GAO’s 2016 legacy IT report made a dated plan the test for old systems; its request to OMB stayed open
On May 25, 2016, congressional auditors reported federal systems decades old and asked for dated modernization plans. Agencies wrote theirs one by one; as checked in November 2026, the request that OMB direct every agency to identify legacy systems was still open.
Looking Back revisits past weeks one at a time, up to the most recent, and asks what it meant for the banks, health systems and public agencies we serve. The series was written in 2026, with the benefit of hindsight, and each article says which week it looks back on.
Two-minute summaryThe dek and the first sentence of each section, quoted from the article
On May 25, 2016, congressional auditors reported federal systems decades old and asked for dated modernization plans. Agencies wrote theirs one by one; as checked in November 2026, the request that OMB direct every agency to identify legacy systems was still open.
- The week On Wednesday, May 25, 2016, the Government Accountability Office (GAO) released a report titled Information Technology: Federal Agencies Need to Address Aging Legacy Systems. [1] Its number was GAO-16-468.
- The spending signal: what the operations share measured GAO said operations and maintenance was 68 percent of the federal IT budget in fiscal year 2010.
- The oversight gap: how at-risk investments escaped review Agency CIOs rate major investments on the federal IT Dashboard.
- The obsolescence evidence: age was one signal among several GAO reported agency-stated ages, and its notes qualified them.
- The plan as the control: dates, activities, functions The test is modest.
- How the findings apply beyond federal agencies GAO’s report covered federal agencies.
- Objections to the argument, taken in order GAO’s spending data support the money argument: development and modernization funding fell while operations rose. [1] But GAO also found at-risk investments with requested funding and only general plans. [2] Money without a plan does not produce a replacement, and a fund still needs a list of what to fund first.
- Checks a CIO could run These checks apply GAO’s test to any organization’s portfolio.
- With hindsight As checked on November 16, 2026, GAO’s page for the 2016 report listed the status of all 16 recommendations. [1] Three were listed as open: the two to OMB and Treasury’s recommendation to plan legacy modernization.
The week
On Wednesday, May 25, 2016, the Government Accountability Office (GAO) released a report titled Information Technology: Federal Agencies Need to Address Aging Legacy Systems. [1] Its number was GAO-16-468. The same day the office published testimony, GAO-16-696T, summarizing the report for Congress. [3]
The report was addressed to the chairmen of the Senate Committee on Homeland Security and Governmental Affairs and the House Committee on Oversight and Government Reform. [2] It had three objectives: to assess federal agencies’ IT operations and maintenance spending, to evaluate oversight of at-risk legacy investments, and to assess the age and obsolescence of federal IT. [2]
The auditors reviewed OMB and 26 agencies’ IT operations and maintenance spending for fiscal years 2010 through 2017. [1] It looked more closely at the 12 agencies that reported the highest planned IT spending for fiscal year 2015. [1] They asked those 12 to list their three oldest systems; where agencies did not track system ages, the auditors asked for their three oldest investments. [2]
The headline finding was about money. The report said the federal government spent about 75 percent of the total amount budgeted for IT for fiscal year 2015 on operations and maintenance. [1] It said that spending had increased over the past 7 fiscal years, which resulted in a $7.3 billion decline from fiscal years 2010 to 2017 in development, modernization and enhancement. [1] It said 5,233 of the government’s approximately 7,000 IT investments were spending all of their funds on operations and maintenance. [1]
The second finding was about age. The report said agencies reported using several systems with components that were, in some cases, at least 50 years old. [1] It reported that the Department of Defense used 8-inch floppy disks in a legacy system that coordinates the operational functions of the nation’s nuclear forces. [1] It said the Treasury used assembly language code, a language first used in the 1950s and typically tied to the hardware it was developed for. [1]
The report’s table of examples listed Treasury’s Individual Master File and Business Master File, each with an agency-reported age of about 56 years. [1] It described the Individual Master File as the authoritative data source for individual taxpayers, where accounts are updated, taxes assessed and refunds generated. [1] The auditors recorded that neither had specific, defined plans with a firm date for replacement. [1]
The table also listed Veterans Affairs systems written in COBOL, a Justice Department inmate system, and Social Security’s Title II systems. [1] The report said the Title II investment, which determines retirement benefits eligibility and amounts, comprised 162 subsystems written in COBOL. [1]
The report tied the problem to policy. It said OMB had drafted guidance that would require agencies to identify, prioritize and plan to modernize legacy systems. [1] It said that until the policy was finalized and fully executed, the government ran the risk of maintaining systems that had outlived their effectiveness. [1]
The report made 16 recommendations, two of them to OMB. [1] [2] One asked OMB to commit to a firm date for issuing its draft legacy guidance and then direct agencies to identify legacy systems needing modernization or replacement. [1] Twelve agencies were asked to identify and plan to modernize or replace legacy systems as needed, including time frames, activities and functions to be replaced or enhanced. [1] [2]
The report’s highlights said nine agencies agreed with the recommendations, two partially agreed and two had no comment. [1] Defense and Energy were the two that partially agreed; the report said their plans were consistent with the intent of its recommendations. [1]
The 2016 legacy report’s most vivid detail was the floppy disks. The image was vivid and accurate. [1] It also pointed attention at the wrong variable. GAO’s own remedy did not turn on how old a system was. It turned on whether anyone had written down when, how and in what order the system would be replaced. [1] [2]
Our argument is that GAO treated a dated modernization plan for each system as the unit of control, and that this is where the federal effort stalled. Agencies wrote plans one by one, and GAO closed every agency recommendation but one. [1] The government-wide step, OMB directing agencies to identify legacy systems needing attention, was still listed as open when we checked. [1]
We take the report in four parts:
- The spending signal: what the operations and maintenance share measured, and what it could not tell anyone.
- The oversight gap: how at-risk investments escaped the reviews that existed, and why general plans did not count.
- The obsolescence evidence: age, language, hardware, vendor support and skills, and how GAO weighed them.
- The plan as the control: what a specific, defined plan had to contain, and what OMB’s draft guidance was meant to require.
The parts are separate questions. The first is about money at portfolio level, the second about review processes, the third about the condition of individual systems, and the fourth about the instrument GAO proposed. After them we look at how the findings apply to organizations other than federal agencies, take the objections in order, and end with checks a CIO could run.
The spending signal: what the operations share measured
The numbers GAO reported
GAO said operations and maintenance was 68 percent of the federal IT budget in fiscal year 2010. It said agencies planned to spend 77 percent of their IT funds on it in fiscal year 2017. [2] It said 15 of the 26 agencies increased their operations spending from fiscal year 2010 to 2015. [2] Agencies also reported planned spending of nearly $55 billion in fiscal year 2015 on what OMB called non-provisioned IT. [1]
Non-provisioned IT was OMB’s term for spending that did not use solutions viewed as more efficient, such as cloud computing and shared services. [1] GAO said OMB had developed a metric for this spending but had not identified a goal for it, which could limit agencies’ ability to evaluate progress. [1]
What the share could not show
A rising operations share is a symptom. It can mean old systems are expensive to keep alive. Agency officials gave GAO other reasons: major systems moving from development into operations as development finished, and improved reporting of spending that had been booked as development. The share alone does not say which cause is at work. [2]
GAO’s own table illustrates the point. The largest operations line in fiscal year 2015 was the Medicare Management Information System at $4,381.0 million. [2] GAO’s note said this represented the federal share of state Medicaid systems’ cost. It added that HHS said it did not manage those assets or control how the money was spent. [2]
In our reading, that note is a caution against reading the top-line share as a modernization score. The single largest operations line in the report funded systems that, by HHS’s account, a federal agency did not manage.
How spending gets classified
The operations share depends on how agencies label spending. OMB collects it through each agency’s IT Portfolio Summary, which identifies all IT investments, and a Major IT Business Case for each major investment. [2] Spending is split between operations and maintenance and development, modernization and enhancement.
GAO reported a revealing exchange on that split. OMB officials said that where agencies are unsure whether to report costs as operations or development, they default to operations. [2] According to OMB, agencies tend to categorize investments that way because operations attract less oversight, require less documentation and carry a lower risk of losing funding. [2]
In our reading, that explanation changes how the headline number should be read. Part of the rising share may reflect accounting incentives rather than older technology. A modernization effort booked as maintenance also escapes the reviews that development work receives.
The low-development tail
GAO looked at the other end of the split. It said 5 of the 26 agencies reported spending less than 10 percent on development, modernization and enhancement in fiscal year 2015. [2] They were Housing and Urban Development, the Interior, the Environmental Protection Agency, NASA and the Army Corps of Engineers. [2] GAO said 51 of 151 bureaus spent less than 10 percent on development. [2]
Agency officials gave reasons. They cited bureau size and mission, and several bureaus having recently finished major development work that had moved into operations. [2] An OMB policy analyst told GAO that a low development share could indicate maintenance costs were reducing an agency’s flexibility. [2] Both readings can be true; GAO’s later work looked at systems rather than ratios.
Why costs rise
OMB staff identified several factors behind rising operations spending, including two that GAO set out. Maintaining legacy hardware costs more over time. [2] And programmers who know older languages are becoming increasingly rare and thus more expensive, according to OMB staff. [2] Both reasons are about the condition of specific systems, which a portfolio ratio cannot show.
Why GAO asked for a goal
GAO’s first recommendation to OMB was to identify and publish a specific goal for the non-provisioned spending measure. [1] The logic was managerial. A metric without a target cannot tell an agency whether it is moving fast enough. [1] The recommendation did not prescribe the goal.
What the money signal did
The spending figures framed the policy debate around funding. GAO noted that in April 2016 the IT Modernization Act had been introduced in the House, which would establish a revolving fund of $3 billion to retire, replace or upgrade legacy systems. [2] In our reading, the money framing was accurate but incomplete. A fund pays for a modernization. It does not decide which system goes first or prove the plan is real.
The oversight gap: how at-risk investments escaped review
The risk ratings
Agency CIOs rate major investments on the federal IT Dashboard. GAO found that CIOs at the 12 selected agencies reported 23 of their 187 major operations investments as moderate to high risk as of August 2015. [2] The agencies requested $922.9 million in fiscal year 2016 for those investments. [2]
Of the 23, agencies had plans to replace or modernize 19. GAO said the plans for 12 of those were general or tentative, lacking specificity on time frames, activities or functions to be replaced or enhanced. [2] Agencies did not plan to modernize or replace 4 of the 23. [2]
The Dashboard behind the ratings
OMB deployed the IT Dashboard publicly in June 2009. GAO said it displayed cost, schedule and performance information on over 700 major federal IT investments at 26 agencies. [2] Agencies submit CIO ratings that, under OMB’s instructions, should reflect the level of risk facing each investment. [2]
GAO had made over 20 recommendations in earlier years to improve the accuracy and reliability of Dashboard information. [2] A rating system with known accuracy problems is a weak trigger for reviews. The 2016 report showed what happens when the trigger fails for systems in steady operation.
Two review tools
Two tools existed for such investments. The first was the TechStat, a face-to-face meeting to terminate or turn around IT investments that are failing or not producing results. [2] GAO said the Federal CIO began leading TechStats in January 2010, and OMB empowered agency CIOs to hold their own by June 2012. [2]
The second was the operational analysis. GAO said OMB directs agencies to monitor operations investments through operational analyses performed annually. These assess costs, schedules, whether the investment still meets customer and business needs, and performance. [2]
How often the tools ran
GAO’s background section noted that in June 2013 it had recommended OMB require agencies to address high-risk investments, and OMB generally agreed. [2] It then reported that as of October 28, 2015, OMB had conducted only one TechStat review in the prior 2 years. [2] In our reading, a review tool used once in two years at the center cannot be the main safeguard for thousands of operating investments.
What GAO found
GAO said agencies generally conducted the required operational analyses but did not consistently perform TechStats on the at-risk investments. [2] Although TechStats were not required, agencies had performed them on only five of the at-risk investments. [2] TechStats were not performed on three of the four investments rated high risk. [2]
Operational analyses were not conducted for four at-risk investments, two of them Treasury infrastructure investments. [2] GAO recommended that Commerce and Treasury ensure required analyses were performed on investments in operations. [1]
Why ratings drifted
Agencies gave reasons for skipping TechStats. GAO said officials reported that several investments’ risk levels had been lowered in the months since the Dashboard was publicly updated. [2] GAO noted that the public portion of the Dashboard is not updated during formulation of the President’s budget. [2] One HHS investment did not meet its internal TechStat criterion of a cost variance over 10 percent. [2]
In our reading, this is the core of the oversight gap. Review tools triggered on cost and schedule variance. A system running on budget, on unsupported software, without a replacement date, produces no variance. The tools were built for projects, and legacy systems are not projects until someone makes them one.
Why operations are hard to review
Development projects have baselines: a planned cost, a planned schedule and planned capabilities. Variance against those baselines is what review tools measure. An operating system has a budget line and a service level, and as long as it meets both it shows no variance.
OMB’s operational analysis was meant to fill that gap by asking each year whether an investment still met business and customer needs. [2] GAO found agencies generally performed them, yet several at-risk investments still lacked specific plans. [2] In our reading, the analysis asked the right question but did not require the answer to be a dated decision.
What the law had added
GAO noted that recent legislation required agencies to review each major investment rated high risk for four consecutive quarters. [2] It said OMB’s associated guidance required a TechStat when an investment had been rated high risk for three consecutive months. [2] Both triggers depend on the rating. A system whose risk rating is lowered, or never raised, never reaches them.
The obsolescence evidence: age was one signal among several
Age
GAO reported agency-stated ages, and its notes qualified them. Age was reported by agencies, and systems may have components newer than the reported age. [1] GAO also said not all agencies tracked systems and their ages in the same way, and some had to approximate initiation dates. [2]
That qualification matters. A system can be 56 years old in its business logic and recent in its hardware. GAO said the Individual Master File’s hardware had been upgraded to a more modern IBM mainframe, while its code remained assembly language. [2]
Language and hardware
GAO pointed to two kinds of obsolescence. One was software written in languages such as assembly and COBOL. [1] The other was hardware no longer made or supported, as with the floppy disks in the Defense command and control system. [1] The table recorded that Defense planned to update data storage, processors and terminals by the end of fiscal year 2017. [1]
Vendor support
GAO reported that each of the 12 selected agencies used unsupported operating systems and components in their fiscal year 2014 reports under the Federal Information Security Management Act. [2] It said Commerce, Defense, Transportation, HHS and VA reported using 1980s and 1990s Microsoft operating systems that stopped being supported more than a decade earlier. [2]
This is the obsolescence signal with the clearest security consequence. GAO said lack of vendor support created security vulnerabilities and additional costs. [2] Unlike age or language, it can be checked against a vendor’s published support dates.
Security exposure
The security signal grew in GAO’s later work. In 2025 GAO said seven of the 11 systems it selected were operating with known cybersecurity vulnerabilities. [5] It said the Environmental Protection Agency’s system had obsolete hardware with known vulnerabilities that could not be remediated without modernization. [5]
That last phrase marks the point where a legacy question becomes a security question. Patching and compensating controls can manage a vulnerability for a while. When the fix requires replacing the platform, the security team’s risk acceptance and the modernization plan become the same document. In our reading, an organization should be able to show both.
Skills
GAO said legacy systems may become more expensive as agencies pay a premium to hire staff or contractors who can maintain outdated systems. [2] It reported that SSA had re-hired retired employees to maintain its COBOL systems. [2]
In our reading, the skills signal can force a decision sooner than the others. A system can run on old code for decades if the people who understand it remain available. When they retire, the cost of a change, even a small one, rises.
The profiles
GAO profiled 12 legacy investments or systems in an appendix, one per agency, chosen because they were among the agency’s oldest or had been rated at risk. [2] One was the National Weather Service Telecommunication Gateway. GAO’s table recorded that Commerce planned to retire it in fiscal year 2017 and replace it with a new system. [2]
That example shows how the signals combine in one system. The gateway was rated high risk, had both a TechStat and an operational analysis, and had a dated plan. [2] It was the case GAO’s framework was built to produce: a known risk, reviewed, with a date for its end. GAO later recorded Commerce’s statement that it had retired the gateway, as noted under Hindsight. [1]
How GAO combined the signals
The 2016 report did not score systems on a composite index. It reported examples and agency-reported ages, and paired each with whether a specific plan existed. [1] [2] GAO’s 2025 review added a scoring method, discussed under Hindsight. The 2016 report’s pairing of a condition with the presence or absence of a plan is the thread that runs through all of them.
The plan as the control: dates, activities, functions
What counted as a plan
GAO’s tables used one test: whether the agency had specific, defined plans for modernization or replacement. [1] Where it found only general or tentative plans, it said the agency had not provided time frames, activities to be performed, or functions to be replaced or enhanced. [2] Those three elements reappear word for word in the recommendation to the 12 agencies. [1]
The test is modest. It does not ask whether the plan is funded, approved by a board, or technically sound. It asks whether the plan names when, what work and which functions. In our reading, that modesty is the point. A plan with those three elements can be tracked by someone outside the program.
Examples from the report
GAO’s examples show the difference. Veterans Affairs planned to replace its time and attendance system with a Human Resources Information System Shared Service Center in 2017, which counted as specific. [1] Its benefits system had general plans to roll capabilities into another system, with no firm time frame, which did not. [1]
Of the ten oldest systems listed, GAO said agencies reported plans to modernize or replace each. It said the plans for five were general or tentative. [2] Being among the oldest did not ensure that a system had a specific plan.
What each element does
The three elements serve different readers. A time frame lets a budget office and an oversight body test progress. A description of activities lets an engineering lead judge whether the time frame is realistic. A list of functions to be replaced lets the business owner see what changes for users and what will be retired. [1]
Missing any one weakens the plan in a specific way. Without dates, nothing can be late. Without activities, the dates cannot be tested. Without functions, a replacement can go live while parts of the old system keep running, because nobody listed them.
GAO’s later reviews added a fourth idea, the disposition of the legacy system. [4] [5] In our reading, that addition closes the gap the third element leaves open. Listing functions to replace does not say when the old system is switched off and its data archived or migrated.
Plans and budgets
The 2016 report did not require plans to be funded. It treated funding as a separate issue, noting the proposed modernization fund. [2] In our reading, a dated plan is still the prerequisite for funding. A fund that pays for modernization needs proposals, and a proposal needs the dates, work and functions GAO asked for.
Who owns a plan
GAO’s recommendations went to department heads, who were to direct their CIOs to identify and plan. [1] In our reading, the addressee matters. A legacy plan that replaces business functions needs the program owner’s agreement on what the new system must do, and the budget owner’s agreement on when it is paid for. A plan written only by IT staff can meet GAO’s three elements and still lack both agreements.
The examples in the report bear this out. SSA’s Title II modernization was underway, but GAO said one effort was experiencing cost and schedule challenges due to the complexities of the legacy software. [1] Having a plan did not remove the difficulty. It made the difficulty visible, which is what oversight needs.
OMB’s draft guidance
GAO said agencies were, in part, maintaining obsolete investments because they were not required to identify, evaluate and prioritize their operations investments. [2] That decision is whether each should be kept as-is, modernized, replaced or retired. [2]
According to OMB staff, OMB had created draft guidance to require agencies to identify and prioritize legacy systems needing replacement or modernization. [2] GAO said the guidance was intended to develop criteria for identifying the highest priority legacy systems, evaluate and prioritize agencies’ portfolios, and develop modernization plans. [2] The draft included time frames for those steps, but OMB had not committed to a firm date for issuing it. [2]
Why GAO asked for a date
GAO’s recommendation to OMB was procedural: commit to a firm date for the guidance, then direct agencies to identify legacy systems needing modernization or replacement. [1] It did not ask OMB to choose systems itself. In our reading, GAO asked OMB to do for the guidance what it asked agencies to do for systems: put a date on it.
The agencies’ answers
The agencies’ comments showed the range of positions. Defense partially concurred and said it had modernized, upgraded or retired hundreds of systems through an investment review process. [2] It said it would continue under existing policies, consistent with OMB’s draft guidance to the extent practicable. [2] Energy partially concurred and said it would continue to modernize as needed and as funding is available. [2]
Commerce concurred and said it was already replacing and modernizing systems as needed within budget constraints. [2] OMB, through an e-mail from its Office of E-Government on May 12, 2016, concurred with GAO’s recommendations. [2]
In our reading, the comments show a pattern in this report. Agencies agreed with the goal and tied action to existing processes and available funding. That is reasonable, and it is also why GAO kept asking for dates. A commitment to modernize as funding allows has no date until the funding decision is made.
How the findings apply beyond federal agencies
GAO’s report covered federal agencies. Its test, a dated plan with defined work and functions for each old system, travels. The differences lie in who sets the rules and who pays.
A federal agency
For a federal CIO, the report’s instruments were the IT Dashboard rating, the operational analysis and the TechStat. [2] The practical lesson is that those tools do not fire on their own for a stable legacy system. A CIO who wants a legacy decision has to put the system on a review calendar regardless of its variance.
A federal CIO’s authority
The report noted that Congress had enacted FITARA in December 2014, which among other things required OMB to develop standardized performance metrics and report cost savings to Congress quarterly. [2] GAO also said OMB’s instructions call for CIO ratings that reflect each investment’s risk. [2] For a federal CIO, those are the levers that exist. The report’s finding is that none of them, on its own, forced a dated decision on a stable old system.
A state Medicaid agency
GAO’s largest operations line was the federal share of state Medicaid systems’ cost. [2] GAO noted that every state must implement a claims processing and information retrieval system to support its program. [2] A state agency running such a system answers to its own budget process and to federal funding conditions. The 2016 test still applies: a plan with dates, work and functions is what lets anyone else check progress.
A county or city
A county running an old tax, court, permitting or payroll system may have a small IT staff to maintain it. The skills signal GAO described, paying a premium for people who can maintain old code, can bite harder at that scale. [2] A county that cannot name who maintains its oldest system has an answer to the skills question already.
A federal contractor
Contractors that operate government systems hold much of the knowledge GAO’s skills signal points to. GAO reported agencies paying a premium for staff or contractors able to maintain outdated systems. [2] A contractor running a legacy system for an agency can help the agency meet the plan test, by documenting functions and dependencies that only its staff know. That documentation is the input to the functions element of any plan.
A bank or credit union
A bank’s core processing platform may be decades old in its logic. The federal budget rules GAO cited do not apply to them. The transferable piece is the plan test: for each core or ancillary system, is there a date, a description of work and a list of functions to be replaced.
A state IT department
A state that runs shared infrastructure for many agencies faces GAO’s classification problem in another form. When agencies pay a central department for hosting, the cost of an old application can look like a service charge. The plan test applies to the application, whoever runs the servers. The owner of the business function still needs a date, work and a list of functions.
A health system
A hospital that keeps an older system running for laboratory, imaging or device integration may find that replacement depends on a vendor’s roadmap. The vendor-support signal GAO reported for federal agencies, unsupported operating systems, has the same security consequence in a clinical network. [2] A health system’s plan has to name the vendor’s committed dates as well as its own work, because the hospital cannot replace the component alone.
Objections to the argument, taken in order
1. Money was the binding constraint, not plans
GAO’s spending data support the money argument: development and modernization funding fell while operations rose. [1] But GAO also found at-risk investments with requested funding and only general plans. [2] Money without a plan does not produce a replacement, and a fund still needs a list of what to fund first.
2. Old systems that work should be left alone
Some legacy systems are stable and well understood. GAO did not argue that age alone required replacement; its recommendation said to modernize or replace legacy systems as needed. [1] The plan test accommodates a decision to keep a system, if the decision is recorded and dated. What GAO objected to was the absence of the decision. [2]
3. Agency-reported ages are unreliable
GAO said so itself: ages were reported by agencies, some were approximated, and components may be newer. [1] [2] That weakens rankings by age. It does not weaken the plan test, which does not depend on age at all.
4. A government-wide policy adds paperwork
Agencies did write plans without the OMB policy, as the closed recommendations show. [1] But GAO’s later work found plans still incomplete, and recommended that Congress consider requiring them. [5] A common requirement is what makes plans comparable across agencies.
5. The headline images distorted priorities
The floppy disk example was accurate and memorable. [1] The risk is that attention goes to the oddest technology rather than the systems that matter more. GAO’s 2025 review scored systems on 16 attributes, of which age was one. [5]
6. Cloud and shared services make legacy plans moot
OMB encouraged agencies to adopt provisioned services such as cloud computing and shared services. [2] Moving to a shared service is itself a modernization, and still needs a date, work and a list of functions to move. VA’s plan to replace its time and attendance system with a shared service center was the kind of plan GAO counted as specific. [1]
7. Twelve agencies are not the government
GAO’s system-level work covered 12 agencies. GAO said those agencies made up over 90 percent of reported federal IT spending. [2] The spending analysis covered all 26 agencies that report to the Dashboard. [2] Smaller agencies may differ in detail, but the sample covered over 90 percent of reported spending.
Checks a CIO could run
These checks apply GAO’s test to any organization’s portfolio. None requires new tools.
List systems, not investments
GAO found some agencies tracked investments, not systems, and some could not date their systems. [2] Start from a list of systems, each with a business owner and an approximate age of its code and hardware.
Apply the three-element test
For each system older than your threshold, record whether a plan names time frames, activities and functions to be replaced or enhanced. [1] Mark plans that lack any element as general, as GAO did.
Check vendor support dates
For each system, record the support end date of its operating system and key components. This is the signal GAO tied directly to security. [2]
Name the people
For each system written in a language your staff no longer uses, name the people who can change it and their expected availability. GAO’s example of re-hired retirees is a warning, not a plan. [2]
Put stable systems on a calendar
Schedule an annual review for every legacy system, whatever its risk rating. GAO found that rating-triggered reviews missed systems whose ratings were lowered or never raised. [2]
Check how spending is labeled
Review whether modernization work on legacy systems is booked as operations. GAO reported OMB’s view that agencies default to operations when unsure, partly because it attracts less oversight. [2] Mislabeled work escapes the reviews that would test its plan.
Add disposition to every plan
Require each modernization plan to say what happens to the old system and its data, and when. GAO’s 2019 and 2025 reviews treated disposition as a key element. [4] [5]
Report plan quality, not system counts
Report upward the share of legacy systems whose plans meet all three elements, rather than the number of old systems. GAO’s tables reported plan status next to each system for this reason. [1] [2] A count of old systems rises and falls with definitions. The quality of plans is what management can change.
Tie each plan to a budget line
For every plan that meets the three elements, find the budget line that pays for its next milestone. Where none exists, the plan’s dates are aspirations. GAO’s 2016 report kept funding and planning separate; a CIO cannot afford to. [2]
Test one plan end to end
Pick the oldest system with a complete plan and walk through it with the program owner, the budget owner and the engineering lead. Ask each to restate the next date, the work behind it and the functions it moves. Where their answers differ, the plan exists on paper only. That exercise takes an afternoon and shows whether the rest of the portfolio’s plans can be trusted.
Record the decision to keep
Where the answer is to keep a system as-is, write that down with a date for the next decision. A recorded choice to keep is a plan. Silence is not.
With hindsight
As checked on November 16, 2026, GAO’s page for the 2016 report listed the status of all 16 recommendations. [1] Three were listed as open: the two to OMB and Treasury’s recommendation to plan legacy modernization. GAO listed every other agency recommendation as closed as implemented, describing the evidence for each. [1]
The agency closures show the plan test working. GAO recorded that SSA issued an IT Modernization Plan in October 2017, and that VA provided a comprehensive plan with a roadmap in May 2018. [1] It recorded that Defense issued its Digital Modernization Strategy in July 2019, and that State issued an IT Modernization Plan in December 2019. [1] USDA began tracking legacy systems in June 2021 and, by July 2021, had identified 275 legacy systems needing modernization. [1] GAO also recorded Commerce’s statement in May 2019 that it had retired the National Weather Service Telecommunications Gateway. [1]
Treasury’s recommendation stayed open. GAO said Treasury reported in February 2021 that it manages IT in a federated way and has no single plan for modernizing all its IT, though some bureaus have plans. [1] GAO said that as of February 2026 Treasury had not provided an update. [1]
The OMB recommendation on legacy guidance was marked a priority recommendation and remained open. [1] GAO said OMB stated in April 2021 that its December 2018 guidance directed agencies to manage risk to High Value Assets associated with legacy systems. [1] GAO said that guidance did not require agencies to identify, report, assess and remediate issues for all legacy systems. [1] It said OMB stated in March 2025 that it planned to act, and had not provided an update as of February 2026. [1]
GAO returned to the subject in 2019 and 2025. In June 2019 it identified 10 critical federal legacy systems most in need of modernization. [4] It found that seven agencies had documented modernization plans, and that only the Interior and Defense plans included the key elements GAO identified. Those were milestones, a description of the work, and a plan for disposing of the legacy system. [4]
In July 2025 GAO published an updated review. It said that as of February 2025 agencies had completed three of the 10 modernizations from the 2019 list. [5] For the new review it obtained 69 systems from the 24 CFO Act agencies, scored them on 16 attributes such as age, vendor support, language, cybersecurity risk and operating cost, and selected 11. [5]
GAO said agencies had modernization plans for nine of the 11 systems, and that three of those plans included all three elements. [5] The systems without plans belonged to Defense and Energy. [5] GAO said Treasury’s two selected systems run on COBOL and assembly language code. [5]
GAO wrote that it had recommended nearly a decade earlier that OMB direct agencies to identify legacy systems needing modernization, and that OMB had not yet acted. [5] It suggested that Congress consider requiring major agencies to develop modernization plans for their most critical legacy systems. As of February 2026, GAO said legislative action had not yet occurred. [5]
What held up is GAO’s instrument. The three elements it used in 2016, time frames, activities and functions, have close counterparts in the milestones and work description it applied in 2019 and 2025, which added the disposition of the legacy system. [1] [4] [5] What did not hold up was the expectation that OMB’s draft guidance would make the instrument universal. Agencies that wrote plans did so under their own policies, and the government-wide requirement remained a recommendation. [1]
Spotted an error? Tell us and we will check it and correct the article where needed.