The recent SaaS sell-off did not happen for no reason. The market was reacting to two real concerns, and both matter.
The first is that AI may weaken the traditional seat-based SaaS model.
The second is that the barrier to building software is falling fast.
Both are valid concerns. But the market is still making a mistake by treating all software companies as if they are the same.
That is where the misunderstanding begins.
Why the valuation reset happened
The reset in software valuations did not come out of nowhere. Investors started to realise that the old way of valuing SaaS may not hold as cleanly as before.
For years, the model was simple enough. More employees at customers meant more software seats. More seats meant more recurring revenue. That created a fairly understandable link between adoption and revenue growth.
Now that link looks less certain.
1. AI agents may reduce seat growth
If AI agents and automation tools start taking over repetitive work, some roles may disappear, and others may change. We are already seeing signs of this. Even SaaS companies themselves are doing it. They are automating repetitive internal work, reducing certain types of roles, and shifting hiring towards more value-added functions.
Salesforce CRM 0.00%↑ is one example people point to. The company has talked about using AI to automate some internal tasks while still investing in areas like sales. That matters because many software businesses have historically grown through seat expansion. If fewer people are needed to get the same amount of work done, then that old seat-based revenue model becomes less predictable.
That creates a valuation problem.
If revenue gradually shifts from seat-based pricing toward outcome-based, usage-based, or resolution-based models, investors have less certainty. They do not yet know what the new steady-state economics look like. They do not know how durable the pricing will be, how usage will behave, or what margins look like over time.
So part of the sell-off is really the market saying: we are no longer sure how to value these businesses the way we used to.
2. The barrier to building software has fallen
Tools like Claude Code and others are reducing the time, cost, and human effort needed to build software. You no longer need the same size engineering team to get a product off the ground. Development cycles are getting shorter. Similar products can be built faster. A lot of code can now be generated with AI assistance, and that percentage is likely to keep rising.
That does not mean the software is automatically better. That still has to be proven. But it does mean that the resource intensity of starting a software company is falling.
If it becomes easier, cheaper, and faster to launch a product, then competition increases. More teams can enter. More tools get built. More categories get crowded. What used to require a serious technical organisation may now require fewer people, less time, and less capital.
That is another reason valuations reset. Investors are asking whether software should still command the same multiples if the supply of software products increases and the cost of creating them keeps falling.
But the market is still oversimplifying it
Where the market gets it wrong is by taking those two concerns and applying them equally to all SaaS businesses.
That is too simplistic.
Not all software is equally exposed to these risks. Some software is shallow, easy to replicate, and lightly embedded. Some is deeply integrated into how an enterprise actually functions.
Those are completely different businesses.
Where the fear is valid
The fear is real for thinner software businesses.
If a product is basically a narrow interface on top of a database, with limited workflow depth and little integration into the customer’s operations, then yes, it is vulnerable.
A lot of standalone apps, lightweight productivity tools, and simple workflow products fall into that bucket.
If a customer can recreate something close enough quickly and cheaply, pricing power weakens. If the core value is mostly the interface, then the moat was never that strong.
For those businesses, lower barriers to entry are a real threat.
Where the market gets it wrong
The problem is that the same logic is being applied to large enterprise platforms, where the economics are very different.
For scaled enterprise SaaS, code is only one part of the value. Often it is not even the main part.
These businesses are not just selling software. They are selling systems that sit inside the operating structure of the customer. Over time, they become tied to workflows, approvals, compliance, reporting, audit trails, internal controls, training, and decision-making.
A coding tool can help write software faster. It cannot instantly recreate years of trust, integration, implementation experience, security review, legal accountability, and organisational adoption.
The real moat is not just the code
This is the part many people miss.
When they look at a big SaaS company, they assume the value sits mainly in the software itself. But in many enterprise platforms, the visible software is only part of the story.
The real moat often comes from:
deep workflow integration
switching costs
trust built over years
compliance and security approvals
embedded data relationships
global support and implementation
internal standardisation across large organisations
distribution and customer reach
That is not something a prompt can reproduce.
Yes, code is becoming cheaper. But enterprise trust is not. Institutional adoption is not. Deep embeddedness is not.
The data and trust issue matters more than people think
This becomes even more important once AI starts doing more real work inside companies.
These systems need data. Not just surface-level data, but real operational data. Customer information, legal contracts, employee permissions, financial records, security controls, internal workflows.
Large enterprises are not going to hand that over to a two-week-old startup just because the demo looks clever.
In regulated industries especially, vendor trust takes years to build. Procurement, legal, compliance, security, and internal audit all matter. These customers do not just buy features. They buy reliability, accountability, and continuity.
So when these companies adopt AI capabilities, the likely outcome is not always that the incumbent gets replaced. Often the incumbent is the one best positioned to deliver those capabilities because it already has the relationship, the permissions, and the trust.
That matters a lot.
What these companies are actually solving
This becomes much clearer when you ask the right question.
Not: Can this product be rebuilt?
But: What job is this company actually doing for the customer?
That is the better lens.
Salesforce is not just a contact database. In many enterprises it is tied into quoting, approvals, pipeline management, territory rules, contracting, and revenue workflows. The value is not just storing customer data. The value is providing reliable commercial infrastructure.
ServiceNow NOW 0.00%↑ is not just an IT ticketing tool. In many companies it connects IT, HR, legal, security, and internal controls. It creates process discipline and auditable records. That is much harder to replace than people assume.
Adobe ADBE 0.00%↑ is not just image creation software. For large creative organisations, it provides workflow consistency, asset control, collaboration, legal comfort, and a talent base trained on its ecosystem. That is far more valuable than just the tool itself.
In each case, the visible software is only part of the value. The rest comes from being trusted, embedded, and integrated into the way the customer operates.
Ironically, AI may help the incumbents most
If coding becomes faster and cheaper, the businesses with scale may benefit the most. They already have customers, distribution, trust, and installed workflows. Now they can potentially improve products faster and do it with lower cost.
That can increase operating leverage.
Historically, one of the biggest expenses in SaaS has been engineering talent. Alongside that came heavy stock-based compensation to attract and retain it. If AI meaningfully improves developer productivity, large software companies may be able to ship more, maintain products more efficiently, and eventually rely on fewer people for the same output.
That could mean:
faster product cycles
lower engineering intensity
better margins
less dilution from SBC over time
stronger free cash flow conversion
So yes, AI may compress parts of the market. But it may also make the strongest platforms more profitable.
The real mistake is that the market is mixing up two very different categories.
One is software that is easy to copy, lightly embedded, and vulnerable to falling barriers to entry.
The other is software that sits deep inside enterprise workflows, with trust, integration, compliance, and switching costs built over many years.
The first category deserves more caution.
The second may actually become stronger.
Bottom line
The SaaS sell-off reflects two real fears: that AI may weaken the old seat-based revenue model, and that software is becoming easier and cheaper to build.
Both concerns are valid.
But the market is going too far by treating all SaaS as if those risks hit everyone equally.
If a software business is shallow and easy to replicate, then yes, the threat is real.
But for deeply embedded enterprise platforms, the story is different. Their value was never just in the code. It was in the trust, the workflow integration, the distribution, the compliance layer, and the fact that they are already part of how large organisations operate.
In that world, AI is not just a threat. It can also be a productivity tool that helps the incumbents build faster, run leaner, and become even more profitable.
Disclaimer: This article is for informational purposes only. It is not financial advice. I am not a financial advisor. I may buy or sell these stocks at any time. You must do your own research before investing
By reading this email, you agree to the full disclaimer found here: https://www.thevaluethesis.com/p/disclamier