Why the IRS’s ten-year budget needed a one-year payoff

The week

For years, calling the IRS had been an exercise in patience that often ended with nothing at all. Most callers could not get through on the phone lines. During the pandemic, large backlogs of paper returns and correspondence piled up, envelopes that simply waited to be opened. Underneath it all were core systems that were old, including the “master files,” the agency’s central record of every taxpayer’s account, built decades earlier. The agency and its overseers had long described them as outdated.

On August 16, 2022, President Biden signed the Inflation Reduction Act. Among many other things, it gave the Internal Revenue Service roughly 80 billion dollars in additional funding over ten years. The money was split across four areas. Enforcement, meaning audits and collection, received about 45.6 billion dollars. Operations support, the behind-the-scenes infrastructure that keeps the agency running, received about 25.3 billion. Business systems modernization, the replacement of aging technology, received about 4.75 billion. Taxpayer services, the phones, letters and online tools that ordinary people touch, received about 3.2 billion.

The law also set aside 15 million dollars for a task force to study a free, IRS-run electronic filing system, with a report due to Congress within nine months. Such a system would let people file their returns directly with the government rather than through commercial software. Smaller amounts went to the Treasury Inspector General for Tax Administration, the Tax Court and Treasury’s tax policy office.

The path to that signature shaped what followed. The Senate passed the bill on August 7 through budget reconciliation, a procedure that avoids the filibuster, with the Vice President breaking a 50 to 50 tie. The House passed it on August 12 on a party-line vote. Money approved without votes from the other party was always likely to be a target if control of Congress changed, and the midterm elections were less than three months away.

For an agency used to tight annual budgets, this was a rare thing: a large, multiyear appropriation, meaning money Congress approves to be spent over a set period. Most of it was to remain available until September 30, 2031. Because it did not depend on the annual appropriations cycle, the IRS could in principle plan for a decade rather than a year.

The arithmetic behind the funding was explicit. The Congressional Budget Office estimated that the additional enforcement would bring in substantially more revenue over the decade than the funding cost. That made the money part of how the law was paid for, which tied the IRS budget closely to the politics of the bill as a whole.

The administration also tried to answer a fear before it took hold. In a letter that month, Treasury Secretary Janet Yellen directed that the new resources not be used to raise audit rates for households and small businesses earning under 400,000 dollars, relative to historical levels. The commitment shaped how the enforcement money could be spent and how the agency would later be judged.

But the funding was contested from the start, especially the share for enforcement. Opponents seized on an estimate, drawn from an earlier Treasury analysis, that the money could support roughly 87,000 additional hires over the decade, and portrayed them as new auditors. Fact-checkers noted that the figure covered all kinds of staff, including replacements for retiring employees. Opponents in Congress said they would seek to reduce the funding.

That put the IRS in an awkward position. It had a ten-year budget on paper and no guarantee that the money would still be there in year two or three. It also faced a leadership transition: the commissioner’s term was due to end that November, so the plan for spending the money would be finished under new leadership.

That tension, between a long horizon and a short leash, is what makes the week worth revisiting. It applies to any public body, and to many companies, that receives a large but reversible pot of money for transformation. How do you spend it so that something valuable is left standing, however much of it eventually arrives?

Consider a large public agency that learns it will receive a decade’s worth of extra funding to fix its technology and its service. It is the budget the agency has asked for, sometimes for years. The natural instinct is to plan a program to match: a long roadmap, foundations first, visible results in later years. That is how many large technology programs are drawn up, and on a whiteboard it looks sensible.

Large multiyear appropriations for modernization are rare in government, which makes that instinct stronger. The IRS funding showed the risk in it. Money appropriated by one Congress can be rescinded by the next, and a program that spends its early years on invisible foundations may lose its budget before anyone outside the agency sees a result.

The more durable approach treats a reversible appropriation as a series of tranches, each of which must be worth having even if the next never arrives. A tranche is a slice. The approach builds a house one usable room at a time, rather than pouring a foundation for a mansion and hoping the walls get funded. That principle sorts the budget into three kinds of spending. The first is service improvements that show up within a filing season. The second is multiyear system work that has to be cut into usable stages. The third is recurring commitments, mainly people, that temporary money should be careful about creating.

The sections below take those three in turn, after a short account of what modernization means inside a tax agency. They then cover how to run a program this way, how to report to funders, and how the same arithmetic applies outside the IRS.

What “modernization” actually means at a tax agency

The core and everything around it

First, what lies inside a tax agency, and why it is so hard to change.

At the center are the core processing systems. These hold the authoritative record of every taxpayer’s account: what they filed, what they owe, what they paid, what notices were sent. At the IRS these include the master files, which date back decades. The Government Accountability Office has described the Individual Master File as one of the oldest systems in the federal government. Much of it is written in assembly language, a low-level programming language few developers now learn.

“Legacy system” is the polite term for software like this. It still works, often reliably, but it was written in older programming languages, by people who have largely retired, for a world of paper and batch processing. Batch processing means the system collects work and processes it in large runs on a schedule, rather than updating instantly the way a modern banking app does. The IRS had already spent years on a program to move individual accounts to a modern database and daily processing. That effort showed how slow such a migration can be.

Around that core is everything else. There are the systems that read incoming returns and the tools that call center staff use. There are the websites and online accounts that taxpayers see, the case management systems for audits and collections, and a great deal of paper. Every one of these has to talk to the core. When people say an agency’s systems are “tightly coupled,” they mean you cannot change one piece without risking a knock-on effect somewhere else.

Why the calendar dominates

Modernization, then, is not one project. It is a long sequence of changes to a running machine that cannot be switched off, because tax season arrives every year whether the new systems are ready or not.

The calendar shapes everything. Systems are usually frozen in the months before and during filing season, because a failure in February or April affects tens of millions of people. That leaves limited windows each year for major changes. A program that slips past one window can lose most of a year.

Tax law itself adds to the load. Every new credit, deduction or reporting rule has to be programmed into the agency’s systems, often on a tight timetable. The Inflation Reduction Act was an example. Alongside the IRS funding, it created and changed clean energy credits, including new ways for some organizations to receive or transfer them. The IRS had to build new processes, including a registration system, to administer them. Modernization staff compete for time with the work of keeping up with Congress.

What the money could and could not buy

The split of the money mattered too. The business systems modernization line was the smallest of the four large pots. Operations support, which covers technology infrastructure among other things, was much larger. Much of what the public would experience as better service depended on technology funded from several lines at once.

The budget had to be spent in that context: an old core, a crowded calendar, competing demands from new legislation, and money divided across accounts with different purposes.

Year one: service people can feel

The quickest gains

The quickest gains are often the least glamorous. In the 2022 filing season, many taxpayers could not reach the IRS by phone at all, and millions of paper returns and letters waited to be opened. Hiring and training phone staff, adding callback options and clearing the paper backlog produce results within one season, and the public notices.

Year-one spending has its own trap. A phone agent who can answer the call but cannot see the case, change an address or release a hold only moves the backlog from the queue to the callback list. The early money works best when it pairs people with small tool changes that let them resolve issues on the first contact. Those include read access to the right systems, scripted authority for common fixes, and a way to route the rest to staff who can act. First-contact resolution, the share of problems fixed in a single conversation, is a better year-one measure than answer rate alone.

Consider a state revenue department that gets a one-off boost for customer service and hires fifty new phone agents. If those agents can only take messages, the department has bought a politer queue. Suppose that, alongside the hiring, it gives them a screen showing the status of a refund. It also gives them authority to fix a mistyped address on the spot. Then a large share of calls end with the problem solved. The headcount is the same. The outcome is entirely different.

The IRS put early money there, hiring thousands of customer service staff and adding callback options, and phone service improved noticeably in the 2023 filing season. That was valuable twice over. Taxpayers got answers, and the agency had evidence, in terms anyone could understand, that the money was doing something.

Measures that survive scrutiny

Early measures need to be honest to be useful. The IRS’s headline phone measure, “level of service,” had long been criticized by the National Taxpayer Advocate for describing only part of the picture. It counted calls on certain lines and did not capture how many callers were disconnected or gave up before reaching the queue. A program that reports a flattering metric invites a skeptical funder to discount everything else it reports.

A sturdier set of service measures would include several kinds of evidence:

• Total calls attempted, answered by a person and resolved, reported side by side.
• Average and worst-case wait times, reported together.
• The age of the oldest unprocessed paper return and correspondence, as well as the total count.
• The share of notices that taxpayers can answer online.
• Repeat contacts within thirty days, which show where first contacts failed.

Measures like these are harder to game and easier to defend. They also tell managers where the next tranche of money should go.

Paper as a service problem

Paper deserves its own line. A tax agency that receives returns and letters on paper must open, sort, transcribe and file them before anyone can act. When that pipeline backs up, refunds are delayed, notices go out on stale information, and taxpayers call to ask what happened. Phone volume rises because paper is slow.

Scanning paper at intake, and letting people send documents electronically, cuts that loop at its source. The benefit shows up in the same filing season, in fewer calls and faster refunds. It also produces digital records that later systems can use. Few investments serve both the near term and the long term so directly. Scanning also reduces the risk of misplaced documents, and of taxpayers being asked to send the same papers a second time.

The parallel elsewhere

For other agencies, the parallel is contact center capacity, callback technology, backlog reduction and simple status tools that tell people where their case stands. These are cheap relative to system rewrites. They also generate the measures that sustain a budget: wait times, backlog age and the share of contacts resolved on first try. Publishing those numbers regularly builds a record that is hard to argue with.

The mix differs by type of agency. A state revenue department may already run a commercial integrated tax system and can often add online services faster than the IRS. A state unemployment insurance agency, many of which received federal modernization grants after the pandemic, faces the same sequencing choice with a far smaller staff. A county benefits office may find its quickest gain in a document upload portal and a status page. The principle holds in each case: spend first where the public will notice within the year.

Multiyear systems: stages that each pay off

Why the big replacement is fragile

The second kind of spending is the slow, important work. Core tax processing systems hold the authoritative record of every taxpayer’s account. Replacing them takes many years and carries real risk of failure. Done as one large replacement with value only at the end, it is the part of a program most exposed to a budget cut, because a half-built system delivers nothing.

Large government technology replacements have a long history of overruns and cancellations. Part of the reason is scale. Part is that requirements change over a multiyear build, as laws and policies change. A design fixed in year one may be partly obsolete by year five.

What a useful stage looks like

The alternative is staging. Each stage should leave the agency with something it can use if funding stops. A modern data layer that copies records from the old master files can power better online accounts years before the old systems are retired. A data layer is a newer, more flexible copy of the information that modern applications can read easily, while the old system carries on underneath. Digitizing paper returns and correspondence at intake helps today’s processing and tomorrow’s systems alike. Letting people respond to notices online removes mail handling whether or not the core is ever replaced.

A useful test for any proposed stage has three parts. It should retire a manual process, power a service the public can use, or remove a dependency on the old core. A stage that does none of these is preparation, and preparation should be kept short.

The order of stages matters too. Work that reads from the old core, such as online account views, is lower risk than work that writes to it. A sensible sequence lets new systems read first, then handle narrow transactions, and only then take over the authoritative record for a category of accounts.

Buying in stages

Procurement has to match the staging. A single large contract for a complete replacement locks the agency into the full scope and makes a mid-course stop expensive. Federal acquisition rules already encourage modular contracting, which buys large systems in smaller, separately useful increments. Smaller contracts with clear deliverables let the agency stop, redirect or re-compete work as funding and priorities change.

That approach has its own costs. More contracts mean more procurement work, more integration between vendors and more coordination for the program office. Agencies need staff who can manage several vendors and own the architecture themselves, rather than handing it to one prime contractor.

The honest cost of staging

Staging has costs. Interim integrations with legacy systems are throwaway work in an ideal plan, and engineers dislike building them. An integration is the connecting plumbing that lets two systems exchange data; temporary plumbing is still plumbing, and it has to be built, tested and maintained. Running old and new side by side for longer adds operating expense. Those costs are the price of insurance against an abrupt stop, and program directors should present them that way rather than hide them inside the roadmap.

Identity as a stage of its own

Every online service depends on knowing who is on the other end. Earlier in 2022, the IRS had faced criticism for requiring facial recognition through a private identity provider to create online accounts, and in February it announced it would move away from that requirement. The episode showed that identity verification is part of modernization, not a detail to settle later.

A stage that improves identity proofing, offers more than one route for people who fail automated checks, and strengthens fraud detection pays off on its own. It makes existing online tools more usable and safer, and it is a prerequisite for anything built later. It also has trade-offs. Stricter checks reduce fraud but exclude more legitimate users, who then call or write instead. Agencies need to measure both the fraud stopped and the people turned away.

Pilots as stages

Free filing belonged in this category too. The study the law funded led to a pilot, Direct File, in 2024. A small pilot that can be measured and expanded is the staged approach in miniature, though it also showed how exposed a visible new service is to a change in policy.

A pilot is only a useful stage if it is designed to produce evidence. It needs a defined population, measures set in advance and a decision point. Those measures should cover usage, error rates, cost per return and taxpayer satisfaction. A pilot that cannot show results by its decision point is easy to cancel, whatever its merits.

Recurring commitments: people on temporary money

Costs that outlive the appropriation

The third kind of spending creates costs that outlive the appropriation. Permanent staff are the largest example. Hiring is slow in the federal government, training takes months, and staff hired on money that is later cut face layoffs or hiring freezes that undo the investment. Software subscriptions and new operating teams carry the same risk on a smaller scale. A cloud service or software license bought with one-time money still sends an invoice every year.

Matching the type of hire to the type of money

Agencies have a few tools here. They can hire into roles that the base budget could sustain if supplemental money disappeared, and use contractors or term appointments for peaks and one-time work. A term appointment is a government job with a fixed end date; in the federal civil service it can run for up to four years. That matches temporary money to temporary work. Tax agencies also have long experience with seasonal staff, who work during filing season and are released afterward.

They can prioritize people who build lasting capability, such as engineers who understand both legacy and modern systems, over roles that only make sense at the full funding level. And they can track, for every new recurring cost, which funding source pays for it in each future year.

Each option has a trade-off. Contractors can be released quickly, but they take knowledge with them when they leave. Term employees may leave early for permanent jobs elsewhere. Permanent hires are the most stable, but they are the hardest to unwind. A workforce plan for reversible money usually mixes all three, with the mix chosen deliberately rather than by whichever hiring route is fastest.

Training and attrition

New staff are not productive on day one. A customer service representative at a tax agency needs months of training on tax law, systems and procedures before handling complex calls. Experienced staff spend time coaching new ones, which reduces their own output in the short run. A hiring surge can lower productivity before it raises it.

Attrition compounds the effect. A large share of a long-serving public workforce may be eligible to retire within a few years. New hiring partly replaces departures rather than adding capacity. Plans that count gross hires, rather than net change in trained capacity, overstate what the money buys.

Labor relations

Much of the IRS workforce is represented by a union, the National Treasury Employees Union. Changes to roles, work methods and performance measures can require bargaining. That adds time to any plan that relies on new ways of working, such as giving phone staff broader authority to fix problems. Plans should build that time in from the start, and involve the union early, rather than treat it as an obstacle discovered late.

The funding map

The last step is a planning document, not a technology. It shows leadership and appropriators exactly what would stop if a rescission passed. A rescission is Congress taking back money it previously approved. Laying out its consequences in advance makes the effects of cuts concrete before they are made, which is a far stronger position than explaining them afterward.

The map should list every recurring cost created by the multiyear money: positions, licenses, cloud services, support contracts and facilities. For each, it shows the funding source in every future year and what happens if that source shrinks. Finance teams usually have the pieces; the work is assembling them in one place and keeping them current.

How tranche planning works

Planning each year as if it were the last

Turning the principle into a working plan is less exotic than it sounds. It mostly changes the questions asked at the start of each budget year.

In a conventional program, the plan is drawn as a single line from today to the finished state, and each year’s budget funds the next stretch of that line. In a tranche plan, each year’s budget is drawn as its own small program with its own finish line. The team is asked, in effect: if this were the last money we ever received, what would we build with it? The answer is then checked against the long-term destination, so that each self-contained slice also moves the organization in the right direction.

Take a county that receives a multiyear grant to replace a paper-heavy benefits system. A conventional plan might spend year one on requirements and procurement, year two on building the new core, and year three on moving cases across. A tranche plan might instead spend year one on an online form and document upload that cut mail handling immediately. Year two might bring a status tracker that answers the most common phone question, and year three would move the first category of cases to the new core. Both plans head to the same place. Only one leaves the county better off if the grant is cut after the first year.

Governance that follows the tranches

The governance changes too. Each tranche gets a short business case that names what it will deliver, how that will be measured and what it depends on. Steering committees review tranches against those measures, not against percentage-complete figures on a long roadmap. And the finance team keeps a running view of which costs are one-off and which recur, so the question “what if the money stops?” always has a current answer.

The federal government already has a working model of this kind of discipline. The Technology Modernization Fund, created in 2017, lends money to agencies for technology projects. It typically releases funds in increments tied to milestones and expects agencies to repay. That structure forces each phase to justify the next.

The base-budget trap

One risk is easy to miss. Multiyear money and annual appropriations are separate, but they interact. If annual funding for an agency’s base operations is held flat or cut, the agency may have to use the multiyear money to keep existing services running. The transformation budget then turns, without anyone deciding it, into an operating subsidy.

A tranche plan should track this directly. It should show how much of the multiyear money is going to new capability and how much is replacing base funding. When the second share grows, funders should be told plainly, because it changes what the money can deliver. The same view helps internally. Program managers can see when their transformation budget is being drawn down for running costs, and can escalate before the planned deliveries slip.

Talking to the people who hold the purse

From advocacy to reporting

The next piece is communication with funders, whether that means appropriators, a grant agency or a corporate board. Organizations often treat these conversations as advocacy: asking for more and defending what they have. A tranche approach turns them into reporting. Each report covers what this year’s money bought, how that was measured, what next year’s money would add and what would stop without it.

That framing has two advantages. It makes cuts harder to impose casually, because their consequences are specific and public. And it builds trust, because funders see results in the same terms that taxpayers or customers do. An agency that can say “wait times fell and the backlog shrank” is having a very different conversation from one that says “the platform is forty percent built.”

A reporting calendar

Reports work best on a fixed schedule that matches how funders decide. For a federal agency, that means results in hand before appropriations hearings and before the budget request is built. For a grant recipient, it means aligning with the grantor’s reporting periods. For a company, it means results before each budget cycle and board meeting where the program’s funding is reviewed.

A fixed schedule also disciplines the program. Teams know when results are due and plan deliveries around those dates, rather than around internal milestones only they can see.

Oversight bodies as allies

Independent oversight helps when it is used well. Inspectors general, auditors and taxpayer advocates publish their own assessments of how money is spent. An agency whose own reporting matches what the auditors find gains credibility. One whose claims are contradicted by an audit loses it quickly.

Sharing measures and data with those bodies early, and responding to their findings in public, makes them a check that funders trust. It also surfaces problems while they can still be fixed.

The case against front-loading

The strongest objection to this approach is that it can starve the foundations. Visible quick wins are easier to fund and to report. A program run on tranches may keep putting off the core replacement, because each year a nearer, more visible project looks better. The result can be a polished front end on a failing core.

That risk is real, and it is the reason the staging test matters. A good tranche plan includes core work in every year, cut into pieces that each retire something. It does not postpone the core; it refuses to make the core all-or-nothing.

A second objection is political. Visible services attract attention from people who oppose them, as Direct File later showed. Staging cannot remove that exposure. It can only ensure that what was built before a reversal still has value, in data, in skills and in evidence, after it.

A third objection is cost. Interim integrations, parallel running and more frequent procurement add up. In a world of guaranteed funding, a single well-run replacement might be cheaper. Few public programs have that guarantee, and the extra cost buys protection against a far larger loss.

Hospitals, cities and boards face the same arithmetic

The pattern outside tax

A hospital system receives a large grant for digital transformation. A city wins a multiyear federal award for broadband or permitting systems. A company’s board approves a three-year technology budget that the next CEO may cut in half. In each case, the money is real but conditional, and the program’s survival depends on what it can show before the conditions change.

State and local governments learned a version of this from pandemic relief. The American Rescue Plan’s State and Local Fiscal Recovery Funds had to be obligated by the end of 2024 and spent by the end of 2026. Governments that used the money for recurring costs, such as permanent positions, faced a cliff when it ran out. Those that used it for one-time investments faced fewer hard choices later.

How the risk differs by organization

The source of reversibility differs, and so does the right response.

• A federal agency faces rescission by Congress and shifting annual appropriations. Its best defense is public, measurable results and a funding map that shows exactly what a cut would stop.
• A state or county often relies on federal grants with fixed deadlines and conditions. Its main risk is the cliff at the end, and its main tool is keeping recurring costs off one-time money.
• A health system may fund digital work from grants, capital budgets or operating margin. Margin pressure can cut a program mid-stream, so each phase should improve clinician or patient experience on its own.
• A bank or credit union usually funds technology from its own budget. The reversal comes from a new CEO, a merger or an earnings shortfall, and staging protects against all three.

What is at stake

The stakes run through every part of the organization. Money is the obvious one: a half-finished system is often worth nothing, so a cut mid-program can turn spent dollars into a write-off. Risk follows close behind, because long programs with no interim results are where large technology failures tend to happen. Customers, or citizens, feel the difference between a program that improves service in its first year and one that asks them to wait. Staff hired for an ambitious plan carry the human cost when it is reversed. And for public bodies, credibility with legislators and the public is itself a resource, and an agency builds it by pointing to shorter waits rather than to a roadmap.

Signs of a well-sequenced program

A well-sequenced program tends to have a few recognizable features.

• Every tranche stands alone. If funding stopped after this year, the organization would still have something useful to show for it.
• Visible measures, published on a schedule. Wait times, backlog age and first-contact resolution are reported regularly, including when the news is bad.
• Honest staging costs. Interim integrations and parallel running are budgeted openly as insurance, not buried.
• Recurring costs mapped to funding. Every new hire, license and operating team is matched to the budget line that will pay for it in each future year.
• Small pilots before big launches. New services start at a scale that can be measured, adjusted and either grown or stopped.

Weaker programs show the reverse. Programs spend the first two years on architecture and procurement with nothing the public can see. Teams treat interim work as wasted effort and skip it, leaving the organization all-or-nothing. Leaders hire permanent staff against temporary money because hiring is slow and they want to start early. And success measures focus on activity, such as calls answered or systems procured, rather than outcomes, such as problems solved.

What each year of spending leaves if the money stops

• If our funding stopped at the end of this year, what would we have to show for what we spent?
• Which improvements will customers or citizens notice within twelve months, and how will we measure them?
• How is our system replacement cut into stages, and what does each stage retire or enable on its own?
• For every new permanent role or subscription, which budget pays for it in each future year?
• If a cut of a quarter or a half arrived next year, what exactly would stop, and have we told our funders?

With hindsight

The funding did not survive intact. The Fiscal Responsibility Act of 2023, the debt ceiling law signed that June, rescinded about 1.4 billion dollars of the IRS money immediately. A side agreement redirected about 20 billion dollars more, which the 2024 spending law carried out. Further cuts were proposed, and the IRS workforce shrank in 2025. The Congressional Budget Office scored the rescissions as increasing federal deficits, because the enforcement they cut was expected to collect more than it cost. The ten-year budget behaved exactly like the reversible money it always was.

The midterm elections changed control of the House. In January 2023, the first bill the new House majority passed would have rescinded most of the new IRS funding. The Senate did not take it up, but the vote made the political risk plain.

The plan itself arrived later than the six months Treasury had originally asked for. The IRS published its Strategic Operating Plan in April 2023, under a new commissioner confirmed the month before. It set out five objectives, from improving service and resolving problems quickly to focusing enforcement on complex, high-dollar noncompliance, modernizing technology and data, and building the workforce. The task force report on direct filing reached Congress in May 2023.

Some early gains held. Phone service improved in the first filing seasons after the law, more notices could be answered online, and paper scanning expanded. In 2023 the IRS announced a paperless processing initiative intended to let taxpayers send correspondence and respond to notices digitally. These were the unglamorous, near-term improvements: easier to deliver, easy for the public to notice and, once in place, hard to undo.

The newer, more visible service had a rougher path. Direct File ran as a pilot in 12 states in 2024 and expanded to 25 states in 2025. It was reported later in 2025 that it would not continue for the 2026 filing season. A service the public can see is also a service that policy makers can see, and a pilot’s fate can turn on decisions far outside the program team’s control. What the pilot left behind, in usage data, design work and evidence about cost, remains on the record either way.

Leadership changed repeatedly. The commissioner who launched the plan left at the start of 2025, and the agency went through several leaders that year. Each change of direction tested whether the program’s early pieces could stand without their original sponsors.

What did not hold was the premise that the law had bought a decade of planning certainty. Within ten months of signing, part of the money was gone, and the larger redirection followed within two years. What did hold was the value of improvements already delivered. A scanned letter, a faster refund or an answered call did not need next year’s funding to stay useful.

Core modernization continued in parts, more slowly than planned. That is the typical fate of long-horizon work funded by money that does not last. It does not stop entirely, but it stretches out, and the pieces designed to stand alone are the ones that make it.

By 2026 the IRS case shows the sequencing argument at work. The improvements taxpayers saw early were the easiest to defend, and the long-horizon work depended on money that did not last. Anyone handed a large, multiyear budget that someone else could reverse should assume the money might stop at any point, and spend it in an order where each year leaves something worth keeping.