In the private sector, product teams pick their customers, generate demand, and ship into something close to a green field. In the public sector, none of that holds. Ayushi Roy — Chief Program Officer at New America's New Practice Lab and a lecturer at the Harvard Kennedy School — joins Randy to unpack what changes when your user base is already sitting in front of you, your scrutiny is congressional, and the right answer is sometimes to delete ten systems rather than build an eleventh.
Drawing on her work on IRS Direct File, the Children's Health Insurance Program (CHIP), the Illinois childcare voucher system, and a text-based 911 alternative that rolled out to 800,000 students across 13 universities, Ayushi makes the case for a distinct public-sector product playbook: thin-slicing for safe failure, designing for the lowest digital denominator, separating design problems from engineering problems, and treating unbuilding as a first-class option.
Chapters
Key Takeaways
Public-sector product starts with the curb cut, not the early adopter. Civil servants don't choose their customers. Success isn't ROI — it's reaching outcomes across the full public, fringes included.
Demand already exists. The job is lowering barriers. There's no product marketing in the public sector. People are already struggling to claim what they're owed. Replace demand generation with burden reduction.
Build for the lowest digital denominator. Test on the devices your users actually have — $20 flip phones and library PCs, not the newest MacBooks. Procurement that flatters the team can blind the build.
Thin-slice to make failure survivable. Direct File rolled out internally to IRS staff first, then by tax-scenario complexity rather than household income — building internal sponsorship and creating safe ground to fail on.
Treat the service, not the product. Buy-in, legacy systems, colours of money, oversight reporting and analog workflows aren't adjacent to public-sector product — they are the job.
Separate design problems from engineering problems. Policymakers often frame design problems (open-ended outcomes) as engineering problems (reverse-engineerable solutions), forcing compliance-driven builds instead of customer-focused ones.
Close the loop between policy designers and policy implementers. Train the next generation to be bilingual across both disciplines — the FAFSA simplification crisis is what happens when they aren't.
Sometimes the answer is unbuilding. A request to build a 15th CHIP system became a programme to remove ten. Ask candidates: what's something you built that you wish you never had?
