AI’s real promise for government: Shortening the distance from data to decision

Artificial intelligence is often discussed in government as a tool for automation, one that helps agencies do the same work faster or with fewer resources. That framing understates its real potential.

The most important impact of AI in the public sector may not be automation at all. It is the ability to shorten the time between data becoming available and leaders gaining the insights they need to act. In other words, AI’s real promise lies in improving time to insight.

For governments responsible for complex programs affecting millions of people, the speed of decision-making matters. Whether responding to emerging challenges, evaluating programs, or analyzing regulations, policymakers must turn vast amounts of information into decisions that shape service delivery, outcomes and people’s lives. Historically, this process — from analysis to action-ready insight — has been measured in months or even years.

AI is beginning to change that.

Across the public sector, AI-enabled tools are helping compress the cycle from data ingestion to insight to action. Tasks that once required hours of manual document review, synthesis, or data exploration can now be completed in minutes. But the real value is not simply faster processing. It is giving leaders more time to interpret findings, weigh trade-offs and make better decisions.

This shift suggests that governments should rethink how they measure the value of AI. Traditional outputs, such as reports, dashboards and other analytic products remain essential tools for decision-makers. But AI’s impact should also be evaluated by how much it reduces the lag between evidence and action, helping agencies move from insight to implementation more quickly.

Consider a few emerging examples.

At the Centers for Medicare & Medicaid Services, AI is helping staff respond faster to hospitals seeking guidance on complex quality measures. An AI-powered tool developed by Mathematica and Telligen generates draft responses to hospital inquiries regarding the Severe Sepsis and Septic Shock measure — an assessment of whether hospitals follow evidence-based protocols — and integrates them into CMS’s workflow so experts can review and finalize answers more efficiently. 

The approach helped reduce inquiry turnaround time by 35%, increased the share of questions resolved within the same month by 18%, and reduced reopened inquiries by 2% — signs that hospitals are receiving clearer, more complete guidance.

AI is also accelerating the policy development process itself. In the annual rulemaking cycle for Medicare’s Inpatient Prospective Payment System, which governs how hospitals are reimbursed for inpatient care, analysts must review thousands of pages of regulatory text and compare changes across years. Generative AI tools can now perform those comparisons in minutes rather than hours, allowing experts to focus on interpreting policy implications rather than manually tracking changes across thousands of pages.

Even in research and convening work, the gains are significant. In a recent initiative involving 18 national symposia on the future of health services research, AI-assisted transcription and synthesis tools enabled Mathematica researchers to analyze discussions and identify cross-cutting themes within days, compressing what would traditionally take weeks of manual synthesis.

AI can also help agencies manage the growing scale and complexity of public programs. In supporting CMS oversight of Medicaid Section 1115 demonstrations, Mathematica has developed AI-assisted workflows that help analysts review lengthy monitoring and evaluation documents against federal guidance and prior examples. These tools help teams review complex reports more consistently and rapidly, freeing experts to focus on interpreting findings and advising policymakers.

Taken together, these examples illustrate a broader shift: AI is beginning to function as decision-support infrastructure for government.

But faster insights alone are not enough.

If AI accelerates the production of insights without strengthening governance, transparency and analytic rigor, it risks creating a new problem: decisions made more quickly but not necessarily more wisely.

Responsible adoption therefore requires embedding AI within well-designed analytic workflows. Human experts must remain accountable for interpreting results. Models must be transparent about their limitations. Agencies must establish clear governance structures to ensure AI-assisted insights are reliable, reproducible and aligned with policy goals.

It’s also important that AI help leaders cut through complexity. Government executives already face overwhelming volumes of information. AI systems should help synthesize evidence, highlight uncertainties and surface decision-relevant insights, not generate more noise.

When deployed thoughtfully, AI can help governments operate at the speed of modern challenges while preserving the rigor that public policy demands.

For decades, the public sector has invested heavily in collecting data. The next frontier is ensuring that data translates into insight — and insight into action — fast enough to matter. That could mean identifying emerging risks in near real time, accelerating regulatory analysis that once took hours, or synthesizing evidence across programs in days rather than weeks.

If we want AI to deliver real value for government, we should judge it by a simple question: How quickly can they close the gap?

Time to insight may soon become the most important metric in public-sector innovation.

Paul Decker is president and CEO of Mathematica.