The Unread Record

In 2005, Guam's lawmakers took an honest look at their own government online and wrote down what they found. Many departments had no website at all, and many that did were out of date or so thin as to be useless.¹ That year they passed the Transparency Act, requiring every department and agency, including the Legislature and the Judiciary, to keep a public website and post specified information.²

Twenty-one years later, the problem has flipped. The Bureau of Budget and Management Research publishes consolidated revenue and expenditure reports online and tells residents they don't need to file a request or visit an office to see them.³ That is a good-faith effort. But a monthly PDF of financial tables is a record, not an answer.

This spring showed what that costs. When the government's fiscal 2024 federal compliance audit flagged $22.6 million in questioned costs, senators convened an investigative subcommittee, and Adelup accused them of politicizing the findings.⁴ Under federal rules, a questioned cost is one an auditor flags because it may violate grant terms, lacks adequate documentation, or looks unreasonable; it isn't treated as an improper payment until someone reviews and confirms it.⁵ A resident who wanted to weigh both sides would have needed to know which report to open, find it among dozens of PDFs, and understand that definition. When people can't read the record, they don't reach their own conclusions about it. They choose whose version to believe.

Transparency should be measured by what a resident can learn, not by what an agency has uploaded. Any office could run a simple test: take an ordinary question, such as what the office spent on travel this quarter, and time how long a resident needs to answer it using only public tools. If the answer takes an afternoon, the office isn't transparent yet. Passing that test takes three things.

To know what an office did this week, residents need summaries that arrive on their own. Each day's actual payments should appear as they clear, drawn straight from the financial system rather than retyped by staff. Every week, a plain-language note should explain what the office accomplished and what that work cost. For an office that performs inspections, that could mean a public chart of how many requests arrived and how many were completed, and a way for each business to see roughly when its turn will come. Every month, spending should be set against the budget and last year, with anything unusual explained before a reporter has to dig it up. A report that depends on someone remembering to write it will eventually be late, and a late report looks like something is being hidden.

To find one payment, residents need search. Typing a vendor's name, a dollar amount, or a date should bring up every related payment and contract. In its 2018 report, U.S. PIRG found that only 24 states let users narrow results by department and spending category at the same time, even though at least 1.5 million users had visited state transparency sites the year before.⁶ Ohio went from a D-minus in 2014 to the nation's highest score in 2015.⁷

To ask what no report anticipated, residents need a way to pose a question in ordinary words, such as the travel question above, and get a direct answer tied to the underlying records.

That last idea draws the strongest objection. Letting software answer questions about public money sounds risky, and it can be. In 2024, The Markup found that New York City's official AI chatbot was telling business owners they could take workers' tips and turn away tenants with housing vouchers, both generally illegal in the city.⁸ This year, the new mayor called the bot "functionally unusable," noted that it had cost the city about half a million dollars, and the city took it down in February.⁹ A government tool that answers confidently and wrongly wastes money and trust at once.

That lesson should shape how such a tool is built, not whether it is built. The New York bot was interpreting law. A spending tool has a narrower, more checkable job: retrieving records the government already keeps. Every answer should display the payment, contract, or audit page it came from, so anyone can verify it in one click. When the records don't answer a question, the system should say so and name the office that can. It should be tested before launch and corrected in public when it errs. A tool that admits "I don't know" earns more trust than one that guesses.

Privacy raises a second concern. Not every payment belongs online line by line. Assistance to individual families, medical matters, and settlements involving private citizens deserve protection. Daily posting has to come with clear, published rules about what is withheld and why, so that the redactions themselves are transparent.

Cost is usually raised first, yet it may favor transparency. PIRG reported in 2010 that the federal spending site, covering more than $2 trillion a year, cost less than $1 million to build, and that Missouri built its daily-updated site with existing staff and funds. Texas reported $4.8 million in administrative savings in the two years after its site launched, and Utah expected millions saved from fewer information requests.¹⁰ For an office short on staff, every question a portal answers is one fewer request someone has to process by hand. Guam's books are far smaller. Having built data systems for about twenty-five years, I can say the hard part is rarely the technology. It is the decision to do it.

A portal also has to be kept alive. As of 2018, PIRG found that Hawaii had posted no spending data for years after 2016.⁶ A portal left to go stale can be worse than none, because it tells people nobody is watching.

None of this requires waiting for new laws. The first step costs nothing: a branch head could run the time-to-answer test on the office and publish the result. From there, the office could turn on automatic daily posting of payments and commit to weekly and monthly plain-language summaries. It could publish payment and contract data in searchable, downloadable form instead of scanned images, answer every public question about spending in writing, and post those questions and answers for everyone to see. It could also invite outside verification. South Dakota, PIRG reported in 2014, audits its online checkbook every year so users can trust the data.¹¹ Guam's Office of Public Accountability could provide that check.

Openness won't guarantee that every dollar is spent well. But it changes what happens next. Waste that can be seen gets questioned early, while it is still small. And a mistake an office discloses on its own looks very different from one a reporter uncovers.

There is an old idea, older than any audit standard, that those who manage what belongs to others will one day give an account of it. Good stewards don't wait for the auditor. They keep the books open because the money was never theirs. Leaders who hold that view will find that trust isn't something they have to ask for. It returns on its own, one answered question at a time.