Five minutes before they called my boarding group for the last hop to Kupang, I get an alert from Trello, the way I run my one task at a time AI fractional service.
Books reconcile. Allocations aren't all done. How can the books balance if things are not reconciled?
The words from the CFO I have spent months building an ai financial controller for. I read it twice in the boarding line with my bag already on my shoulder, then had to keep moving because the queue moved and I still had a flight to catch. For two hours, I had nothing to do but sit with this and let my thoughts run rampant.
The following are raw excerpts from my journal, which I quickly opened to help me think through the reality of the situation.
What does that mean, books can't balance without the payments all allocated? What if I did something wrong and the books need to be recovered from backup, and redone manually? How is this going to affect the entire project? A hundred and four days of writing about this in public, gone. What of my relationship with the founder, will he be embarrassed if I hosed the books? Of course he would, how could I salvage this?
What will I do if this fails?
I did not know yet how bad "not right" was. What I did know, that I may need to make a backup plan and fast. This felt like the first time in twenty years I have had to seriously ask myself what happens if this does not work out.
Some background, for anyone who has been reading since Day 0: I sold my software agency I had run since 2018, after the sale of my 3rd or 4th company. I then walked out of a well-paid COO position over differences with the founder, and committed in public to a year of building a fractional AI ops business, one person, one laptop, whatever showed up. What showed up first was a monthly retainer from a good friend, someone I still ski with, who took a real risk on an idea with zero case studies behind it. I have never said that part out loud on this blog before. It has always just been "an enterprise client." He is also someone whose opinion of me I care about more than I care about the money, and that is what was sitting in my chest on that plane, not the income. The relationship was 100 times more valuable to me.
The job itself was the piece of the business I understood least: reading a live set of books across several bank accounts and currencies, matching them to the bank, and drafting the entries a CFO used to key in by hand. I planned the go-live carefully in May, watched it run clean and still miss half a statement in June, and then watched it close its first fully clean month at the end of that same June. I wrote that last post feeling like I had built something with my hands for the first time in years.
July has been an intense month. I took over running the books almost entirely myself, leaning harder on the automation than before, and over about thirty days it started to slip: entries repeating themselves, small errors nobody would notice but a proper audit would catch immediately. I am not an accountant. I know how to balance a ledger. I do not know how to check an allocation is correct, which turns out to be a much deeper piece of expertise than I gave it credit for when I started this. I had put a great deal of confidence into a system I built myself, and somewhere in July that confidence started to look a lot like hope.
So I sat with the actual question for the length of a flight, and more. I had set up a call with the CFO to get clarification on this the next day in the afternoon. That was the soonest I could get on to find out where we stood.
What do I do if the fractional work does not hold? I need roughly twelve thousand dollars a month to cover the basics, nothing close to comfortable, just the floor. I have not had to think seriously about finding a job in twenty years. The things I am good at should be worth something in this market: I can run a team. I am not a software developer, but what I would say when people ask. I'm a software entrepreneur. Which makes it questionable if I'm even hirable at all. I am technical enough to direct a build and know when something smells wrong, which is exactly the skill that got me into this and also the one keeping me from panicking about it now.
If the allocations turn out to be badly broken, the plan is simple and not fun: tell my friend everything, tonight, before he hears a version of it secondhand. Void the invoice I just sent if the books need real repair. Offer to fix it for nothing, because the friendship is worth more than one contract, and I would rather lose the money than lose that. If it cannot be fixed quickly, I would rather hand the books back clean and walk away than let it drag on. There is a smaller, far less technical contract I could lean on in the meantime, and there is a remote AI ops job market I am genuinely qualified for, even if the title would be a step down from founder. None of that is the plan I want. It is the plan that exists if the one I have been building in public for the last hundred and four days does not hold.
I have the call with the CFO in about 18 hours. I have time to dig more into this. I'm now on a remote island in Indonesia, on a different time zone. I'm jet lagged, the Internet is spotty, I feel like I'm taking risks, and it isn't a great feeling. The timing could not be worse for a month I need to fix quietly and fast, and there is nothing to do about it now except stew and let my imagination play out 100 worst case scenarios I'd rather not repeat.
Tonight, before anything else, I open the ledger myself and find out exactly how bad "not right" is. I do not know yet. Somewhere over open water I spent the better part of an hour meditating on not knowing, which felt like a strange kind of luxury: a decision that was not due for a few more hours, and permission to just sit with the fear of it instead of solving it at ten thousand feet.
This is building in public, and it's not always glamour or the way people make it sound on Linkedin.
Monthly Revenues $11,000 | Clients 2 | Prospects 1 (AI outbound agent now live)
Day 105 of 365.