Hey y'all, I don't think I'm a SaaS buyer anymore.

A founder friend pitched me his new product recently. It checks a lot of boxes: he’s a good founder, he’s got early traction from quality logos, and the product itself looks like it solves a real problem well.

A couple years ago I would've signed up on the call.

This time I spent the whole demo doing math in my head on whether my agent could just build it cheaply enough for it to not make sense to add a subscription for.

The toughest part of the call was that he did have something else I would’ve paid for… but he’s getting rid of it.

This week I want to talk about who's actually buying your product now, how they’re deciding whether to try it, and how you can protect yourself from the changes.

Build your company's brain with me, live.

This is the setup behind the math I was doing on that call: one agent that knows every call, email, client and decision we've made.

This Thursday I'm building one from scratch, live. Build yours alongside me, or just watch and bring your questions.

(If you’re going to build one alongside me, I recommend setting up a Hermes account beforehand.)

The first thing I asked was whether my agent could connect via API or at least an MCP.

The product had an AI-powered onboarding, where you give it a bunch of context about your business and it fills out a more robust underlying profile about you.

I’ve seen more products move to this high-friction onboarding in the last couple years.

LLM-powered tools want to make things more personalized for you, but founders are less willing to invest that much in onboarding for a new tool they’re just trying out.

Company brains are the answer to this.

As more founders set up a brain that already has all that context about them (and, realistically, more than they’d be able to give in a tool’s onboarding), this friction will go away.

The brain can give the tools whatever context it needs (that the founder allowed).

(That's the thing I'm building live on Thursday, btw.)

So when I look at a new tool now, the question is a lot simpler: how much would it cost for my agent to build this, and how does that compare to the monthly subscription cost for the tool?

The tradeoffs are clear.

If you build in-house you get:

But you also take on responsibilities:

What I realized on the call was that on that call, I wasn't really the customer. My agent was.

It would do the setup.

It would decide whether a product beats what it could build itself for a reasonable cost.

And it would be the one using it every day, not me clicking around a dashboard.

So the bar you're selling over is my agent building it for me, and the incremental knowledge you give it.

Your tool isn’t going to add functionality, most likely, that the agent couldn’t build itself.

Rather it’s whether you can deliver that functionality at a low enough cost to make building a bad economic decision, or if there are network effects built into your product that someone’s own version wouldn’t be able to match.

A good example of this is Slack.

Slack is expensive.

The UI has gotten worse over time.

It isn’t agent-native (my agent is shoehorned into as a bot).

But the chances of us moving off of Slack feel very low, because we use Slack Connect channels with the vast majority of our clients.

They have Slack, so we have Slack to meet them where they are, even if new entrants like Ando (or a self-built version) seem better for the way our own team operates internally.

Another place where SaaS still wins is beehiiv.

My friend Tyler Denk, cofounder of beehiiv, made the case for buying really well this week.

His point was that building it is the easy part. Then you own it forever. And when your emails start landing in spam, you're on your own.

Under the hood there's deliverability, bot filtering, click verification and relationships with every inbox provider. That's years of specialized infrastructure, and I don't want that job. Same reason I'm not rebuilding Stripe.

So I lied a little up top. I'm still a happy SaaS buyer when the hard part is under the hood.

But a lot of software isn't like that. It's a workflow sitting on top of your data: a form, a list, a dashboard, some scoring, a few reminders.

That's the layer agents are eating.

Back to the pitch.

While my buddy was walking me through the product, I was clicking around his site. On the pricing page there was a done-for-you option where his team runs the whole thing for you.

That's the one I'd actually pay for.

I asked about it. He said they offered it to learn the process so they could productize it, and they're phasing it out.

From a traditional venture lens, that makes total sense.

But the world has changed. He told me I was their exact target customer, and what I need is the exact opposite.

I can build most of the product pretty cheaply now, but it’d be another system for me to monitor, and in this product’s particular case it’s one that does require some human judgement. So I’d love to hand it off.

There’s a bigger conversation here, but it’s why I’m building Rye Valley to deliver services instead of software.

We find experts in their verticals, work with them on new concepts, and both power and help grow the businesses with our AI-native centralized operations.

In a world where founders and small teams can do more with less, we’re going to see them outsource more to experts who bring the years of domain knowledge they don’t have on their team and can’t trust they even know what “good” looks like when working with AI themselves.

His pricing page had a verticalized AI-native service business sitting right on it. And he's taking it down.

If you want to try this on your own stack, here's a quick test for every tool you pay for:

P.S. Want to see what "my agent could just build it" actually looks like? Save your seat here for Thursday’s company brain workshop.