ClearFund
What ClearFund is, how fund accounting works, what has been built, and where it honestly stands.
Narrated walkthrough · 18 min · 8 chapters
Fund accounting in five minutes, for people who have never had to care.
Fund accounting 101
When a foundation gives a youth shelter $75,000 for its kitchen program, that money cannot buy a van. Accountants call the pot a fund and the rule a restriction.
Money the organization may spend on anything in its mission. Individual donations, most earned revenue.
Money given for one program, one project, or one kind of cost. Most foundation and government grants.
Money that may only be spent inside a window. Spend it early or late and it was never allowed.
Fund accounting 101
Inside are the rules that decide whether a dollar was spent correctly. The same four questions come up in almost every one.
Which costs the funder will pay for, and which it will not. Food, travel and equipment are the usual arguments.
The period of performance. A service delivered the day before the contract starts is not covered.
A budget by category, and how far you may drift from it before asking permission.
Reports on a schedule, in the funder's format, with numbers that tie to your books.
Fund accounting 101
Organizations that spend more than one million dollars of federal money in a year get a Single Audit. The auditor tests whether the rules were followed and files the results publicly.
The auditor's written conclusion that a rule was broken. It goes on a federal record with your name on it.
Dollars the auditor thinks may have to be paid back, out of money you did not set aside.
The same failure two years running. Funders read these. Boards read these. The next application asks about them.
What the public record shows
of audited years at small nonprofits came back with at least one finding.
organizations in the last two years failed the same way twice.
had findings in three or more separate years across the decade.
Reporting, then allowable costs, are the most common categories. Both are about rules that lived in a filed document.
Source: Federal Audit Clearinghouse, public under 2 CFR 200.512. Counts are of organizations, not individual findings.
Why careful people still end up with findings, and why it is nobody's fault.
The problem
The agreement is read the week it is signed. From then on its rules live in a drawer, and in one person's memory.
Every expense is assigned to a grant by whoever is closest to it, between a staffing call and a board meeting.
The auditor finds the drift long after the money is spent, when it is hardest and most expensive to fix.
At a nonprofit under five million dollars there is usually nobody whose whole job is finance. The executive director does it nights and weekends, or a bookkeeper does it twelve hours a week.
The problem, from the inside
A fictional twelve-bed youth shelter in Michigan. Sixteen staff, 1.8 million in revenue, four grants, nobody whose job is finance.
Executive director. Does compliance after hours because a past audit found payroll charged to a federal award without time records. Twice.
Keeps the books in QuickBooks Online twelve hours a week. Knows the money. Has not read the agreements.
The outside auditor. Wants evidence, not explanations, and wants it before fieldwork starts in November.
A compliance layer that reads the promises and checks the money against them.
The idea
Your books stay in QuickBooks. ClearFund reads them. You upload each grant agreement as the PDF it already is.
AI reads the agreement into plain-language rules, each one citing the page and clause it came from. A person confirms every rule once.
Every expense is checked against its grant's confirmed rules. Anything that trips a rule becomes a flag that names the amount, the grant and the clause.
The AI does one job, once, under review. The checking itself is deterministic: the same transaction and the same rule always give the same answer. That is what an auditor can trust.
The promises ClearFund makes back
ClearFund reads. It does not become your system of record.
Every flag points to a page and clause you can open and read yourself.
Every version of every transaction and rule is kept. Corrections are new entries, so the history an auditor sees is the real one.
A time-limited, read-only seat for the CPA, with every page view logged.
Every surface, with real screens from the Harborlight sample data.
Home
Grant portfolio
Transactions
The flag queue
Reporting calendar
Funder reports
Audit binder
Recode workbench
Costs: allocations, indirect rate, payroll effort
Books: ledger, funds, journals, statements
Money in and money out
Federal pack and activity log
What is built, what is not, and what has never been tested against the world.
Built
Nothing can be edited or deleted after the fact, and the database itself refuses, not just the app.
Frozen artifacts that hash identically every time. Proven by test, not asserted.
Deterministic checking, obligation tracking, the queue, the recode trail with separation of duties.
Six thousand lines of effort-certification machinery alone. Most of it waiting on a connector.
Not built, or not proven
Uploading an agreement and confirming its rules one at a time is the core idea, and that screen is still pending. The extraction behind it has been measured once on real documents and was weak.
The sync is built and tested against sandbox data. It has never pulled one real customer's books, because no customer has connected one.
The binder's whole thesis is that a CPA will trust it. No CPA has seen it. That gate is open and waits on one meeting.
The honest number
Everything in this deck was built from public audit records, competitor history and careful reasoning. None of it has been checked against a person who does this job. That is the gap, and it is the next thing to fix.
An outside review found several of our own market numbers were miscounted, including the claim that executive directors, not finance staff, are the buyer. The split is about even.
The incumbent weakness is real. Nobody has shipped a Single Audit binder in eleven years. The code quality is high and the repo does not lie to itself.
Three studies, and the one thing none of them could tell us.
What the research found
Blackbaud, MIP and Sage Intacct. Grants got 23 of them. Accounts payable got 387. Single Audit, SEFA and Uniform Guidance appear zero times.
Who failed, how, and how often. Reporting and allowable costs lead. The trigger is rare: a given organization is cited roughly once in three years.
Not conferences, not ads, not Reddit. A finance forum where they already ask, one-to-one email to people with a public finding, and the bookkeepers who serve dozens of them.
The one thing research cannot provide is a conversation. Everything from here is about getting eight of them in fourteen days.
Stop building blind. Start with the people.
The next fourteen days
Look up your own nonprofit's ten-year audit history, free, from the public record. Then one question: what did your last audit cost you?
To people whose public finding we can name, asking for twenty minutes about their last close. No product pitch. A corrective-action-plan template as the gift.
Six to ten finance people, one hour, no product. The platform comes later, if twenty people show up twice.
Success in fourteen days looks like eight conversations, four people volunteering to show a spreadsheet, and the same pain named unprompted by more than half of them.
ClearFund
That is the promise. The next step is finding out, from the people who keep these books, whether it is the promise they need.
Built by Nick Vander Kolk · clearfund.us