26:03Greg Isenberg · Talking Head
AI Agents are the new SaaS
A step-by-step playbook for building and selling AI agents as done-for-you labor instead of software seats.
July 1stThe argument in one line.
Because AI makes software cheap to build, the tool should be free, and the money moves to upselling AI agents and human services on top of the data and network effects the free tool collects.
Read if. Skip if.
READ IF YOU ARE…
- A founder or indie builder deciding how to price a new software product now that AI makes the code itself cheap to produce.
- Someone running a per-seat SaaS who wants to understand how the pricing model is likely to shift over the next few years.
- A builder hunting for niches where people still pay for an ordinary tool and a free version could take over the market.
SKIP IF…
- You want a step-by-step build tutorial; this is a single whiteboard model with one consumer example and no implementation detail.
- You sell mainly to Fortune 500 buyers, where he says per-seat pricing mostly survives.
The full version, fast.
For 15 years SaaS meant paying engineers to build a tool and charging per seat. AI now makes that tool easy to build, so the model is breaking. Isenberg's replacement is smart SaaS: give the tool away free, take data and network effects as payment, then upsell intelligence from AI agents, charged by token or by outcome, and services from humans, charged weekly or monthly. A free calorie tracker that sells personalized meal-plan agents is his example. SMBs and individuals trend toward free, while enterprises still pay something per seat for security and reassurance.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 00:47
01 · Intro
States that SaaS is dead and promises the clearest breakdown of where it is going.
00:47 – 02:56
02 · SaaS 1.0
How per-seat SaaS worked for 15 years, the Slack example, and why AI breaks it.
02:56 – 05:12
03 · SaaS 2.0: the new business model
Build the tool, give it away, and take data and network effects as the payment.
05:12 – 06:18
04 · How to monetize
Upsell intelligence from agents and services from humans, the idea behind Sequoia's services-as-software thesis.
06:18 – 07:22
05 · Calorie tracker example
A free consumer app with paid agents for meal plans and a paid weekly human coach.
07:22 – 08:58
06 · SaaS is dead, long live SaaS
Product studio idea, pricing by token, outcome, week or month, and a larger market.
08:58 – 10:31
07 · SMB vs enterprise
Free trends dominate SMBs and individuals; enterprises still pay per seat for security and comfort.
10:31 – 11:12
08 · Closing thoughts
Asks for engagement and promises a follow-up on niches and opportunities.
Lines worth screenshotting.
- When AI can build the tool in days, the tool stops being the thing customers pay for.
- In smart SaaS the user pays with data and network effects instead of money, and those become the moat.
- Dumb SaaS is only a tool. Smart SaaS is a free tool with intelligence sold on top.
- Agents can be priced by token or by outcome, so the customer pays only when the work gets done.
- Human services layered on a free tool can charge weekly or monthly, like a coach attached to a free tracking app.
- Per-seat pricing capped the market because many people simply don't want to pay for tools.
- The market for people accomplishing things is larger than the market for people using tools, so the total SaaS market grows.
- A product studio shipping many free tools into open niches can feed one upsell engine of agents and humans.
- Look for niches where people already pay for a tool under the old model and give them that tool for free.
- Enterprises will still pay per seat, partly for security and partly because paying feels like being taken care of.
- SaaS is not dead period; the dumb per-seat version is dead and a smarter version is replacing it.
Give the tool away, sell the intelligence
WHAT TO LEARN
When AI makes the tool cheap to build, the paid product shifts from access to the tool toward agents and human services that do the work for the customer.
- Per-seat pricing worked because building quality software was hard; once AI collapses that cost, the tool alone stops justifying a monthly fee.
- A free tool still pays you, in proprietary data and network effects that improve the product and build a moat competitors cannot easily copy.
- Revenue comes from upselling intelligence: AI agents priced by token or by outcome, and human services priced weekly or monthly.
- A calorie tracker shows the shape: free logging, then paid agents for personalized meal plans, then a paid weekly human coach.
- The addressable market grows, because far more people want a result accomplished than want to operate a tool themselves.
- Enterprises will keep paying something per seat for security and reassurance, so SMBs and individuals are where free matters most.
Terms worth knowing.
- Per-seat pricing
- A software billing model where a customer pays a fixed fee, such as $9.99 or $29.99, for each person who uses the tool.
- Dumb SaaS
- Isenberg's label for the old model: software that is only a tool, paid for per seat, with no built-in intelligence.
- Smart SaaS
- His label for the new model: a free tool that gathers data and network effects, with paid AI agents and human services sold on top.
- Network effects
- A product gets more valuable as more people use it, which makes it harder for a competitor to pull users away.
- Pay per outcome
- A pricing method where the customer is charged only when an AI agent actually completes a defined piece of work.
- TAM
- Total addressable market, the full amount of revenue available if everyone who could buy the product did.
Things they pointed at.
06:20linkSequoia, services as the new software
Lines you could clip.
03:38
“Give away the CRM, give away the dashboard, give away the app, get people onto the tool.”
08:10
“It's not that SaaS is dead, period. It's SaaS is dead, long live SaaS.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
SaaS is dead, everybody. And in today's video, I'm going to basically break down why I think that's the case, where the opportunities are, and what's really happening here. And I think the people that understand where SaaS is really going, I think they're going to understand how they can reorient themselves to build something that people actually want.
So I'm going to make this the clearest breakdown on the internet for... what's happening with SaaS, where it's going, why it's dead, quote unquote, why I think it's evolving into something else, what that something else is. And yeah, by the end of it, you're going to understand it.
How did SaaS typically work for the last 15 years was people basically built some code. That was the software, right? Software as a service.
They created some code, and that code was essentially a tool. So the hard part about SaaS was, can you get enough smart engineers in a room, pay them to code something out that was high quality, that worked, that was a tool for a human being to go and use, right? And in exchange for using that tool, you paid money.
And the whole business model was a per -seat basis. So you would pay per seat $9 .99, $29 .99. And you would go and access that tool as a service in the cloud.
And that was basically how it worked. And for a long period of time, this model made people billions of dollars. I mean, there's so many examples of SaaS companies that made billions of dollars.
Slack, for example, I think sold for $29 billion. This model really worked for a while. And the reality is it does not work anymore.
AI makes the ability to create the SaaS super easy. I think when we hit the Opus 5 .5, GPT -6, Astra, that level of model, even Argon, Carbon, these models are so good that building the SaaS is not hard.
Cool, Greg. That's very academic of you. How can I actually go and build a SaaS that works in today's environment?
Let me explain where we're going with all of this in the simplest way I possibly can. The new version of SaaS, it has actually a different business model. In a world where the tool is so easy to create, and by the way, we're going to talk about the ICP and who to target later, so stick around for that.
You're still building a tool, right? You may be using AI to create the tool, or AI with one human or two humans instead of... no AI and 20 humans, but you're creating a tool.
But the business model is actually free. So you're actually giving away the tool. I don't know why more people aren't actually doing this.
I tweeted this the other day, like give away the CRM, give away the dashboard, give away the app, get people onto the tool, find niches where... There's something in this old model that people are paying for and literally give away the thing to the person for free. And we're going to get to how you can actually make money from this in a second.
But the value exchange here initially isn't the money, right? So it used to be that you would go and pay for this. So what are you giving here?
You're not giving money. Well, you are giving data. So that is going to help you make a better tool.
You are creating network effects and that's going to help you create a moat. So that's the idea. You're in the free tool business.
And you're going to be trying to acquire proprietary data, which you can eventually sell or use to make your tool a lot better. And then a tool that helps create network effects. Now, how do you actually make money from this tool then?
Well, and by the way, I actually think that the market in this world, we'll call it SaaS 2 .0.
or SaaS 1 .0, this is what I would call dumb SaaS because it's just a tool. And this is what I would call smart SaaS because it's intelligent. And I'll get to that in a second.
You're basically upselling intelligence. You're upselling intelligence. So you're basically saying, hey, I'm going to give you away this CRM, say.
But if you want the CRM to go and do something for me with AI agents, that's going to cost you money. And you can upsell that. The other thing that you can upsell is some service.
So a service provided by humans. And this is intelligence provided by agents. So this is, you know, when Sequoia talks about, you know, trillion dollar service is the new software, they're talking about this.
And they're basically saying like, hey, there's going to be a tool and that's going to get you in and that's going to bring data and network effects. And if you want a upsell, you can have a service. Now, what's a concrete example of this?
In the consumer space, it would be something like a calorie tracking app. The tool is free, but if you want to actually use agents to go and help you make personalized meal plans or give you more precise calorie tracking, you would have to pay for that. Or if you want it to be upsold to a weekly...
to help you get on track, that is something that you would pay money for. So it's pretty interesting. I think why it's interesting is this was kind of limiting.
The tool per seat was kind of limiting because not everyone wants to pay for tools. But if you create essentially a product studio around free tools, get people in, get their data, get their network effects, and then upsell them either to... ai agents or to human beings um there's a lot of niches that have are are just wide open right now that i think you can go and build stuff in um so it excites me so when i see things tweets and stuff like that around like sas is dead period it's not that sas is dead period it's sas is dead long live sas uh this is like you know we're evolving from this which is no longer a thing anymore.
It's dead. Don't create this, right? The other thing to mention is the business model of, okay, it's free over here, upsell to agents.
How do you charge for that? I mean, you can either charge by token, you can charge by pay per outcome, so if the agents actually do something. And then for human beings, it could be weekly or monthly.
So I think that the total TAM for total addressable market for SaaS is now bigger because the market for people accomplishing things is just a lot bigger than people actually using a tool to complete something. When I posted about this whole smart SaaS thing and how SaaS is going to be free, my point is that SaaS is trending.
towards being free. So I think that, here, this is sort of the price, where things are going now. I think for a lot of SMBs, prosumer stuff, it's going to be free.
But I do want to say, when it comes to enterprise, when I'm talking about enterprise, I'm saying like Fortune 500, 2000, those companies, they still might pay, Something per seat. I think that number goes down over time.
And I still think that there's upsells to humans and superintelligence.
But the point is, some enterprises are just not going to want free stuff. They're going to want to pay. They're going to want to pay for the security.
They're going to want to pay because it just feels good to pay because it feels like you're getting taken care of. So I think that enterprise, they will still pay something. But again, the main point I'm trying to make is it's going towards free.
I think for SMBs, you're going to see it becoming more and more free. And individuals too. I hadn't seen anyone really...
clearly explain this, so I wanted to put it in here. I think I'm happy to do more on this. Just comment below, like, comment, and subscribe.
And I can do more on where SaaS is going, the niches, the opportunities. If this gets 5 ,000 likes, comments, and subscribes on YouTube, I will go and do more of that. But yeah, I think this is a really interesting time to be creating SaaS.
This is the map. Enjoy. Go and do it.
Thank you for watching. Thank you for listening. Have a creative day, and I'll see you next time.
The Hook
The bait, then the rug-pull.
Greg Isenberg opens with the blunt claim that SaaS is dead, then spends eleven minutes qualifying it into something more useful: the per-seat tool is finished, and a free tool with paid intelligence on top is what comes next.
Named ideas worth stealing.
05:30model
Dumb SaaS vs Smart SaaS
- SaaS 1.0: tool, per seat, person pays
- SaaS 2.0: free tool, data and network effects, paid agents and humans
Old model charges per seat for a tool. New model gives the tool away and charges for agent intelligence and human service on top.
Steal forpricing a new software product in a niche
How they asked for the click.
VERBAL ASK
10:40subscribe
“If this gets 5,000 likes, comments, and subscribes on YouTube, I will go and do more of that.”
A conditional engagement threshold tied to a promised follow-up video.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
