To The Moon

Anthropic’s Claude models have created a massive sell-off is Sofware-as-a-Service businesses, as investors run away from disruption.

My colleague Hazel prepared the above overview of the worst-performing stocks YTD. As you can see, it is dominated by SaaS names.

• $Intuit(INTU)$ is down 33%

• $Gartner(IT)$ is down 33%

• $ServiceNow(NOW)$ is down 28%

• $Workday(WDAY)$ is down 26%

This fear is rational, as historically, SaaS monetised its services through seat-based pricing.

This was a very straightforward way to monetise software. The more features were added, the more usefull software became. Each incremental feature served as a selling point to customers, convincing management to increase the number of seats they were paying for.

This business model is getting disrupted by AI!

Just recently, Claude released new AI demos that demonstrated how their models automated Excel grunt work and legal document analysis. These new tools promise to make employees significantly more productive. But if a single employee can do the work of five, then, at least in theory, that implies four fewer SaaS seats will be needed for a company to generate the same output.

The math seems simple.

More AI = Fewer Seats = Less Revenues

That’s the narrative that is causing the havoc in the market now, but the reality is more complicated than that, as software isn’t dead just yet.

It’s just that the unit of value is shifting from seats to outcomes!

Instead of charging customers based on how many employees are using the software, SaaS will charge based on the value of the outcome created. This means that companies that provide low-value, commoditized software will see their growth disappear.

However, that also means companies whose software actually makes their customers money will see their business accelerate!

1. Who gets hurt first?

The reality is that AI won’t replace apps, AI replaces app usage.

Agents become the interface layer for routine work, so anything whose value is mostly humans clicking on buttons is in the blast radius.

A graph of a bar graph AI-generated content may be incorrect.

Researchers at Bain and Company believe that 50-60% of employee time in customer support can be automated by AI. R&D and engineering 40-50%. HR, marketing, and finance 35-45%.

There are a lot of SaaS providers that are heavily exposed to these verticals. As AI automates the roles of these employees, companies will cancel SaaS seat based subcriptions.

Here is who gets hurt first:

• Point solutions with low switching costs

• Human middleware SaaS

• Workflow layers on top of systems of record

• Analytics and dashboards

• Land and expand SaaS

In software, a point solution is a tool that solves one narrow problem really well and almost nothing else. It’s built to optimize a single use case, such as meeting notes, schedulers, data visualization, expense tracking, and so on. These tend to have very low switching costs, so if you find a new and better tool, you can switch relatively easily. These point solutions can be easily built with AI either internally, or by other SaaS that can add it as a feature of an already existing product.

Next, human middleware SaaS refers to tools that exist only because humans operate the process. It’s software whose core function is to coordinate people across systems. People read the request, figured out what it was, copied context into the right place, nudged the right person, and followed up until it closed. Think ticketing, incident, approval, and request solutions. In the world of AI agents, there will be fewer seats that subscribe to these tools.

Think of procurement platform Coupa and signature SaaS DocuSign. During my finance internship, I uploaded invoices in Coupa so the customer would pay us. AI Agents will automate invoice processing to a degree that you won’t need a dedicated platform such as this. Whilst some executives sign hundreds of contracts a week in DocuSign, this too can be automated with AI.

Systems of record save all the data, and there are various SaaS that use that data to provide workflows on top of it. These SaaS companies specialise in making interesting features and pretty buttons, so users use more of the software.

Essentially, their core monetization engine is humans spending time in their software. In a world of AI agents, there is little use for these workflow layers, as agents don’t need pretty buttons. In fact, AI doesn’t need any buttons at all, they can access all the data through code and APIs with the system of record.

Think of ServiceNow, which is one of the biggest workflow layer platforms in the world.

The stock is down 41% in the last 6 months, as investors reassess whether their platform, which depends on systems of record to provide workflow and process orchestration services, can survive the AI.

Simply put, ServiceNow monetises process friction, but AI removes friction!

Less friction means fewer system users, which in the ServiceNow per-seat business model directly translates to less revenue.

Another SaaS that is set to suffer is analytics and dashboards. At my previous job, every Monday, I created a cash flow dashboard and emailed it to various executives. We had a whole team based in Morocco whose job was to assist various departments with creating dashboards. The objective of these dashboards is to visualise KPI’s to help people make better decisions. If the cash position was not up to the CFO’s liking, he would direct people to take action.

Tableau Dashboard Examples

In the above picture, you see an example of a Tableau dashboard for medical clinics. Salesforce paid $15.7B to acquire Tableau in 2019, despite it reportedly only making $1.3B in revenues.

Tableau charges customers not only per dashboard, but also per viewer. As AI models learn to build great dashboards in seconds, there won’t be a need to pay for such an expensive solution. Furthermore, if AI reduces the number of middle managers as some predict, the demand for dashboards will decrease.

In the future, instead of having a team dedicated to gathering and structuring data to feed Tableau, executives will simply ask a chatbot, “How much cash did we collect last week?”

“$12M, sir. It was below our target of $18M. The majority of the $6M discrepancy comes from Customer X not paying invoice 5152525 for $4.3M. The Overdue Invoice Reminder Agent sent the customer a reminder on Thursday. Please see below a graph of weekly cash collection trends and deviations from the forecast.

Would you like me to analyse Customer X payment behaviour to identify if this is a recurring issue?“

Lastly, a key customer acquisition strategy is about to go completely out of the window, the land and expand model.

The pitch of many SaaS goes like this: “Take 50 licenses of our software, try it out. We are confident that we can deliver value to your organization.”

The company gives these licenses out for cheap, with the hope that the user will love them and get more and more seats, leading to higher revenues. This model doesn’t work in the Agentic AI world, where the revenues go to the SaaS that can deliver superior outcomes.

2. Who thrives in the AI world?

So far, everything sounds bleak for Saas companies. While some of them will adapt and figure out new monetisation streams, most will not.

However, as I said earlier, SaaS that use AI to actually make their customers money will see their business accelerate.

Here is a 6-point checklist to score any SaaS to assess whether a company will be an AI Winner or Losser

1. System of Record

2. Switching Costs

3. Regulatory or Audit Requirements

4. Data Advantage

5. Ecosystem Moats

6. AI Costs and Monetization

A system of record is the place where data is created, stored, audited, and used. This is not just a place where data is viewed or analyzed. So a dashboard SaaS is not a system of record.

For example, SAP is possibly the world’s biggest system of record. The German SaaS company provides software that is widely used in finance, manufacturing, supply chain, HR, and more. This is an important criterion because a system of record is where all the data is stored, which any functioning AI desperately needs. As models get more advanced and cheaper to use and implement, it will be relatively easy for systems of record to build AI products on top of these systems.

Simply put, AI can turn a System of Record into a System of Action.

Next is switching costs.

If it’s quite difficult, expensive, and time-consuming to switch providers, then the SaaS is at a smaller risk of AI disruption.

Returning to the SAP example, at my previous job, we used it for finance and accounting. When I started working there, my CFO said that there is a plan to move to a newer version of SAP. After plenty of meetings with the SAP sales team and various consultants, the project was delayed due to budget limits. When I left that job after almost four years, we were still using the same SAP. And this is when moving to a newer version from the same provider, imagine the cost and complexity of moving to a different provider.

Furthermore, investors must consider whether there are regulatory and audit requirements that strengthen the moat.

In regulated environments, it’s not enough for the AI Agent to do the task. It must do the task with defensible controls, traceability, and retention, every time, so they can pass scrutiny under audit.

That’s why compliance-heavy SaaS tends to be safer, as the product isn’t the features, it’s trusted rails.

In customer service, coding, dashboards, or in IT maitanence there are fewer, if any, audit requirements. Meanwhile, in payroll, human resources, cybersecurity, pharmaceuticals, and healthcare, there are strict regulations and audit rules. Here, AI deployment will be slower and more expensive, protecting incumbents.

Another important point is data.

A diagram of a cloud AI-generated content may be incorrect.

SaaS companies that have access to a lot of data will be able to turn that data into more advanced AI. The more data AI has, the more useful it is. This creates a self-reinforcing cycle, as companies with the best AI get more customers, which generate more data that feeds into the AI, making it even better.

And here I just don’t mean data from systems of record. I mean outcomes, results, steps to get there, what worked, and what didn’t. Zeta, for instance, is a perfect example. They analyse data from millions of advertising and marketing campaigns in real time to deliver better results. Each campaign parameter is saved to be analysed by AI.

The last, and possibly most important point. Can the company turn AI costs into AI revenues?

AI development and deployment is very expensive! So there is a scenario in which even in the case when a SaaS survives, it will have structurally lower margins than before.

The main reason why SaaS companies have historically traded for high multiples was that they had very high 75-85% gross margins, making their business extremely scalable. From each incremental $1 of revenue, these companies made 80 cents in gross profit. It doesn’t take that many more operating expenses to generate that additional $1, so the theory was that from those incremental 80 cents in gross profit, SaaS companies could make 40-50 cents in operating profit.

If this process kept going, in a decade, SaaS companies could have incredible margins. But AI is putting this theory in doubt.

If SaaS companies are not systems of record, don’t have their own data, don’t have their own AI models, then in an Agentic AI world, they must share more of the economics.

This resets terminal value expectations. In a DCF valuation model, the terminal value is a significant contributor to the present value of all cash flows. If that terminal value is 30-40% lower than previously expected, the present value falls, leading to a fall in share price, which we are seeing right now.

Let’s talk about Salesforce, which recently released the “Agentic Enterprise License Agreement” (AELA). It allows customers to convert unused human seat licenses into AI credits.

This is Salesforce’s attempt to stop churn from customers who might use AI to reduce the number of seats they are using. The problem is that 1 seat has significantly better economics, as it costs nothing to Salesforce. Whilst AI credits incur significant operating expenses, as AI compute is notoriously expensive.

So Salesforce is losing money on AELAs in the short term, to hope a long-term customer lock-in. Analysts are skeptical that Salesforce will be able shift AI costs to consumers through higher prices or new subscriptions.

3. A Few Winners

I already discussed a few companies that the market views as losers, Salesforce, ServiceNow, and DocuSign. So let’s talk about a few winners.

• $Zeta Global Holdings Corp.(ZETA)$

• $SAP SE(SAP)$

• $Palantir Technologies Inc.(PLTR)$

Let’s assess them on the 6 AI SaaS winner criteria checklist.

Zeta Global – Ad Choices

Zeta is an AI marketing and advertising SaaS that helps large enterprises acquire, retain, and upsell customers.

1. System of record, yes.

Its marketing platform acts as a centralized system that unifies customer identity, data, audience profiles, and activation across marketing channels. It is effectively a system of record for customer data and campaign execution.

2. Switching costs, yes.

To move all data from one provider to another is complex and risks disrupting ongoing campaigns. If things go wrong, lost revenue could be in the hundreds of millions of dollars.

3. Regulatory and audit requirements, yes.

While regulations are not as strict as in healthcare or cybersecurity, Zeta still must comply with data privacy regulations.

4. Data advantage, hell yes.

This is arguably Zeta’s strongest moat. The company leverages a massive proprietary data cloud that includes billions of consumer identifiers and predictive signals. This data fuels its AI models and enables deeper personalization and predictive analytics compared to many competitors.

5. Ecosystem moat, no, but emerging.

The company has a large platform that is being expanded through acquisitions and the development of new offerings. However, their ecosystem of partners and integrations is not as vast as giants such as Salesforce, Microsoft, or AWS.

6. AI costs and monetization, 100% yes.

AI is central to Zeta’s narrative and growth. Rather than bolt-ons Its AI features are deeply integrated and drive measurable client outcomes, which translates into monetization ability and pricing power. The company can easily monetise AI through new offerings and shift AI costs to customers through higher prices.

From this checklist, I would give Zeta a score of 10 out of 12.

1. System of record, yes.

SAP’s core value is being the system of record for the most critical enterprise functions.

2. Switching costs, yes.

Moving EPR systems could cost millions and take 1-2 years.

3. Regulatory and audit requirements, yes.

SAP operates in functions with intense regulatory scrutiny, such as financial reporting, tax compliance, supply chain traceability, data privacy, and audit trails.

4. Data advantage, yes.

SAP arguably has one of the richest and stickiest enterprise data fabrics globally. Embedding AI on top of it has huge potential for revenue generation.

5. Ecosystem moat, yes.

SAP has a deep global ecosystem of developers, consultants, and partners that has been built over decades.

6. AI costs and monetization, yes.

The company can easily shift the AI compute cost to customers, and it can monetise new AI products.

From this checklist, I would give SAP a score of 12 out of 12.

1. System of record, yes.

Palantir’s platforms of Foundry and Gotham serve as systems of record for its enterprise and government clients

2. Switching costs, yes.

Incedibly high.

3. Regulatory and audit requirements, yes.

In addition to healthcare, Palantir literally works with the CIA and the military. The company operates in the strictest regulatory environments in the world.

4. Data advantage, yes.

While Palantir doesn’t own any data, its customers do, they transform raw data into contextualized, interconnected assets that power AI and analytics.

5. Ecosystem moat, yes.

Strategic partnerships with hyperscalers, Nvidia, Microsoft, and a network of consultants.

6. AI costs and monetization, yes.

From this checklist, I would give Palantir a score of 12 out of 12.

Thanks for reading Global Equity Briefing! This post is public so feel free to share it.

4. Final Thoughts

The easy days of SaaS are over.

Gone are the days when SaaS would trade for 30x ARR and grow 40-60% per year. Investors are scared and will not be loyal to SaaS with decelerating growth and high multiples.

Most SaaS won’t survive, but those that do will be huge winners. This means that consolidation is coming. Companies such as DocuSign that are more of a feature than a standalone company won’t remain relevant.

If SaaS revenues are not as durable as investors expected, then terminal values are much lower, warranting a valuation reset. However, today we are seeing indiscriminate selling in all SaaS names. This is an opportunity to purchase companies that are poised to become AI winners, before the market realizes which one is which.

For SG users only, Welcome to open a CBA today and enjoy access to a trading limit of up to SGD 20,000 with unlimited trading on SG, HK, and US stocks, as well as ETFs.

🎉Cash Boost Account Now Supports 35,000+ Stocks & ETFs – Greater Flexibility Now

Find out more here.

Complete your first Cash Boost Account trade with a trade amount of ≥ SGD1000* to get SGD 688 stock vouchers*! The trade can be executed using any payment type available under the Cash Boost Account: Cash, CPF, SRS, or CDP.

Click to access the activity

Other helpful links:

• 💰Join the TB Contra Telegram Group to Get $10 Trading Vouchers Now🎉

• How to open a CBA. How to link your CDP account. Other FAQs on CBA. Cash Boost Account Website.