stop selling subscriptions. sell work minutes.

Here's a number that should ruin your day.

The average knowledge worker is "productive" for 2 hours and 53 minutes per day.

Not 6 hours. Not 5. Not even 4. Two hours and fifty-three minutes of actual output across an entire 8-hour workday. The rest is meetings about meetings, Slack threads that go nowhere, searching for the thing you saved somewhere but can't remember where, and context switching between 350 SaaS apps (real number, per Okta) at a cost of 9.5 minutes of refocus time per switch.

If you're paying a knowledge worker $120,000 a year - which is pretty median for tech in a major city - their loaded cost is roughly $75/hour. that's salary plus benefits plus overhead plus the office space plus the software subscriptions plus the snacks in the kitchen.

But they're only productive for ~3 hours.

So the effective cost of a productive hour from that person isn't $75. It's closer to $165. Almost $3 per productive minute.

You are paying $3/minute for a human to do work. And most of what they're doing during those productive minutes isn't even the work you hired them for. It's context transfer - moving information from one place to another, getting people up to speed, formatting spreadsheets, compiling reports, copy-pasting between tabs.

The actual thinking? The judgment calls, the strategy, the creative work? That's maybe a third of the productive time. So your REAL cost for an hour of human judgment is approaching $500/hour.

and we wonder why companies keep laying people off.

I've been thinking about this for six months because Tyler and I had to figure out how to price something that doesn't have a precedent.

We built an AI that does work. Not "helps you do work" like every other productivity tool - actually does the work. You give it a task, it connects to your apps, it executes, you review. We wrote about the workflow and how we apply it to everything from code to content to admin.

So how do you price that?

The SaaS playbook says: charge per seat per month. $20/user/month. $50/user/month. Enterprise tier with SSO for $200/user/month. Everyone on the same plan gets the same features, regardless of whether they use 2% of them or 90% of them.

This model has been the backbone of software for 20 years. It's generated hundreds of billions in market cap. Salesforce, Slack, Notion, Asana, Monday.com - all per-seat.

And it's dying.

61% of SaaS companies now have a usage-based component, up from 34% in 2021. Usage-based models see 25% faster revenue growth and significantly higher retention (130% vs 110% net revenue retention). Gartner predicts 40% of enterprise SaaS spend will shift to usage- or outcome-based models by 2030.

Why? Because AI broke the per-seat model. AI agents don't sit in chairs. They don't need seats. They consume tokens and compute and they work 24/7/365. Charging "per seat" for a product that deploys AI agents on your behalf is like charging per passenger on a self-driving car. The metaphor doesn't fit anymore.

Zendesk figured this out early. They started charging $1.50 per automated resolution instead of per agent seat. Not per month. Per outcome. You only pay when the AI actually solves a customer's problem.

That's the model that makes sense in an AI-native world. Pay for work, not for access.

So we did something that felt obvious to us but apparently confuses people: we charge per work-minute.

Not per seat. Not per month. Per minute of actual AI work completed on your behalf.

And I've watched people's faces when they hear this. The first reaction is always: "that sounds expensive."

It's not. Here's why.

Remember Jan, my co-host's executive assistant? The one who made 200+ individual edits over 5 working days updating a spreadsheet with event attendee info? Tab-switching between LinkedIn, the RSVP list, and the spreadsheet. Copy. Paste. Format. Next row. For hours.

Jan is good at her job. Let's say she costs $35/hour fully loaded (which is generous for a skilled EA in a major city - the real number is probably $45-60).

That spreadsheet task took her roughly 15-20 hours across those 5 days. At $35/hour, that's $525-700 worth of human labor.

I did the same task in 3 prompts. Maybe 30 minutes of Doozy work-minutes.

Even at $2/work-minute, that's $60. For the same output. Done while I was drinking coffee.

$60 vs $600.

Now tell me the per-minute price sounds expensive.

OK but here's the thing that ACTUALLY keeps me up at night about this.

It's not the cost savings. The cost savings are obvious. It's what happens AFTER you realize the cost savings.

I run an AI consumer meetup in Seattle. I talk to founders and operators every week. And a pattern has emerged that I think reveals something much bigger than pricing models.

The smartest companies I've seen aren't replacing their EAs with AI. They're doing something much more interesting: they're giving their EAs AI tools and watching them become 10x.

Think about it. You have a person who already understands your business, already has your context, already knows how you like things done. They have the taste. They have the judgment. What they DON'T have is the ability to tab-switch between LinkedIn and a spreadsheet 200 times without wanting to die.

The AI handles the context transfer. The human handles the thinking.

Same split. Same discuss-plan-implement-review workflow. But now the human who was spending 60% of their day on "work about work" is spending 60% of their day on the work that actually matters. The judgment. The relationship management. The strategic thinking. The stuff you hired them for.

The math gets insane when you lay it out:

US Executive Assistant: $35-60/hour. Productive for ~3 hours/day. Effective productive rate: $95-160/hour.
Overseas VA: $7-20/hour. Same productivity constraints. Effective productive rate: $19-53/hour.
Either of them WITH AI tools: suddenly productive for 6-7 hours/day because the context transfer is gone. Effective productive rate drops by half. Output doubles or triples.

The cheapest option isn't "replace the human with AI." It's "give the human AI and remove the garbage from their day."

And HERE is where the pricing psychology clicks.

If you're paying an EA $40/hour, and they spend half their day on context transfer, you're burning $20/hour on clipboard work. That's $160/day. $800/week. $41,600 per year on copy-paste.

If an AI subscription that does all the context transfer costs you $200/month? That's $2,400/year replacing $41,600/year of wasted human time.

You just gave your EA superpowers AND saved $39,000. And they love you for it because you just deleted the worst part of their job.

I keep coming back to a question that I think is the most important question nobody's asking:

How do you make people feel good about this shift instead of threatened?

Because the natural reaction is fear. "AI is going to take my job." And honestly? If your entire value is context transfer - if you're a human clipboard - that fear isn't irrational. Context transfer is dead. I wrote 3,000 words about why.

But context transfer was never the good part of anyone's job. Nobody became an EA because they loved copying data between spreadsheets. Nobody became a marketer because they loved reformatting the same report for the third stakeholder this week. Nobody became a developer because they loved searching for the right file in a 300k-line codebase.

The good part of the job - the thinking, the problem-solving, the creativity, the relationship-building - that part isn't going anywhere. AI can't taste. AI can't judge. AI can't decide whether a deal is worth pursuing based on a gut feeling developed over 15 years of negotiations.

What AI can do is remove every single thing that was standing between you and the good part.

The framing shouldn't be "AI replaces you." It should be "AI deletes the worst 60% of your job so you can spend all day on the best 40%."

That's not a threat. That's a promotion.

There's a specific company I'm watching that I think shows where this is all going.

I can't name them - NDA stuff - but they're a mid-size company with about 150 employees. they were spending roughly $17,000 per office worker per year on pure administrative overhead. Not salaries. Just the productivity cost of coordination - the meetings, the status updates, the context switching. For 150 people, that's $2.55 million per year in invisible waste.

They rolled out AI tooling across their operations team. Not as a replacement. As an augmentation. Their ops team of 8 people went from processing ~200 tasks/week to over 600. Same team. Same hours. Same salaries. Triple the output.

The cost of the AI tooling: ~$15,000/year.

The value created: measured in the millions.

This is the part that breaks people's brains. The ROI on AI tooling isn't 2x or 5x. For the right tasks, it's 100x+. Because you're not paying for intelligence. Intelligence is commoditized. You're paying for the removal of friction. And friction was eating 60% of everyone's day.

Let me bring this full circle.

The SaaS industry spent 20 years perfecting a pricing model that charges you for access to tools. $20/seat/month for the privilege of organizing your own work.

But in a world where AI can DO the work - not organize it, not track it, actually do it - charging for access makes no sense. You should be charging for outcomes.

Work-minutes instead of seat-months. Per-resolution instead of per-agent. Pay for what gets done, not for what you have access to.

This feels obvious to me. But I think the reason it confuses people is that it requires a fundamentally different mental model. You're not buying a tool. You're not buying software. You're buying work. Actual labor. At a fraction of the cost of human labor for the same output.

$2/minute for AI work vs $3/minute for productive human work.

Except the AI minute is fully productive. No context switching. No Slack breaks. No meetings about meetings. 60 seconds of pure output.

And if that output saves your $120k knowledge worker 2 hours a day? That's $30,000/year in recaptured productivity for maybe $3,000/year in AI work-minutes.

10x ROI. Before you even count the things the human does with those recaptured hours.

I'm a 22-year-old founder. I fully understand how weird it is for me to be writing about pricing philosophy and the economics of knowledge work.

But I've seen the numbers from the inside. Tyler and I run our own business on Doozy. We used it to do our tax prep and saved $1,200. We use it to source leads while we sleep. We use it to manage a 300k-line codebase with zero engineers. We used it to create a week's worth of content in 30 minutes that would've taken a week of effort before.

Those aren't hypotheticals. Those are receipts.

And when I watch Jan make 200 spreadsheet edits over 5 days and I do the same thing in 3 prompts over coffee, I'm not thinking "AI is going to replace Jan." I'm thinking "what could Jan do if she never had to make those 200 edits again?"

The answer is: the things that require a human. The things that require taste, and context, and judgment, and generosity, and all the other things that make people irreplaceable.

The edit-the-spreadsheet part was never the valuable part. We just pretended it was because we didn't have an alternative.

Now we do.

The question isn't whether the economics work. They do, embarrassingly well. The question is whether we're brave enough to reorganize work around the thing that's left after the clipboard work disappears.

I think we are. But only if we make people feel like they're being upgraded, not replaced.

That's the whole game.

-parsa