Companies like Figma, HubSpot, and

Dualingo are all down more than 65%. And

the entire SASbased industry has lost

more than $300 billion in market cap

recently. And the most common rhetoric

out there is that AI is to blame for

this. If you look at different sectors

outside of SAS, like auto manufacturing

and credit services, they tend to trade

in low PE ratio as opposed to software.

For example, auto manufacturers like

Ford will easily trade on multiples like

8 to 12p, while software companies like

Dualingo can go up to 50 or even 200 PE

ratios because well, most software

companies actually don't make money in

the first place. And unlike highly

cyclical industries that depend on

supply chain, inflation, supply and

demand, and policies, SASbased companies

have always been easier in valuation

because the underlying math to calculate

the annual recurring revenue is just

subscription multiples. And not only

that, unlike auto industries, software

companies always had room to grow in

comparison. But despite all these facts,

Dualingo still lost 68% of their market

cap in just 6 months alone. This kind of

progression is leading some people to go

as far as saying that SAS is dead and AI

killed SAS through and through. But why?

How is AI exactly impacting this?

Welcome to Kale Bryce Code where every

second counts. Quick shout out to

OnSpace. More on them later. When we

look at SAS, it's the highest level of

abstraction where users pay monthly or

yearly fees to use services provided by

software. And unlike other offerings

like on-prem where you have to buy your

own equipment, networking server and

databases, SAS allows you to just log

into Dropbox or Google Drive to store

documents without setting up the

database or even log into HubSpot or

Salesforce to manage CRM without having

to worry about what's underneath it. But

how did AI exactly change this where

SASbased companies are losing a large

share in market cap? If you look at the

progression of AI like MCP back in

November 2024, A2A or agent to agent in

April 2025 and more recently skills in

October 2025. What we are seeing here is

not a replacement of applications but

rather an abstraction on top of

application by connecting these systems

by leveraging MCP connections using best

practices written in skills. And while

the common rhetoric is that AI just

killed software, in reality, we haven't

really seen AI replace Slack or AI

replace HubSpot, Salesforce, Dualingo,

Docu Sign, all the other applications

that we use on a daily basis. In fact,

agents are actually orchestrating on top

of solid build applications that's

already been built. So while AI hasn't

really replaced software as we know,

what this new level of abstraction is

forcing is when it comes down to

pricing. SAS companies typically charge

people by per user or per seat. For

example, if we go to Salesforce, their

pricing plan goes from free to $100 per

user per month. And this kind of pricing

model allowed investors to project the

growth by using ARR as a metric. But as

AI agents took on the role of

interacting with the applications where

before we had to learn the application

by clicking through websites and

navigating the applications, now agents

are essentially absorbing this entire

interactions away from us to agents.

Which means you could have one agent

with a single user subscription but now

shared across 10 people outside of the

system rather than having 10 different

users having their own subscription. And

the result of all of this is

commoditization of the application

layer. Since what used to be the remote

or competitive edge by providing an

intuitive application to users is now

becoming less relevant since agents can

connect to applications through MTP and

use their skills that explain the best

practices around how to actually

interact with the application. And this

kind of rotation is not only forcing

SASbased companies to switch from per

user pricing model to usage based

pricing model. And now what used to be

predictable valuation based on user

growth measured in ARR is starting to be

valued as a commodity based on usage or

supply and demand of agentic use cases.

And this all brings us to the next point

which is how the broad market is

changing because of this. Quick shout

out to OnSpace.ai. OnSpace is a platform

that lets you build mobile apps by

talking with its AI chats like this

where you can go from idea to publishing

on app store with a single click. Here

I'm sending a prompt called GPU cost

calculator where I can calculate the

projected cost of GPU usage. Once I have

the basic app generated, I asked the

follow-up question to fix the UI and UX

to be more friendly like this. After

that, I asked Onspace to add a login

screen and it generated a login screen

like this where I can sign up with an

email code being sent so I can apply

them so that I can access the app. Once

I'm happy with the app, OnSpace allows

me to publish it on the app store for

Apple like this. You can also download

the codebase directly, add payments and

databases on the app itself. OnSpace

also has a vibrant tutorials on their

YouTube channel that helps you with

anything from writing better prompts to

building a more secure and managing data

throughout the app. They also have great

support from AI and human support in one

day. I'll leave a link in the

description below. According to Goldman

Sachs projection, the total addressable

market for agents is expected to grow

above $50 billion while traditional SAS

is set to decline and 2026 is where we

start to see a divergence between the

two. And consequently, investment from

venture capitals is moving away from

traditional software products into

agents. That's not to say traditional

software companies aren't adopting AI

into their existing platform. According

to Deoid survey, more than half of the

companies are expected to spend up to

50% of their budget into AI automation.

In other words, instead of losing their

business to agents, a lot of companies

are choosing to also build their own

agents directly in their application. so

that users are inclined to continue user

service on their platform instead which

is seen in Salesforce agent force

service now now assist HubSpot also has

their own agent and well basically every

software company now offers agents that

know how to navigate their own

application out of the box. So what used

to be 50 to 100 PE ratios that investors

were willing to spend up to is now

reduced because investors are no longer

sure how much longer these companies are

viable beyond five or even 10 years down

the line as pricing models changes with

AI. And if you pair that with the fact

that last June around 10% of the

websites in the world were viodated

using lovable which goes to show that at

least in the application side how easy

it is to replicate existing

applications. That's not to say vioded

software will replace software since

well-written software requires more than

just pure vibe coding. But what it does

reveal is that SASbased companies are in

the position to change how they price

and how fast they need to innovate given

the pace of innovation that AI brings.

What do you think? Do you agree with AI

and agents complementing applications or

do you think that agents will replace

applications entirely?