Why I'd Log 50 Hours for a 30-Minute Job · Issue #180

Opening

Dear reader, let me start with a question.

There’s a developer who charges ₩60,000 an hour. A client asked her to build a feature, and it happened to be something she’d built before — three years earlier. Back then, it took a full 50 hours. This time, she pulled the old code out, dropped it in, and it worked, all in 10 minutes. It passed the tests too.

Should the company pay for 50 hours, or for 10 minutes?

The moment the answer is 10 minutes, that developer will never reuse old code again. The moment the answer is 50 hours, the contract has already stopped being hourly. Whichever way you answer, the contract breaks.

Let me give you the conclusion up front. This isn’t a matter of personal honesty. It’s a structural accident, happening at the exact spot where the premise that hours invested equal value created has collapsed. And right now, this accident is happening simultaneously in the human labor market and the software market.

🔴 Four Moments When Hourly Billing Logically Collapses

Hourly billing works perfectly under certain conditions — when hours invested are proportional to output. The problem is that in knowledge work, that proportionality breaks down constantly. Let’s look at four scenes.

First, the zero-hour idea. An idea that would change the company’s entire cost structure comes to you in the shower. No research, no meetings. How much should hourly billing pay for this idea? Zero hours means zero won. Because companies don’t buy ideas separately.

Second, reuse. This is the case above — a 50-hour asset bolted on in 10 minutes. Hourly billing has no way to convert the value that those past 50 hours created into today’s 10 minutes.

Third, agent delegation. A 10-hour task gets handed to an AI agent; you review the plan, check the results, run the loop a few times, and finish in 30 minutes. Those 30 minutes have the skill of knowing what to delegate compressed inside them — and hourly billing can’t read that compression.

Fourth, the failed 10 hours. Only after spending 10 hours do you realize the direction was wrong from the start. Is this billable? Hourly billing recognizes these 10 hours as legitimate labor. The three scenes before it, it doesn’t.

You can see the inversion here. Hourly billing is generous toward time spent wandering, and stingy toward time that’s been compressed. It’s a contract where the more skilled you become, the more you lose.

So the contractor is left with only two options: refuse hourly clients altogether (losing half the potential customer base), or pad the timesheet.

Let me be clear — I don’t recommend the second option. But when a structure consistently rewards one specific behavior, that’s not a problem with the people doing it. It’s a sign the structure is badly designed. A contract that runs on personal conscience isn’t really a contract. It’s a hope.

The Company’s Response Was Surveillance, and It Doesn’t Work

To stop padded timesheets, companies reached for surveillance tools — software that captures screenshots every 15 minutes, counts keystrokes, and demands an activity log every hour. The industry calls this bossware¹. According to market research, 78% of employee monitoring tools include a screenshot feature.

Does it work? There’s a study I found quite interesting here.

It’s a randomized controlled trial² published at NBER in 2025 by Namrata Kala of MIT and Elizabeth Lyons of UC San Diego. In an online labor market, they randomly assigned digital surveillance, split workers into groups where the reason for the surveillance was explained or not explained, and measured performance.

The result has two layers. First, surveillance itself had no statistically significant effect on performance, on average. The common assumption that people work harder when watched wasn’t supported by the data. Second — and this is the more important part — when surveillance was introduced or removed without explaining why, worker output dropped significantly. It wasn’t the act of surveillance that hurt performance. It was the failure to explain why.

So surveillance tools don’t prevent dishonesty — instead, they erase the accountability of explanation between managers and workers. And what this method filters out isn’t dishonest people. It’s people with bargaining power. Skilled people with other options refuse the screenshot conditions and leave. What’s left are the people who have to accept those terms anyway. By tightening surveillance, the company has, in effect, narrowed its own talent pool.

“Then why not just charge three times the hourly rate?” That doesn’t work either. When someone charging ₩300,000 an hour sits next to someone charging ₩100,000, a company judging by price picks the latter — even if the latter logs three times the hours and the total ends up the same. An honest rate hike gets punished by the market.

🌏 The Same Disease Is Spreading to the Software Industry

If everything so far has been about people, what follows is about software. Surprisingly, the symptoms are identical.

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

In a business that sells time, once AI cuts the time down, what do you sell?

The math: if a $300-an-hour lawyer finishes a brief that used to take 25 hours in 10 hours with AI, they’d need to raise their rate to $750 an hour just to keep revenue flat. According to the 2025 Legal Trends Report, 74% of law firms’ hourly-billed work is exposed to automation. The very root of revenue is exactly what automation is aimed at.

But this wasn’t a disease unique to law. The exact same thing is happening in the SaaS industry.

Per-seat pricing⁴ is collapsing. The share of SaaS companies using seat-based pricing fell from 21% to 15% over 12 months. Meanwhile, hybrid models blending subscription fees with usage climbed 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 “seat count,” then the more efficient the customer gets, the less the vendor makes. It’s a structure where the customer’s success becomes the vendor’s loss.

So “per-resolution” pricing emerged. Zendesk charges $1.50 per automated resolution by its AI agent (under committed pricing). Intercom’s Fin charges $0.99. Instead of billing per agent per month, they bill per resolved ticket. They’ve started pricing outcomes, not time.

And let me layer in a number from two issues ago.

A World Where Robots Cost $5 an Hour: What Do Humans Do?

The hourly operating cost of a US warehouse robot is $5.71 — about a third of the average hourly wage of a human warehouse worker.

Line up these three scenes, and one picture emerges. The price of the unit called “time” itself is collapsing — whether it’s human time, seat time, or robot time.

Why Did “Time” Become the Billing Unit in the First Place?

Let’s take a step back. Hourly billing isn’t a law of nature. It’s a legacy of the factory.

In the factory, hours invested and output were nearly perfectly proportional. Stand at the line for 8 hours, and 8 hours’ worth of parts came out — time was an excellent proxy metric, and easy to measure besides. 20th-century knowledge work adopted this unit without much thought. Law firms, consulting firms, agencies, freelance platforms — all of them.

That proxy has now snapped. The correlation between time and value in knowledge work was always loose, and AI has cut that loose thread entirely. Let me bring back last issue’s question: “In a business that sells time, once AI cuts the time down, what do you sell?” This was never just a question for the legal industry. It’s a question for everyone who gets paid by the hour.

📊 So Should We Just Switch to Outcome-Based Pricing?

Let’s not jump to conclusions here. If time doesn’t work, pay for outcomes instead — that’s an intuitive answer, but the data shows outcome-based pricing⁵ has walls of its own.

Measurement and attribution are hard. Real outcomes — revenue growth, cost savings — have multiple inputs. It’s hard to tell whether AI contributed, the client’s own team contributed, or the market was simply good. That’s why, as of 2022, only 17% of enterprise SaaS vendors had adopted genuine outcome-based pricing.

Sales slows down. One analysis found that outcome-based contracts — which come bundled with baseline measurement, PoCs, and legal safeguards — extend the sales cycle by 20-30%.

Finance hates it. 64% of SaaS finance executives named “unpredictability” as their top concern about outcome-based models. If revenue is tied to the customer’s outcomes, you can’t even plan your own company.

Only mature products can pull it off. 78% of companies that succeeded with outcome-based pricing had had their product on the market for over 5 years. A new team has no track record with which to promise outcomes.

Project-based pricing has the same problem. If requirements change, you have to renegotiate; drawing up a quote takes days on its own; and if the quote is off by 3x, the contractor eats the loss.

So the real-world answer isn’t pure outcome-based pricing — it’s hybrid. A base fee plus an outcome-linked component. The industry expects the share of hybrid models to climb to 61% by the end of 2026. That means outcome-based pricing won’t immediately replace hourly billing where it collapses — an awkward hybrid will dominate for a while.

Oswald’s Lens

Honestly, I rarely use hourly or project-based billing myself. I settle accounts weekly or monthly, and my reports are either very short or don’t exist at all.

There’s one thing I’ve learned doing consulting and GTM strategy work: measurement has a cost of its own. People usually assume “accurate measurement” is free. It isn’t. There’s the time spent filling out timesheets, the time spent writing reports, the manager’s time reviewing them, and the cost of the feeling of being watched eating away at performance. When you actually run the numbers on hourly projects, it’s not rare to find this administrative overhead eating up 10-20% of the billed amount. You’re spending 15% on overhead to raise the accuracy of the result by 1%.

So here’s my choice: the alternative to hourly billing isn’t outcome-based pricing — it’s the price of trust. Trust has a measurement cost close to zero. Every few weeks, I look at the value that person has created — not just completed tasks, but ideas proposed, time spent helping colleagues, documentation left for the team. I don’t ask how many hours they worked. I already know, every day, anyway.

But let me be honest about the limits. This model doesn’t work just anywhere. It only works at a scale where a leader can see what team members are doing with their own eyes, day to day — roughly a startup of around 10 people. Beyond that, a leader ends up reaching for a proxy metric, and the easiest proxy metric is, again, time. And from a contractor’s perspective, propose this model and half your clients walk away, because you can’t answer the question: “So what exactly am I buying?”

Finally, let me answer the hardest question. If someone who’s performed well for months did nothing last week, should you pay for that week?

I do. What I’m buying isn’t that week’s output — it’s the state of that person staying attached to the team. Output is inherently uneven. Some weeks it’s zero, some weeks it’s ten times the norm. If I dock pay for a zero week, I’ve effectively gone back to hourly billing. But this answer only holds for someone who has already built up a store of trust. If zero weeks keep repeating and you keep paying anyway, that’s not trust — that’s neglect. The real cost of a trust-based model is right here: the manager can’t dodge the judgment call. This might be the real reason hourly billing is popular. Because the number makes the judgment call for you.

Closing

To sum up:

Hourly billing is generous to time spent wandering and stingy toward time that’s been compressed. As AI widens the scope of compression, this contract pushes out the skilled. Surveillance tools don’t solve this — if anything, they filter out people with bargaining power first.

The same collapse is happening simultaneously in the legal industry’s billable hour and SaaS’s per-seat pricing. The price of the unit called time itself is collapsing.

That doesn’t mean outcome-based pricing is a cure-all. Measurement and attribution are hard, and finance hates unpredictability. For now, hybrid is the realistic answer, and if you’re a small team, trust-based settlement is one more option.

If you’re currently billing — or being billed — by the hour, tell me in the comments about the moment in that contract when you felt “logging it honestly means I lose.” Was it reuse? Agent delegation? Or the failed 10 hours? Once enough cases come in, I’ll organize them by type in a future issue.

💬 Share your experience with this question in the comments · 📨 If you have a colleague who bills by the hour, send this along

References & Further Reading

Primary sources

Namrata Kala & Elizabeth Lyons, “The Effects of Digital Surveillance and Managerial Clarity on Performance”, NBER Working Paper 33348, 2025. : The finding that it’s the failure to explain — more than surveillance itself — that hurts performance is the core evidence behind today’s piece. Even just the abstract makes the argument clear.
Monetizely, “The 2026 Guide to SaaS, AI, and Agentic Pricing Models”, 2026.1. : Covers both the collapse of seat-based pricing and the structural limits of outcome-based pricing. The ‘Structural Limits’ section toward the end is the most useful part.
Thomson Reuters Institute & Georgetown Law, “2026 Report on the State of the US Legal Market”, 2026.1. : This is the original report where the phrase “productivity-revenue paradox” appears.

Background

Zendesk, “Guide to AI Agent Per-Resolution Pricing” : 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 sheet that sells outcomes, not time.

In a Business That Sells Time, What Do You Sell Once AI Cuts the Time? : The legal-industry section of today’s piece is an extension of this issue.
A World Where Robots Cost $5 an Hour: What Do Humans Do? : A story about just how far the price of time has fallen.

The author, Kwangseob Ahn, is a professor of business administration at Sejong University and lead consultant at OBF (Oswarld Boutique Consulting Firm). He teaches statistics and data analysis — business data management and business analytics — while leading GTM and AI strategy consulting in the field, designing the seam between technology and business. He has published academic research on a memory architecture for AI dialogue systems (HEMA) and runs Daily Arxiv, a daily curation of global AI papers. He holds a master’s from Korea University’s Graduate School of Technology Management and a KMBA. He is the author of Homo Brainless: The People Who Outsource Their Thinking.

Footnotes

Bossware: Surveillance-style workforce management software that automatically logs employees’ screens, keystrokes, and login times. The term combines “Boss” and “Software,” and it’s generally used with a negative connotation. ↩
Randomized Controlled Trial (RCT): A research method that randomly splits participants into groups, applies an intervention to only one group, and compares the results. It’s the same method used in clinical drug trials. Its strength is that it can establish causation, not just correlation. ↩
Billable Hour: A billing method where lawyers or consultants log the hours they work for a client and charge by multiplying those hours by an hourly rate. It was the standard revenue model in the legal and consulting industries. ↩
Per-seat Pricing: A billing method that charges based on the number of users (seats) using the software — something like “₩20,000 per person per month.” The problem is that when AI reduces headcount, vendor revenue shrinks right along with it. ↩
Outcome-based Pricing: A billing method that prices actual results created, rather than usage or seat count — units like “per ticket resolved” or “per lead generated.” ↩