Why a Developer Bills 50 Hours for a 10-Minute Fix

The Moment a ₩60,000 (~$43)-an-Hour Developer Bills 50 Hours for a 10-Minute Job

Say there’s a developer who charges ₩60,000 (~$43) an hour. A client asks for a feature, and it rings a bell — she built the exact same thing 3 years ago. Back then, it took a full 50 hours. This time, she pulls the old code, drops it in, and it works in 10 minutes. Tests pass, no problem.

So should the company pay for 50 hours, or for 10 minutes?

Pay for 10 minutes, and the developer gets penalized for having been efficient enough to reuse her own work. Pay for 50 hours, and you’ve abandoned the principle that compensation should track actual time spent. Either way, the contract needs to spell out what the same output is worth.

This isn’t a question of one person’s honesty. It’s a problem baked into the structure of the contract itself, because the assumption that “hours in equals value out” has broken down. And this same problem is showing up simultaneously in the labor market for human work and in the market for software.

🔴 Four Situations Where Hourly Billing Breaks Down

Hourly billing works perfectly under specific conditions — when time invested and output are proportional. The problem is that in knowledge work, that proportionality breaks down often. Let’s look at four scenes.

First, the zero-hour idea. You’re in the shower and an idea hits you that could change your company’s entire cost structure. No research, no meetings. How much should hourly billing pay for this idea? Zero hours means ₩0 (~$0). Companies don’t buy ideas separately.

Second, reuse. This is the case of attaching a 50-hour asset from earlier to a task that takes 10 minutes. Hourly billing has no way to convert the value created by those past 50 hours into the present 10 minutes.

Third, delegating to an agent. You hand a 10-hour task to an AI agent, run a few loops of reviewing its plan and checking its output, and finish in 30 minutes. The value of those 30 minutes comes from “knowing what to delegate” — but hourly billing has no way to price that skill.

Fourth, the failed 10 hours. You spend 10 hours only to realize the direction was wrong from the start. Can you bill for this? Hourly billing recognizes these 10 hours as legitimate labor. It doesn’t recognize the previous three scenes.

If the hourly rate stays the same, you can bill for time spent on trial and error — but the more skilled you become and the faster you finish, the less you get to charge.

So contractors start wondering whether to turn down hourly clients, and the temptation grows to inflate the hours actually worked.

I don’t recommend padding your hours. But if a structure consistently rewards a certain behavior, that’s not a sign of a problem with the person doing it — it’s a sign the structure itself is badly designed. If a contract only holds up because individuals rely on their own conscience, that means the terms of the contract itself need to be redesigned.

The company’s response was surveillance, and it doesn’t work

To stop inflated reports, companies reached for surveillance tools — software that captures screenshots every 15 minutes, counts keystrokes, and demands hourly work logs. The industry calls this bossware¹. According to market research, 78% of employee monitoring tools now come with screenshot functionality.

So does it actually work? I came across a study that I found genuinely interesting.

It’s a randomized controlled trial² published in 2025 by Namrata Kala of MIT and Elizabeth Lyons of UC San Diego. They randomly assigned digital monitoring across an online labor market, splitting cases where the reason for monitoring was explained from cases where it wasn’t, and measured performance across both.

Two findings emerged. First, monitoring itself had no significant effect on performance, on average. The common assumption that surveillance makes people work harder simply wasn’t supported by the data. Second, when monitoring was introduced or removed without an explanation for why, worker output dropped significantly. It wasn’t the act of monitoring that hurt performance — it was the failure to explain why it was happening.

In this experiment, surveillance tools alone didn’t improve performance; what mattered was whether managers explained their reasons for introducing or lifting monitoring. I’d also argue that monitoring conditions carry another risk: they can drive away exactly the people who have other job options. If you exclude any applicant unwilling to submit screenshots, you also narrow the pool of people your company can actually hire.

“So why not just charge 3x the hourly rate instead?” That doesn’t work either. Put someone quoting ₩300,000 (~$217) an hour next to someone quoting ₩100,000 (~$72), and a company judging by price will pick the latter — even if the latter logs 3x the hours and the total cost ends up the same. The person who honestly raised their rate ends up losing the contract instead.

🌏 The Same Problem Shows Up in Software Pricing

If the story so far has been about people, what follows is about software. The structure is nearly identical: pricing based on time or headcount.

In the last issue, I covered the collapse of the billable hour³ in the legal industry.

In a business that sells time, what do you sell once AI shrinks the time itself?

Say a lawyer billing $300 an hour used to spend 25 hours drafting a brief, and now finishes it in 10 hours with AI. To collect the same fee, the hourly rate would need to jump to $750. According to the 2025 Legal Trends Report, 74% of law firms’ hourly-billed work is exposed to automation. The very work that generated revenue is the work automation targets.

But this isn’t just a legal-industry problem. The exact same thing is happening in SaaS.

Seat-based pricing⁴ is collapsing. The share of SaaS companies using per-seat pricing fell from 21% to 15% in just 12 months. Meanwhile, hybrid models blending subscription fees with usage-based pricing rose from 27% to 41%. The reason is simple: if an AI agent does the work of 5 junior employees but you’re still charging by “number of seats,” the vendor’s revenue shrinks the more efficient the customer becomes. The customer’s success becomes the supplier’s loss.

So “per-resolution” pricing emerged. Zendesk charges $1.5 per ticket its AI agent resolves automatically (on contract terms). Intercom’s Fin charges $0.99. Instead of billing a flat monthly rate per agent, vendors now charge per ticket resolved. They’ve started pricing outcomes instead of time.

Let’s also revisit a figure from two issues ago.

In a world where a robot’s hourly wage is $5, what will humans do?

The hourly operating cost of US warehouse robots is $5.71 — about a third of the average hourly wage for human warehouse workers.

All three cases point in the same direction. Across legal hourly billing, SaaS seat-based pricing, and warehouse robot hourly operating costs, pricing by time is breaking down — whether it’s hours a person worked, the period a seat was used, or the hours a robot ran.

Why did ‘time’ become the billing unit in the first place

Let me take a step back. Billing by the hour isn’t a law of nature. It’s a legacy of the factory.

In factories, hours worked and output were almost perfectly proportional. Stand there for 8 hours, and you got 8 hours’ worth of parts — so time was an excellent proxy metric. It was easy to measure, too. 20th-century knowledge work borrowed this unit without much thought. Law firms, consulting firms, agencies, freelance platforms — all of them.

That proxy metric doesn’t fit well anymore. Even before AI, hours invested in knowledge work never tracked cleanly with the value of the output — but now that AI can produce far more in far less time, that gap has widened dramatically. Let me bring back last issue’s question: “In a business that sells time, once AI shrinks that time, what exactly are you selling?” This was never just a question for the legal industry. It’s a question for anyone who gets paid by the hour.

📊 So Is Performance-Based Pricing the Answer?

We shouldn’t jump to conclusions here. Getting paid for outcomes instead of time sounds intuitive, but the data shows performance-based pricing⁵ has its own formidable obstacles.

Measurement and attribution are hard. Real outcomes—revenue growth, cost savings—have multiple inputs. It’s difficult to isolate whether AI drove the result, whether the client’s own team did, or whether the market was simply favorable. That’s why, as of 2022, only 17% of enterprise SaaS vendors had actually implemented genuine performance-based pricing.

Sales cycles slow down. Analysis shows performance-based contracts—with their baseline measurements, PoCs, and legal safeguards—stretch sales cycles by 20-30%.

Finance departments hate it. 64% of SaaS finance executives named “unpredictability” as their top concern with performance-based models. When your revenue is tied to a customer’s outcomes, you can’t even plan your own company’s budget.

You need a track record proving your product delivers results. 78% of companies that succeeded with performance-based pricing had been on the market for 5+ years. A new team simply has no basis on which to promise outcomes.

Project-based pricing runs into the same wall. If requirements change, you have to renegotiate; scoping alone takes days; and if the estimate is off by 3x, the contractor eats the loss.

So the real-world answer isn’t pure performance-based pricing—it’s a hybrid: a base fee plus a performance component. Industry watchers expect hybrid models to reach 61% of the market by the end of 2026. In other words, performance-based pricing isn’t going to swoop in and replace time-based billing outright. Instead, an awkward mixture will likely dominate for a while.

Oswarld’s Lens

To be honest, I barely use hourly or per-project billing. I settle accounts weekly or monthly, and reports are either very short or nonexistent.

There’s one thing I learned from doing consulting and GTM strategy work: measurement itself has a cost. People usually assume “accurate measurement” is free. It isn’t. There’s the time spent filling out timesheets, the time spent building reports, the manager’s time reviewing them, and the cost of eroded performance from the feeling of being watched. When you actually run the numbers on hourly-billed projects, it’s not uncommon for this administrative overhead to eat up 10-20% of the billed amount. The management cost incurred to settle accounts more “precisely” can end up exceeding the benefit it produces.

So instead of pay-for-performance, I chose to trust the person and pay a fixed amount we’ve agreed on. There’s almost no cost to logging time and reviewing that log. Every few weeks, I look at the value the person has generated — not just completed tasks, but ideas they proposed, time they spent helping colleagues, documentation they left for the team. I don’t ask how many hours they worked. I already know that day to day anyway.

That said, let me honestly admit the limits of this model. It doesn’t work everywhere. It only works at a scale where a leader can see a team member’s work every day — roughly a startup of around 10 people or fewer. Beyond that size, leaders inevitably reach for a proxy metric, and the easiest proxy metric is, once again, time. And from the contractor’s side, when I propose this model, about half of prospective clients walk away. They can’t get past the question: “So what exactly am I buying?”

Finally, let me answer the hardest question. If someone who’s performed well for months produces nothing at all last week, should they still be paid for that week?

I pay them. What I’m paying for isn’t that one week’s output — it’s the fact that this person continues to stay on the team and keep working with us. Output is naturally uneven from week to week. Some weeks there’s almost nothing; some weeks there’s ten times the usual. If you cut pay for a week with no output, you’ve effectively reverted to hourly billing. But this answer only applies to someone who has already built up trust over time. If empty weeks keep repeating and you keep paying anyway, that’s not trust — that’s neglect. Here’s what actually makes trust-based settlement demanding in practice: the manager has to make a judgment call about each individual person, and can’t defer that judgment. This might be exactly why hourly billing is so widely used — logged hours let you outsource the judgment call.

Looking at the draft, it matches the Korean source well in meaning, structure, and numbers. No Hangul remains, all numbers (10 hours, 10 minutes) are preserved, and headings/rules match.

Closing

Under hourly billing, you can bill for 10 hours spent wandering in the wrong direction, but if you finish the same job in 10 minutes, you get paid less. The more AI lets you cut working time, the worse this contract becomes for skilled people. Monitoring tools don’t solve this problem either — they just push out the people who have alternatives first.

The same issue is showing up simultaneously in legal billable hours and SaaS per-seat pricing. Pricing models built around time and headcount are shaking together.

That said, performance-based billing isn’t a cure-all. Measuring output and figuring out whose contribution counts is hard, and what worries finance teams most is that forecasting revenue becomes harder. For now, a hybrid model is the realistic answer, and for small teams, trust-based settlement is one more option on the table.

If you’re currently billing — or being billed — by the hour, tell me in the comments: when was the moment in that contract when being honest about your hours felt like a loss? Was it reuse, delegating to an agent, or a failed 10-hour stretch?

---

📨 If you know a colleague who bills by the hour, pass this along

---

Looking at this fragment, I need to check for Hangul, number mismatches, and structural drift.

The draft looks clean overall. Let me verify the footnote numbers and structure match the source exactly.

References & Further Reading

Primary sources

• Namrata Kala & Elizabeth Lyons, “The Effects of Digital Surveillance and Managerial Clarity on Performance”, NBER Working Paper 33348, 2025. — This is today’s core evidence: it’s not surveillance itself but the failure to explain that erodes performance. The abstract alone gets you the whole argument.
• Monetizely, “The 2026 Guide to SaaS, AI, and Agentic Pricing Models”, January 2026. — Covers both the collapse of seat-based pricing and the structural limits of outcome-based pricing. The “Structural Limits” section near the end is the most useful part.
• Thomson Reuters Institute & Georgetown Law, “2026 Report on the State of the US Legal Market”, January 2026. — The original report where the phrase “productivity-profitability paradox” comes from.

Background

• Zendesk, “AI agent resolution-based pricing guide” — Shows how per-resolution pricing is actually priced out in practice.
• Intercom, “Fin pricing policy ($0.99 per resolution)” — A real-world example of a price tag that sells outcomes, not time.

Companion issues worth reading

• In a business that sells time, what do you sell once AI shrinks the time itself? — The legal-industry section of today’s piece is a direct extension of this issue.
• In a world where a robot’s hourly wage is $5, what will humans do? — A story about just how far the price of time has already fallen.

The author is Oswarld (Kwangseob Ahn). Current roles: Adjunct Professor at Sejong University, Strategy Consultant at INLEVEL9. Career, research, books, and recent work are kept current on the About page. Latest · July 2026: HEMA-2: A Consolidation-Aware Tri-Memory Architecture with Multi-Channel Scheduling for Lifelong Conversational AI.

Footnotes

1. Bossware: a term for surveillance-style workforce management software that automatically logs employees’ screens, keystrokes, and login times. It combines “Boss” and “Software,” and is generally used with a negative connotation. ↩

2. Randomized Controlled Trial (RCT): a research method that randomly splits participants into groups, applies an intervention to only one, and compares outcomes. It’s the same method used in clinical drug trials. Its strength is that it can establish causation, not just correlation. ↩

3. Billable Hour: a system in which lawyers or consultants log the hours worked for a client and bill by multiplying that time by an hourly rate. It was the standard revenue model in the legal and consulting industries. ↩

4. Per-seat Pricing: a pricing scheme based on the number of users (“seats”) using a piece of software — something like “₩20,000 (~$14.4) per person per month.” The problem is that when AI reduces headcount, vendor revenue shrinks along with it. ↩

5. Outcome-based Pricing: a pricing scheme that charges for actual results produced, rather than usage or seat count. It uses units like “per resolved ticket” or “per lead generated.” ↩