Sense with Cents

I do not know the number. Nobody does. That is part of what bothers me.

The combination I am thinking about is not “small payroll software company.” Shop size is near immaterial. What I am counting is something narrower: one person who can write the code, knows the payroll rules well enough to implement them rather than just use them, understands the accounting underneath, and sells the software they built. Programmer, payroll expert, accountant, and the one taking the support call. Same person. One paycheck. I wrote earlier that being replaceable is part of the job, and I called myself something like a living dinosaur in that field. This post is what that actually looks like, on the ground, watching one of the last shops like ours close.

I have been doing this for about forty years. Jerry Medlin started writing the software in the late 1970s; the first commercial sale was 1984. I came in shortly after. So I have a long view of who else has been doing what I do, and where they went.

The integrated version is what is going extinct

Programmers exist. Payroll experts exist. Accountants exist. Salespeople exist. None of those are endangered. What is disappearing is the version where all four live in the same head.

That matters because the four jobs talk to each other constantly when they share a skull. A support call surfaces a confusing message in the program. The same person who took the call rewrites the message that afternoon. A new tax rule shows up. The person reading the regulation is the person writing the if-this-then-that. The person who sold the software to the customer is the person who hears, six months later, what the customer actually needed. Nothing has to be translated, escalated, ticketed, or scheduled for a meeting. The loop is short because it does not leave the building.

Most of the integrated operators are gone. Retired, sold to a roll-up, absorbed into a larger company where the founder became an employee until the non-compete ran out, or just closed the doors when the founder was tired. Our last direct competitor closed at the end of 2023. Paul Mayer ran ZPAY Payroll Systems out of Illinois — a Vietnam veteran, longtime shareware figure, started writing payroll software for Heathkit computers in the late 1970s and incorporated in 1983, almost the same arc as ours. Paul’s health caught up with him, the consequences of Agent Orange exposure decades earlier, and he had to close. His contract programmer did not want to take over the company. So it just stopped.

I heard about it from a new customer of mine who was shopping for a replacement. Paul and I were not friends, not even colleagues — we were competitors, and I had met him maybe once in passing. I reached out anyway, out of kindness and to offer his customers somewhere to go. Paul tried to help his customers by letting me help them. He shared enough about his product that I could write an import path for his customers’ data into ours, and he graciously shared a list of companies willing to work with them. That was Paul looking out for his people on the way out. I would not have announced a closure without a path forward already worked out — I would have exhausted every option for an arranged handoff first — but Paul’s health did not give him that runway. What he did do was respond to everyone who reached out to help, which could not have been easy under the circumstances. The import path I built was not a sales move. It was a path for people stranded by the closure. Others eventually offered the same. We heard, months later, that someone had bought the source code, but no continuation of the company itself. I was not interested in buying the source code or the customer list. Both felt tainted to me, because the closure had been announced with no path forward worked out.

Candidly, Paul’s software had also stayed pitched to expert users — people who could self-manage their own tax calculations and knew payroll well enough to drive software that assumed they did. I have no way to confirm this, but my read is that the customer base for that kind of software has been thinning out for years. We evolved ours for everyday users — small business owners and bookkeepers who do not want to think about payroll any more than they have to. Paul’s health is what closed the company. I only mention the customer-base point because the integrated operator is the one who notices that kind of shift, since they are in the loop with customers. A split-work organization rarely catches it in time, because the people who hear it from customers are not the same people who could change the software in response.

The companies that look like ours from the outside almost always have the work split now. A coder. A payroll specialist. Someone in accounting. Someone in sales, web design, and marketing. That is a perfectly reasonable way to run a business. It is just not the same animal. And we must plan for that in our own future too, or we will disappear, because we (company and customers) cannot and should not count on another me, another Jerry.

Watching Paul did not change our planning. It pushed us to document it. What stays in one head is not actually a plan — it is a hope. Paul’s closure was the reminder.

Medlin made it through Jerry’s passing because I was already in place. There was no need to replace the integrated operator — I was the next one, and Jerry had given me the time to become that. The mechanics behind the scenes still had to happen, but the company did not have to find a new center of gravity. My version is harder. I have to design a path that does not depend on another integrated operator showing up at the right time, because the post you are reading is the argument that I cannot count on one. That may be one of the hardest things a person has to do.

I had been dreading finding another me for years. I knew I should. I did not know how, and the end was too big to get me to the first step. Something has changed about that recently, and I will get to it.

We — Medlin Software, Paul’s customers, anyone in the small-software world paying attention — actually all owe Paul a thank-you for being open and willing to take help once we and others asked. He could have closed the door and stayed silent. He did not. The lesson cost him a great deal to teach.

The pocket calculator parallel

I learned math the old way. Rote memory, carry the digits, line up the columns. The pocket calculator made most of that obsolete as a primary skill. Nobody needs to do long division by hand anymore, and pretending otherwise is silly. What the calculator did not replace was the ability to know whether the answer is about right. I still do most arithmetic in my head, mostly as a sanity check on whatever the calculator says.

I was not allowed to bring my expensive calculator into school, because it was feared we would not learn math properly. Now nobody cares. There is a calculator in every phone, in the voice assistants, in AI, and even in the address bar of the web browser. When was the last time you worked a problem with a pencil?

That is the right way to think about AI too. It is the new knowledge bank. Encyclopedias, history books, and manuals — all of it — sitting in a device, ready in seconds, provided you have learned how to ask. The job is no longer to memorize. The job is to ask well, and to recognize when the answer coming back is wrong.

I was fearful of AI at first. I started using it after a major update, once I had tested it on edge cases I cared about and found the results good enough to trust. Writing articles is one example of where it earns its keep for me. My voice is a known quantity — I do not need to relearn the mechanics of a good article every time I sit down to write one — so I feed AI my thoughts, ask it to check whether anyone else has already written something similar, and read carefully through what it gives back, pushing on anything that does not sound like me. This article was produced that way. The thinking is mine. The polish is the tool’s. In school, the best part of journalism was the permanent pass to wander the halls at will “for journalism.” AI is that pass for me now. I can wander in and out as thoughts arrive, with zero concern about order or formatting, and the fun part of journalism is back.

Ishi, reluctantly

In 1911 a man walked out of the hills near Oroville, starving. He was the last surviving member of the Yahi people. The anthropologists at Berkeley called him Ishi, which means “man” in his language, because he would not give his real name — there was nobody left to introduce him. He spent his last years recording songs, making arrowheads, and answering questions about a way of life that died with him.

I am not Ishi. The parallel is not the tragedy. It is the integration. The Yahi knowledge did not vanish completely — pieces of it survived in Kroeber’s notebooks and Ishi’s recordings. What vanished was the version that lived in a single person who could speak the language, knew the songs, made the tools, told the stories, and connected all of it without thinking about it. The pieces are not the same as the whole.

The parallel has occurred to me more often as the years go on.

Where the knowledge lives now

AI is going to hold most of what took me forty years to accumulate — the payroll rules, the tax tables, the edge cases I have already seen, the code patterns, the support-call patterns, and the accounting logic underneath all of it. Retrievable in seconds by anyone who knows how to ask. That is genuinely good.

What AI cannot do is the truly new case, or the rare edge that nobody has written down yet. There is not enough data for a reliable answer, and the fewer integrated operators are still around to recognize that kind of problem when it arrives, the more we will all be depending on AI in exactly the situations where AI is least reliable. That is the part that should worry people, and it is not the part most of the AI conversation is about.

The new human skill is the same as the old human skill. Find a problem nobody has solved yet, and figure out a solution. That is what Jerry Medlin did in the late 1970s when he wrote the first version for himself. That is what the next person will do with tools I never had. The job description changes. The work does not.

Once I admitted I was likely the last of my kind, the path forward got simpler. The question stopped being “how do I find another me” — which I could not answer — and became “how do I cover the tasks I do.” That second question is just work. The four jobs in my head — programming, payroll, accounting, sales — do not have to live in one head anymore, because AI fills in the gaps the way a fifth person used to. We already have most of those skills in the family. A programmer. Someone working the marketing and sales side. Others learning the payroll and accounting pieces. Even family who do not want to actively run the company can govern those who do, which is its own kind of succession. Years of trying to find one of me had me stuck. Looking instead at the people already in the room got me moving.

Every piece of the problem has a known answer once you are willing to look at it as separate pieces. The mystery was only ever in trying to solve them as one. Ignore the elephant. Look at the separate tasks.

That principle is not new to us. We have been running the company on it for decades, without naming it that way. We ignored the elephant — the big-name companies, the conventional wisdom about what payroll software has to be — and instead divided the work among ourselves, kept the product focused, and served the customers we knew best. We did not try to be all things to all people. That is a perfectly fine path to success. The succession problem turned out to be the same principle, applied later than I should have applied it.

The same logic works on the customer side too. Most of the people who use our software are running their own businesses — service providers, small shops, professionals who use payroll software the way they use a kitchen knife. They want to do the work, not study the tool. They have their own elephants to ignore — the pressure to master every feature of every piece of software they touch, the feeling that they should keep up with whichever app is trending this quarter. The healthier path is the same one: pick a focused tool, use it to get the work done, and spend the time you save on the work that actually requires you. That is how we have always thought about who our software is for, and it is why I think the elephant principle generalizes well beyond our seat.

Back in the early 1990s, as I remember it, Medlin was bundled on a 5.25-inch floppy disk with a college accounting textbook — the disk in a paper sleeve with a glued flap inside the back cover. One semester, basic accounting on a computer, with our software as the working tool. No separate software class. No software training. Students learned accounting and the software stayed out of the way. Compare that to now, where you can take entire classes — earn certifications — just to learn to operate a particular software product. No accounting credential underneath. Just the certificate to use the tool.

If you are still out there

I would like to hear from you.

I do not know how many integrated operators are left in payroll software, or in any small software business where one person carries the whole stack. I have a guess that the number is small and getting smaller. I would like to find out I am wrong.

If you run a small business where you write the code, know the rules, do the work, and answer the phone — payroll, accounting, anything where the shape rhymes — I would like to talk. I am not selling anything. I am not organizing anything. I think the younger workers in our fields are going to need a few of us still around to ask, and that is harder if we do not know each other.

Reach me directly: dennis@payrollhelp.us

One real conversation with a peer is worth more than another article about how the world is changing. Whether this post finds any of you is, in a quiet way, the experiment.

I hope I am not the last. But if I am, I will wear it with honor, and try to leave a trail.

See also: Being Replaceable Is Part of the Job, Do You Really Have a Backup?, Help Me Help You, and Who Writes In, the Names That Stay — and What’s Next.