The Open Secret of Government IT: A Decade of Failure, Sold as Service

The Number Nobody Wants Said Out Loud

Roughly 70% of government IT and transformation projects fail. That's not a contrarian take — it's the consensus across the only people whose job is to count: McKinsey, the Standish Group, GAO, KPMG, BCG. The Standish CHAOS data is even uglier when you isolate the public sector: government tech projects over $6 million succeed only 13% of the time. NIST has documented average cost overruns on government software at roughly 80%.

The federal government spends more than $100 billion a year on IT. The vast majority of that money doesn't buy new capability — it props up legacy systems already in place, many of them older than the analysts maintaining them. GAO put "Improving IT Acquisitions and Management" on its government-wide High-Risk List back in 2015. In the 2025 update, the area didn't just stagnate — it regressed. Of 1,881 recommendations GAO has issued in this category since 2010, 463 are still open. The VA's electronic health record modernization, a flagship program, has reached 6 of more than 160 facilities as of late 2024.

These aren't anomalies. They are the operating model.

And almost nobody loses their job over it.

"Accountability" Is a Word They Use, Not a Thing They Do

Here is what actually happens when a $200 million state ERP modernization implodes:

1. The vendor produces a "lessons learned" deck.
2. The agency CIO accepts a leadership role at a different agency.
3. The project gets quietly rebaselined, rescoped, or rebranded.
4. A new RFP goes out. Often to the same vendors. Sometimes to a subsidiary of the same vendor.
5. A press release calls it a "transformation."

Nobody refunds the taxpayers. Nobody is removed for cause. Nobody clawed back a bonus. The hearing room theater plays for a news cycle, the GAO publishes another report, and the cycle resets.

Compare this to the private sector. A founder who burns $200 million on a product that doesn't ship doesn't get reassigned to "Director of Strategic Initiatives." They lose the company. Their investors lose patience in months, not years. The market is a brutal accountability mechanism — flawed, but real. Government has manufactured an environment where that mechanism doesn't exist, then acts surprised when the predictable thing keeps happening.

Leadership That Couldn't Survive a Real Hiring Loop

A meaningful share of senior IT leadership in government couldn't pass a technical screen at a mid-sized SaaS company. That isn't a slur; it's an observation that anyone who has sat on both sides of the table can verify in a single afternoon.

The selection mechanism rewards tenure, political navigation, and procurement fluency — not technical judgment. So you get directors and deputy directors who:

• Cannot read an architecture diagram closely enough to spot a single point of failure.
• Cannot evaluate whether a vendor's proposed stack is overkill, underkill, or theater.
• Cannot tell the difference between a real proof-of-concept and a slideware demo.
• Cannot challenge a $30 million change order because they don't understand what it bought.

These same people draw six-figure salaries, generous defined-benefit pensions, full healthcare, and step increases tied to nothing. In the private market, where compensation tracks the value you can actually create, many of them would be out of work inside a quarter. They know it. The institution insulates them from that knowledge having any consequence.

This isn't an attack on public service. There are extraordinary public servants — engineers, analysts, program managers — doing first-rate work for less money than they could make outside. The problem isn't the people who can do the work. It's the layer above them that can't, won't, and never has to.

The Vendor Capture Problem

The "Big Integrator" ecosystem — you know the names — has spent thirty years perfecting a business model that depends on government failure. The model works like this:

1. Bid a fixed-price project at an aggressive number.
2. Use change orders, scope clarifications, and "discoveries" to triple it.
3. Throw bodies — often offshore — at requirements that were never properly defined.
4. Deliver something that technically meets the contract but doesn't solve the actual problem.
5. Win the follow-on contract to fix it.

Members of Congress have started saying this out loud. In a 2025 Oversight hearing on the GAO High-Risk List, Rep. Nancy Mace observed that federal agencies are spending billions a year on software without a comprehensive understanding of what they're purchasing — duplication on top of duplication. That isn't a bug in the procurement system. It is the system, working exactly as the incumbents have shaped it.

Vendors didn't get there alone. They got there because their counterparts inside the agencies — many of them ex-vendor employees, many of them future vendor employees — let them. The revolving door isn't a metaphor. It's a hiring pipeline.

Requirements: The Original Sin

Every failed government IT project I've ever seen up close had the same fingerprint: requirements artifacts that weren't actually requirements. They were wish lists, policy paraphrases, cut-and-pasted RFP language, or flat refusals to engage with the operational reality of the workflow.

You cannot build software for a process nobody has documented honestly. You cannot automate a workflow when the people who run it won't tell you how it actually runs — only how it's supposed to run on paper. And you cannot replace a 1990s mainframe with a modern platform when nobody has the authority, or the courage, to say "this rule the system enforces hasn't matched the law in eleven years."

This is where the worker side of the problem enters. The line staff in many agencies — protected by union contracts, vested pensions, and workloads that would be considered part-time in the private sector — have a clear, rational, self-interested reason to obscure how the work actually gets done. If the process is documented, it can be streamlined. If it's streamlined, the headcount justification disappears. If the headcount disappears, the union dues disappear, the political base disappears, and the comfortable equilibrium disappears.

So the requirements interviews go nowhere. The "subject matter experts" describe the system as a series of exceptions. The vendor builds for the exceptions. The exceptions multiply. The project dies. The workers' jobs are safe.

This is not a conspiracy. It doesn't need to be. It's just incentives.

The Two-Tier Economy Inside Public Service

There are working Americans putting in 50-hour weeks at small businesses with no pension, no defined-benefit anything, healthcare costs that eat their raises, and zero protection if their employer has a bad quarter. Their tax dollars are funding salaries, benefits, and pensions for a class of public employees whose accountability structure they would not recognize as employment at all.

I'm not anti-union. Unions exist for legitimate reasons, and public-sector workers deserve fair treatment. But "fair treatment" and "structurally insulated from the consequences of poor performance for decades on end" are not the same thing, and pretending otherwise has produced exactly the system we now have: one where the people most protected from market discipline are also the people closest to the broken processes nobody is allowed to fix.

Why AI Is the Threat They Can't Procure Their Way Out Of

Here's why the next decade is different.

For thirty years, the diagnostic capability — the ability to look at a process and say "this is what's actually happening, this is what it should cost, this is who is adding value and who is overhead" — has been concentrated in expensive consultants and the people running the process. The consultants were captured. The people running the process had every reason to lie.

AI changes the cost curve of that diagnostic to roughly zero.

A modern model, properly applied, can:

• Ingest a year of process logs, ticket data, and email threads and produce an honest map of what work actually flows through a workflow.
• Compare a written policy to the system's enforced behavior and surface every place they've drifted apart.
• Read a 400-page RFP and a vendor's response side-by-side and flag the change-order ambushes hiding in the ambiguity.
• Estimate, within a useful margin, how many FTEs a process should require if it were designed today — and how many it currently consumes.
• Generate the requirements artifacts that the agency was supposed to produce but never did.

None of this is science fiction. The technology to do it is sitting on every laptop. The reason it isn't being applied at scale to government operations is not technical. It's political. It exposes too many people. It threatens too many headcounts. It eliminates the information asymmetry that the current equilibrium depends on.

That is exactly why it will be resisted — and exactly why it is the most important opportunity to bring real accountability to public spending in a generation.

The Resistance Is Already Underway

Watch what's happening already. AI use is being slow-walked, study-grouped, ethics-reviewed, and pilot-committeed into irrelevance inside agencies that should be deploying it aggressively. The stated reasons — bias, security, hallucination, privacy — are real concerns and real cover. The unstated reason is that an AI honestly applied to most agency operations would produce a report that ends a lot of careers.

So expect the playbook:

• "We need a comprehensive governance framework first." (Translation: years of delay.)
• "Only certified vendors with FedRAMP High can provide AI services." (Translation: only the same incumbents that broke everything get to build the thing that would expose them.)
• "We need a human in the loop for every decision." (Translation: we will continue making the decisions we were already making, and label the AI as advisory.)
• "AI cannot replace the judgment of experienced civil servants." (Translation: please do not measure us.)

Each of these statements contains a kernel of legitimate concern. Each is also being weaponized to preserve a status quo that has cost taxpayers somewhere between hundreds of billions and several trillion dollars over the last two decades, depending on which improper-payment and project-failure estimates you stack up.

What Actually Has to Happen

If you want this to change — and the public, by overwhelming majorities across both parties, says it does — the levers are not subtle:

Tie executive compensation in agencies to delivered outcomes, not occupied chairs. Public-sector pay for IT leadership should look more like the private sector: meaningful at-risk components, real consequences for cost and schedule miss, no pensions accruing on cancelled programs.

Require independent technical evaluation of every IT project over a threshold — by people who don't work for the prime vendor or the agency. The current "independent verification and validation" market is itself captured. Replace it with rotating panels of practitioners who have shipped real software.

Default to AI-assisted process discovery before any RFP is written. If an agency cannot produce a current-state process map that an AI built from real operational data agrees with, the agency is not ready to procure. Period.

Publish post-mortems with names attached. Project failures should be a matter of public record, not a "lessons learned" deck behind a login screen. The same way the private sector publishes earnings.

Sunset legacy systems on a schedule, not on consensus. The current incentive is to keep the old system running indefinitely because nobody wants to be the person who turns it off. Make turning it off the default and require active justification to keep it running.

This Pattern Is Not Confined to IT

Everything described above — the protected class, the vendor capture, the undefined requirements, the workers whose livelihoods depend on dysfunction, the resistance to honest measurement — is the same pathology that runs through other corners of the administrative state where accountability has been deliberately engineered out of the system.

The clearest parallel is family court. There, instead of civil service rules and pensions, the institutional protection is absolute and quasi-judicial immunity: judges, guardians ad litem, and custody evaluators are statutorily shielded from civil liability for almost anything they do in the course of their work. Instead of the Big Integrators capturing federal IT procurement, it is the family law bar capturing family court — the same revolving door, the same self-reinforcing professional ecosystem. Instead of "requirements artifacts that aren't actually requirements," there is the "best interests of the child" standard — an intentionally undefined criterion that gives unappealable discretion to actors with no obligation to justify their reasoning. Instead of federal IT procurement dollars distorting incentives, it is Title IV-D federal incentive payments distorting state enforcement behavior — $25.8 billion paid to families in FY2024 against $6.6 billion in program costs, with federal reimbursement that has no ceiling and rewards enforcement volume rather than enforcement accuracy.

The Bottom Line

A 70% failure rate, sustained for a decade, across hundreds of billions of dollars of taxpayer money, would end any private business and possibly land its leadership in front of a grand jury. In government, it has produced a generation of people who confuse longevity with competence and salary with worth, protected by institutional structures specifically designed to prevent the kind of accountability the rest of us live with every day. We the people deserve more than an uneducated thought around what a dim mind may think are just fields on a screen. There's a reason Elon said they'd kill him if he actually fixed the problem. He was right — where is the revolution?

I have less to lose. America has more to gain. I'll take that stand — for our kids' sake. And the Trump coin — don't tie your sovereignty to the government. Haven't you learned your lesson yet? Tie it to SQUEIL. Make the government squirm — establish sovereignty — get your liberty back. It's as simple as capturing the flag for the correct reasoning.

---

The current system cannot fix itself — that is the entire argument above. The SQUEIL Constitution is one attempt at a different frame: append-only records, cryptographic tenant sovereignty, published recovery objectives, and stakeholder rights that cannot be amended away by the people they constrain. Policy change starts where unaccountable discretion ends. A new government could learn sound IT architectural principles from a sovereign we the people who will forge together to build an economy safe from tyrannical rule.

— FREEEE Ye Ye. Long live Kurt Cobain.

---

Sources for the figures cited above: U.S. Government Accountability Office, High-Risk Series GAO-25-108125 (Feb 2025); GAO duplication and cost-savings reports (May 2025); Standish Group CHAOS reports; McKinsey/Oxford large-scale IT project studies; NIST government software cost-overrun studies; PGPF and CAGW analyses of FY2024 improper payments ($162 billion); House Oversight Committee 2025 hearing record on the GAO High-Risk List.

---

Legal Notice & Disclaimer

Editorial Opinion and Commentary. This essay is a work of editorial commentary, criticism, and policy analysis on a matter of substantial public concern. The arguments, characterizations, and conclusions expressed herein are the personal opinions of the author, Patrick Roden, in his individual capacity. They do not necessarily represent the formal positions of On Top Home Services LLC, SQUEIL LLC, DevForge Academy, or any affiliated entity. Reasonable readers may disagree with any or all of the views expressed.

Sources and Public Record. Statistical figures, agency reports, congressional hearing references, and policy facts cited in this essay are drawn from publicly available sources, including the U.S. Government Accountability Office, the Standish Group CHAOS reports, NIST, McKinsey and Oxford research on large-scale IT project outcomes, the Peter G. Peterson Foundation, Citizens Against Government Waste, and the public record of congressional oversight proceedings. The author has made a good-faith effort to characterize these sources accurately as of the date of publication. Statistics, recommendations, and policy facts may have changed since publication; readers are encouraged to consult the original sources directly.

No Defamation Intended; Critique of Systems and Classes. This essay critiques institutional structures, professional classes, and incentive systems — not specifically named individuals. References to general categories (the "Big Integrators," senior IT leadership, line staff, civil servants, the revolving door, the protected class) are categorical descriptions of groups operating within the federal IT procurement system and are not directed at, and should not be understood to refer to, any specific identifiable person. The author acknowledges and explicitly states that there are extraordinary individual public servants — engineers, analysts, program managers — performing first-rate work within these systems; the critique is of the layer of institutional structure that insulates poor performance from consequence, not of public service as such. No statement in this essay is intended to assert or imply that any specific identified individual has engaged in unlawful, unethical, or professionally negligent conduct.

Public-Figure References and Paraphrase. Where elected officials or public figures are referenced by name in this essay, they are discussed solely in their capacity as public figures speaking on matters of public concern, and the discussion is editorial commentary protected under New York Times Co. v. Sullivan, 376 U.S. 254 (1964), and its progeny. The reference to Rep. Nancy Mace's remarks at a 2025 Oversight hearing is a paraphrase characterization of statements made on the public hearing record, not a verbatim quotation. The reference to a statement reportedly made by Elon Musk regarding consequences of attempting to address federal waste is a paraphrase of public commentary, characterized as opinion, and is not represented as a verbatim quotation; readers seeking the original statement should consult the source. The reference to the "Trump coin" is a metaphorical commentary on the broader theme that ordinary citizens should not tie their economic sovereignty to political branding or government-adjacent instruments of any party; it is not a statement of fact about, or directed at, any specific person, product, or organization.

Forward-Looking Statements; SQUEIL. Statements about the design, intent, or future capabilities of SQUEIL — including any description of platform behavior, sovereignty guarantees, the SQUEIL Constitution, append-only record structure, cryptographic tenant isolation, recovery objectives, or stakeholder rights — are forward-looking and reflect current architectural intent. Actual implementation, capabilities, terms, and protections will be governed by the definitive operating agreements, terms of service, privacy policy, and constitutional documents in effect at the time any tenant, implementer, or stakeholder enters a relationship with the platform. Forward-looking statements involve risks and uncertainties; nothing in this essay should be relied upon as a guarantee of future outcomes.

Not an Offer or Solicitation. This essay does not constitute an offer to sell, or a solicitation of an offer to buy, any security, equity interest, token, membership unit, subscription, or other financial instrument in SQUEIL LLC, On Top Home Services LLC, DevForge Academy, or any affiliated entity. Any such offer or solicitation will be made only by means of definitive offering documents and only to qualified persons in compliance with applicable federal and state securities laws.

Not Legal, Financial, or Investment Advice. Nothing in this essay constitutes legal, tax, financial, accounting, or investment advice. Readers should consult their own qualified professional advisors before making any decision based on the views or information contained herein. Discussion of federal procurement law, civil service rules, public-sector pension structures, and federal IT funding mechanisms is general informational commentary, not legal advice for any specific situation.

Trademarks. SQUEIL™, the Coronel, the SPEED Methodology, Canon, Forge, Arbiter, Worker, and the Sovereign Tribune are marks and architectural designations of SQUEIL LLC. On Top Home Services is a registered Oregon LLC (CCB #244467). Other marks referenced are the property of their respective owners and are used for descriptive and comparative purposes only.

Copyright. © 2026 Patrick Roden. All rights reserved. The text of this essay (excluding quoted, attributed, or referenced third-party material) is the original work of the author. Brief quotation with attribution is welcomed; substantial reproduction or redistribution requires written permission.