The federal government spends more than $100 billion a year on IT investments. About 80% of that — by the Government Accountability Office’s own accounting — goes to operating and maintaining existing systems, including the manual labor and workarounds that keep legacy systems functional. State and local governments run at smaller scale, inside the same dynamic.
Most of that cost is invisible. It is paid out of the same headcount that is supposed to be doing strategic work, constituent-facing work, or work that requires judgment. So those things slow down instead. The cost moves from the budget and away from the services that all of us rely upon and use.
Underneath the cost is a pattern operations leaders in agencies of every size have seen up close. Licensing teams that start the morning sorting applications into stacks, one per program, before any constituent gets touched. Grants analysts who copy data from one system, paste it into another, then re-type it into a third. Eligibility caseworkers who re-key the same information across systems that should talk to each other but don’t.
It is so familiar that nobody mentions it anymore. It has a name.
Human Drudgery
Human Drudgery is what happens when humans do work a computer should be doing — the sorting, the re-keying, the copying, the reconciling, the checking and re-checking. Workarounds that became routine, then invisible. More of the staff day than anyone has measured, consumed by tasks no one would design into a process on purpose.
Government employees work hard, but often at work that computers should do, robbing them of time to do the human-side of the equation. The drudgery question is different: how much of the workday should require a human being at all?
What Drudgery Actually Costs
Here is the math an operations director can do on a team in five minutes.
Take a 100-person operations team. Assume each person spends two hours of an eight-hour day on Computer Work — sorting, re-keying, reconciling, copying between systems, checking that the workaround caught the exception this time. Two hours is a conservative estimate; in operations with three or more disconnected systems, four to six is closer to the truth.
At a fully loaded labor cost of $50 per hour, the arithmetic looks like this:
100 people × 2 hours/day × 250 work days × $50/hour
= $2.5 million per year, spent on work that shouldn't require a humanDouble the staff or double the hours and the number doubles with it. For a 500-person processing team running four hours of daily drudgery, the annual figure climbs above $25 million, before any of it shows up as a discrete budget line.
Why It Persists
Because the workarounds work. That is the trap.
When the process limps along, nobody escalates. When staff absorb the friction, leadership sees stability instead of struggle. Long wait times don’t register as system failures until they show up in the press. Slow enrollment cycles don’t become urgent until a constituent’s complaint reaches a council member or the governor’s office.
The status quo survives because it carries no career risk. Fixing it does. Every technology decision a department leader champions comes with personal exposure: the project that stalls, the implementation that fails, the partner who promised senior people and sent junior ones. Modernization initiatives have failed often enough that every procurement decision feels like exposure. It can be seen in the gallows-humor joke that “CIO” stands for “Career Is Over” – CIOs have to make decisions about these things and each decision carries job risk.
So the drudgery continues. Not because anyone chose it. Because nobody chose to stop it.
What Compounds
Drudgery does not stay flat. It compounds.
In year one, teams absorb the workarounds. Muscle memory builds around the friction. Shortcuts develop that live in people’s heads, nowhere else.
In year three, the people who built those shortcuts start leaving. Institutional knowledge walks out the door. New staff take longer to onboard because the real process is not written down anywhere. The system the workarounds were built on top of is now three years further into technical debt.
By year five, the maintenance work has shifted. Teams are maintaining workarounds, not a system. The cost hides. It shows up as slower processing times, higher error rates, longer onboarding cycles, and the quiet erosion of capacity to do the work people were hired for.
In many of the systems we re-imagine for our clients, this sort of thing has been going on for decades – the human cost is almost incalculable.
People Work and Computer Work
The fix does not start with technology. It starts with a question: which work is People Work, and which work is Computer Work?
The grants analyst who recommends an applicant for a program they do qualify for and talking to them about it? People Work. The caseworker who flags a complex eligibility case? People Work. Sorting, routing, re-keying, reconciling, and enforcing rules that do not require judgment? Computer Work. When a human does Computer Work, that is drudgery.
A better question lives underneath the automation debate: have the right classifications been made about which work belongs to whom?
This is the diagnostic phase of every successful modernization. Carrollton’s delivery methodology calls it Unearth — the requirements work that starts every engagement with deep curiosity about the organization, its people, and how the work actually flows. The principle is straightforward: it is not possible to redesign work that has not first been mapped.
It is the same diagnostic Carrollton has run on the grants, claims, and benefits systems that have moved billions of dollars of public funds through state and federal programs — work that held up in production because the People Work / Computer Work boundary was drawn before the technology was chosen, not after.
The Cost of Waiting
The window for addressing this is while the people who understand the systems are still in the building.
A useful exercise to start: take a half-hour this week and map one workflow handled manually. Count the steps that require human judgment. Count the ones that do not. The ratio is informative. Often, the conversation about modernization gets easier once a specific process and a specific imbalance can be pointed to — not as a sales pitch, just as evidence.
Source : The $100B / ~80% O&M figure is from GAO-25-107795 (July 2025), Information Technology: Agencies Need to Plan for Modernizing Critical Decades-Old Legacy Systems. Cite this report inline where the GAO figure appears.
