Institutional blockages are holding back the abundance we need to solve our biggest challenges. The best response to these challenges is a market response, which requires clearing the bottlenecks suffocating commerce to unleash productivity at home and unlock our competitive edge abroad.

The stakes of these frictions are hard to overstate, while political will to fix them is weak and there are no policy silver bullets. Therefore, more than just incremental policymaking, progress demands a catalyst for rapid transformation: artificial intelligence (AI).

Invoking AI may give some readers pause. Ever since the ChatGPT revolution, AI hype has often run ahead of reality. Muddling the picture further are overimaginative sci-fi fantasists and opportunistic salesmen. Still, today’s AI is both real and remarkable. Its speed and ability to work at massive scale are well matched to the daunting scope of our institutional reform challenge. Applied with care, AI could help cut through backlogs, delays, and inefficiencies that stifle commerce.

To put AI’s potential to transform institutional frictions into perspective, let’s consider three high-impact opportunities where present-day AI could yield immediate, meaningful progress: regulatory reform, processing reform, and paperwork reform.

A well-oiled economy depends on a right-sized rulebook. When restrictions pile up, market dynamism shrinks. Firms saddled with excessive requirements are forced to raise prices, cut R&D, and redirect capital from innovation toward compliance. According to QuantGov, federal regulations contain over a million restrictions, so densely written that reading the federal code would take three and a half years. At that scale, comprehensive reform becomes nearly impossible.

Here AI can help. If modern AI excels at one thing, it’s analyzing text. While human analysts struggle to parse even a few pages of dense regulation, AI can plow through thousands—mapping connections, spotting redundancies, and flagging requirements for revision or removal.

This possibility is no pipe dream. Earlier this year, the City of San Francisco partnered with Stanford’s Regulation, Evaluation, and Governance Lab (RegLab) to build an AI tool to comb through nearly 16 million words of the city’s sprawling municipal code to identify obsolete reporting requirements—time-consuming written memos that are frequently redundant and pointless but still required by law. The result: Of the 528 mandated reports that surfaced, 140 were targeted for elimination.

The Code of Federal Regulations is over 100 million words, about six times larger than San Francisco’s code, but what worked in the Bay Area can work in Washington. Applied federally, AI could help agencies generate reform to-do lists, flagging outdated mandates, regulatory overlaps, confusing language, and the excesses ripe for removal. Reform would still depend on human approval and political will, but with AI the effort required could shrink dramatically, opening the door to the decisive action needed to unburden US commerce.

Beyond reducing the scope of rules themselves, AI has immense potential to reduce the burden of rule implementation by speeding up regulatory processing. Processing backlogs are among the heaviest public-sector burdens, and the most common. Anyone attempting to transition to Real ID has no doubt felt this pain. When the federal government unexpectedly decided to enforce the decades-old program, a flood of applications instantly inundated DMVs. Approval timelines stretched from weeks to months, leaving citizens unable to fly or even drive. Sadly, such delays are not an exception but often institutional business as usual.

Across agencies and processes, AI can lift such burdens. One opportunity lies in easing the rote analysis common across government approval processes. To help speed things up are “deep research” systems, promising new AI tools that can automate information discovery, summarization, and analysis. Already the US Patent and Trademark Office (USPTO) is exploring how such tools might give a speed boost to its commerce-essential approval backlogs. To speed the review of art submissions, the USPTO has deployed a “more like this” search tool that automatically surfaces similar art and relevant documentation. With the right information readily supplied by AI, no doubt time-to-approval will shrink.

A perhaps more important improvement lies in documentation-error handling. Simple mistakes are among the most common causes of bureaucratic delay: Confusing forms get misfiled, filled out incorrectly, or left incomplete. Such errors often push applications to the back of the line, compounding backlogs. At the Department of Energy, officials are now testing a new program called AI4IX that implements a grid-permitting AI concierge to automatically flag paperwork omissions, catch mistakes, and even manage the back-and-forth communication required for resolution. While this kind of technology may sound deceptively incremental, at scale the payoff can be dramatic. In Honolulu, a similar system was implemented to streamline the city’s residential permitting process. The result: Permitting completion times dropped by a remarkable 70 percent. Replicating that level of efficiency throughout our woefully backlogged energy permitting system could unleash projects that are currently stalled in red tape.

These examples only hint at AI’s potential to slash processing delays. While today’s tools mostly assist human-led processes, it is easy to imagine AI reducing frictions even more in the near future through streamlined, end-to-end bureaucratic workflows. To get there, however, requires the groundwork to digitize government forms and processes—groundwork, as it happens, that is perfectly suited to AI’s capabilities.

Americans spend about 10.5 billion hours filling out 106 billion government forms every year, according to a 2022 Chamber of Commerce analysis. A significant driver of this burden is that government paperwork is still just, well, paper. According to the General Services Administration, only 2 percent of federal forms have been digitized. While the private sector has long relied on digital workflows, federal processes remain stuck in the 1980s.

This is a staggering unforced burden. When forms exist only on the physical page, every government interaction moves significantly more slowly. Required information must be handwritten on designated forms. Time must be spent mailing or even walking those forms to the appropriate office. Finally, to be reviewed, each form must be handled individually and often manually re-entered into digital databases.

Once again, AI could help, this time by translating the physical into the digital. Consider Mistral OCR, a state-of-the-art “document understanding” system that can convert scanned paper forms into dynamic, digital formats. Beyond text, it can also digitize diagrams, formatting, and even graphs. Critically, it does so with 95 percent accuracy. With just a sprinkle of human review, today’s AI could finally shift federal paperwork to low-friction digital form.

The possibilities that follow are what make this digitization opportunity so important. Today, there is great interest in AI-powered compliance systems that can verify a company’s regulatory adherence while autonomously completing and submitting any required documentation. For heavily regulated sectors like cybersecurity, where surveys show that compliance consumes up to 50 percent of engineers’ hours, such tech could be a godsend. Just imagine what our economy could produce if AI unlocked 50 percent more engineering time. However, this future lives or dies by a prerequisite: Federal documents must be in a form AI can analyze and interact with. If government runs on paper, this obstacle-clearing automation is impossible.

Institutional frictions must be taken seriously. According to the American Action Forum, the federal paperwork burden alone costs between $276 billion and $422 billion each year—wasted capital, wasted time, wasted opportunity. Such costs cannot be allowed to persist. We must lean into our market system and rapidly remove the undue institutional bottlenecks that impede innovation and growth.

AI offers perhaps the most powerful tool at our disposal. But this technology requires adoption, a political process. Encouragingly, the White House’s AI Action Plan has already identified such adoption as a priority. What’s needed now is follow-through: Agencies must be empowered and encouraged to experiment, apply, and scale these tools so that solutions can move from pilot programs to high-impact reality.

The choice is stark: Continue letting outdated processes stifle potential gains, or leverage AI to unlock productivity improvements. If the United States adopts AI-driven reforms, it could clear barriers to more streamlined government processes. At the same time, these reforms would create an environment that would help free the real economy, fostering the innovation needed to jump-start the domestic productivity that drives growth.