What It Cost

The tax was invisible because everyone paid it. Training, workarounds, double entry, knowledge that walked out the door. We called it the cost of doing business. It wasn't. It was the cost of the wrong interface.

In a family-owned motor repair shop, somewhere in the American Midwest, there are boxes.

Cardboard boxes, stacked on metal shelves, each one filled with 4×5 index cards. One card per motor. The shop has been refurbishing commercial electric motors since the 1940s — the kind that run factory floors, pumping stations, processing plants. Every motor that ever came in has a card. Every repair, every part, every customer, every date. Eighty years of operational history, written by hand, filed in boxes.

The founder's system worked. He built it, he understood it, and for decades it was enough. When a motor came back in, you pulled the card, read the history, knew what you were dealing with. The knowledge lived in the cards and in the people who knew how to use them.

When I met this business, the founder's son was running it. The boxes were still there. The cards were still there. Finding a job's status meant digging through them. Knowing what was in the queue meant walking the shop floor. Searching history meant searching by hand. The system hadn't changed because there'd never been a compelling reason to change it — just a slow accumulation of friction that everyone had learned to work around.

That friction had a cost. Nobody had ever added it up.

The invisible tax

Businesses are drowning in subscriptions for tools that do ninety percent more than they need and somehow still miss the ten percent that matters.

That line came from my own experience after thirty years of watching it happen. But the motor shop version of the problem is older and simpler. It predates SaaS subscriptions by decades. It's just friction — the gap between how work actually flows and how the system requires it to be recorded.

Every business has this gap. The size of it varies. The shape of it varies. But I have never worked with a business that didn't have at least one spreadsheet being maintained by at least one person because the "real" system couldn't handle something important. That spreadsheet is the gap made visible.

The cost of that gap is almost never calculated because it's paid in small increments. Minutes here. Hours there. A double-booking that costs a relationship. A piece of gear that goes out without an invoice. A paper that gets published in a researcher's field while he's still searching last month's PubMed results by hand. No single incident breaks the business. The accumulated weight just slows it down.

Five costs worth naming

After three decades building operational software for businesses that make, move, rent, fix, and manage real things, I've watched the same costs appear in different forms across different industries. They're worth naming precisely because they're so rarely named.

1. The training tax

Every new employee at a business running standard software has to learn two jobs simultaneously. The actual job — how the business works, what the customers need, what quality looks like — and the software job: which screen to navigate to, which fields to fill in which order, what the status codes mean, where things go when they're done.

The software job is pure overhead. It produces nothing. It exists entirely because the system wasn't built for the way the business works — it was built for a generic version of the business, and the employee has to learn the translation layer between generic and specific.

In a fifteen-person company, this cost is manageable but constant. Someone always needs to be brought up to speed. Someone always needs to be corrected when they use the wrong workflow. Someone always leaves and takes their knowledge of the system's quirks with them.

The motor shop had no training tax for the index card system. A new employee could look at a card and understand it immediately. The system was legible because it was built for the work, not for a computer.

2. The workaround tax

A Hollywood lighting rental company was tracking thousands of pieces of high-value equipment across dozens of film and television productions. Their system was a combination of Excel spreadsheets and a literal wall of Post-it notes — one note per piece of gear, moved by hand from column to column as equipment went out and came back.

This was not negligence. This was a rational response to software that didn't fit. The existing tools couldn't handle the specific combination of equipment tracking, production scheduling, and client billing that this business required. So the team built their own system out of the materials available: sticky notes and spreadsheets.

The workaround tax has two components. The first is the direct cost — the time spent maintaining the workaround, moving the Post-it notes, updating the spreadsheets, keeping the parallel system in sync with the official one. The second is the risk cost — double-bookings happened regularly, and revenue leaked through gear that went out without getting invoiced. The workaround system was doing real work, but it was fallible in ways that a purpose-built system wouldn't be.

I have never built software for a business that didn't have at least one workaround I was replacing. Usually several. They exist in every industry, at every scale, and they are universally treated as normal. They are not normal. They are evidence that the software doesn't fit.

3. The double-entry tax

An event production company in Boston was running on three separate systems that didn't talk to each other. Deals lived in one tool. Scheduling lived in a Google Sheets workaround. The core operational database sat in a third system. Every piece of information had to be entered at least twice — sometimes three times.

Their software bill for this arrangement was over eight hundred dollars a month. That's before accounting for the staff hours spent on redundant data entry that produced nothing. Enter a deal in the pipeline tool. Enter the same deal in the scheduling sheet. Enter it again when it becomes an active project in the operational database. For every deal. Every week.

The double-entry tax is particularly insidious because it hides inside normal operations. Nobody notices it as a discrete cost. It's just part of how the job works. Until you add it up.

When we unified their three systems into one — enter once, see everywhere — the staff recovered hours per week that had previously been consumed by redundant entry. Those hours didn't go to overtime. They went to actual work.

4. The knowledge tax

A UK-based photomedicine research company was staying current on published literature the way researchers have done it for decades: searching PubMed manually, copying relevant papers into spreadsheets, categorizing them, removing duplicates. The founder spent hours every week on this process. It was tedious, error-prone, and couldn't scale.

The knowledge tax is what you pay when human expertise is applied to tasks that shouldn't require human expertise. Searching a database. Copying records from one system to another. Maintaining a spreadsheet that tracks what the official system should track but doesn't. Walking the warehouse floor to see which bays are available.

These are tasks that exist because the system can't do them. They consume time that should go to judgment, analysis, relationships — the things that actually require a person. When we connected the research company directly to PubMed's API and automated the nightly import, the founder recovered his evenings. More importantly, he now sees every relevant paper published in his field the morning after it's indexed. The system does the searching. He does the thinking.

That's an eight-times efficiency improvement, measured by the researcher himself. It came entirely from removing a task that shouldn't have been his job.

5. The departure tax

This is the one that bothers me most.

My veterinarian has been practicing for thirty years. He knows things about his patients that no database has ever captured. Which animals are difficult to handle under sedation. Which farms have recurring herd problems that trace back to the same root cause. Which treatment combinations worked last spring when this exact constellation of symptoms appeared. What the previous owner of that dog told him seven years ago that turned out to be important.

None of that is in the system. The system has records — dates, diagnoses, medications, billing. But the pattern recognition, the contextual knowledge, the accumulated judgment that makes a thirty-year practitioner different from a first-year associate — that lives in one person's head.

He's in his seventies. When he retires, that knowledge retires with him. The new associate inherits the records but not the understanding. Every business that has ever lost a key employee to retirement, departure, or illness has paid this tax. It is the most expensive cost on this list and the least visible, because it only becomes apparent after the person is gone.

The system never learned what the expert knew because the system had no way to ask. Until now.

What these costs have in common

Training, workarounds, double entry, knowledge overhead, departure losses. Five different costs with five different shapes. But they share a common origin.

Every one of them exists because the software was built for a generic business and then deployed to a specific one. The gap between generic and specific is where the costs live. The training tax teaches employees to bridge the gap. The workaround tax builds parallel systems to cover it. The double-entry tax fills it with redundant labor. The knowledge tax applies human expertise to problems the system can't solve. The departure tax reveals it when the people who knew how to navigate the gap are gone.

The motor shop's index cards didn't have most of these problems. The cards were specific — built for exactly this business, this workflow, this industry. A new employee could read them without training. There was no workaround because the system was the workflow. Nothing was entered twice. The knowledge lived in the cards, not just in the people.

The cards had other problems. They couldn't be searched quickly. They couldn't be accessed from the shop floor while hands were dirty. They couldn't generate a report or send an invoice. The translation to digital was genuinely overdue.

But the principle the cards embodied — build the system for the work, not for a generic version of the work — was correct. It just needed to be carried forward into software that could do what paper couldn't.

The process didn't change. The friction disappeared.

That's what I told the owner after we replaced eighty years of index cards with a system built for exactly how his shop operates. Technicians update job status on iPads on the shop floor. The office sees the full queue from the desktop. Every motor is searchable in seconds. Productivity improved four times over — measured by jobs processed per week, by the owner himself.

Four times. From the same workflow, the same people, the same shop. The process didn't change. The friction disappeared.

That phrase has stayed with me because it names something important about what good software is supposed to do. It's not supposed to transform your business. It's not supposed to teach you a new way of working. It's supposed to carry your existing way of working forward — and remove everything that was slowing it down.

Most software does the opposite. It asks the business to adapt to the software's model of how work flows. The training tax is the price of that adaptation. The workarounds are the evidence that the adaptation was never complete.

The reckoning

I've spent a long time thinking about why these costs stayed invisible for so long. Part of the answer is that they're genuinely hard to see when you're inside them. The training tax feels like onboarding. The workarounds feel like resourcefulness. The double entry feels like diligence. The knowledge overhead feels like expertise. The departure tax only becomes visible in the aftermath.

But part of the answer is simpler: there was no alternative. Custom software was expensive, slow, and risky. Off-the-shelf software was imperfect but available. Every business made the same calculation and came to the same conclusion. You bought the closest thing that existed, adapted your workflow to fit it, and learned to live with the gap.

That calculation has changed.

The cost of building purpose-built software has dropped by an order of magnitude. What once required a team of five and six months can now be done by one person who deeply understands the business, in days. The AI systems that are rewriting what software can do are simultaneously rewriting what software costs to build and operate.

The businesses that figure this out early will close gaps their competitors are still working around. The ones that keep adapting to software that almost fits will keep paying taxes they've stopped noticing.

The index cards were never the problem. The gap between the cards and what the business needed was the problem. And for the first time in the history of computing, closing that gap is within reach of any business willing to look at it honestly.

The next chapter is about what that looks like when you stop thinking about software as a product and start thinking about it as a conversation.