches, because the capacity it helped build remains.
A FUNDING STRUCTURE MATCHED TO THE MISSION
The 50/50 CCWIS match rate creates real affordability barriers, but the deeper problem is what it signals: that child welfare IT is roughly equivalent in federal priority to any other state administrative expense.
CMS funds Medicaid IT development at 90 cents on the dollar and ongoing operations at 75 cents — enhanced rates Congress wrote explicitly into Section 1903(a)(3) of the Social Security Act. Child support enforcement systems receive similar treatment. Child welfare IT, under Section 474 of the same Act, gets the baseline 50%.
The funding disparity isn't administrative oversight. It's a deliberate legislative choice about relative priority, one that can only be undone by Congress, not by ACF through rulemaking. And the gap is larger than the headline rate suggests: Medicaid's enhanced funding comes paired with MITA, a federal architecture framework that actively rewards modular design. CCWIS asks states to build modular systems without the enhanced match that would make modular investment economically sustainable.
THE WORK AHEAD
The Incubator is not the first initiative to arrive with the right diagnosis. 18F had it too. What it lacked was the authority to change the structural conditions that made good planning insufficient.
The diagnosis has never been the hard part. Changing the budget cycles and oversight structures, the funding disparities, and the state-level institutional defaults that reassert themselves the moment outside coaching ends — that is the work.
Kurt Heisler Is The Founder And President Of Childmetrix, A Consulting Firm Specializing In Child Welfare Data, Technology, AI Integration, And Practice. Previously, Dr. Heisler Served In The Administration For Children, Youth, And Families As Acting Director Of The Office Of Data, Analytics, Research And Evaluation; Project Officer For The National Child Abuse And Neglect Data System (Ncands); And Senior Policy Advisor For Technology And Innovation.
