In my previous post, I introduced a simple but useful lens for thinking about SaaS companies in an era of agentic AI: not all moats are built the same way. Some companies derive their staying power from the data they hold. Others from the complexity of the workflows they govern. And others — more than most would like to admit — from the fact that humans have simply gotten used to their interface.

That last category is the one worth watching most closely. Because if there's one thing agentic AI makes abundantly clear, it's that software built primarily for human navigation is going to face some uncomfortable questions. Agents don't browse menus. They don't need dashboards. They query data, execute logic, and move on. What they need underneath them is infrastructure that is reliable, governed, and deep.

So with that framing in mind, I took a look at ten of the largest SaaS companies in the world — not to render verdicts, but to share some observations about where the moats seem solid and where they feel a little thinner than the market has historically priced in. I'm not an analyst, and I don't cover these companies. But I do spend a lot of time thinking about where enterprise software is headed — and these feel like the conversations worth having right now.

Microsoft

It would be easy to start here and simply say "Microsoft wins" — and there's a version of that argument that is genuinely hard to argue with. Azure is the infrastructure layer that most enterprise AI runs on. Microsoft 365 holds the email, documents, and calendar data that agents will need context from. GitHub Copilot has a commanding early position in the developer workflow. And Copilot Studio is becoming the environment where enterprise builders construct multi-agent systems at scale.

What makes Microsoft's position particularly interesting is that it is simultaneously the platform and the application — the cloud that competitors run on, and a competitor to those same companies in productivity, CRM, and now HR and finance through Dynamics 365. That dual role creates both enormous leverage and real complexity for customers trying to manage their vendor relationships thoughtfully.

Microsoft's commercial cloud surpassed $50 billion in quarterly revenue for the first time in early 2026, and Microsoft 365 Copilot seat adds accelerated significantly through the year. The early traction is real. The more interesting question over the next few years will be whether enterprises lean into Microsoft's full-stack ambition or deliberately build around it to preserve optionality.

On the moat: Microsoft's moat runs deepest at the infrastructure and data layers — and those are exactly the layers that agentic AI depends on most. The application layer is in transition, but the substrate underneath it looks durable.

Salesforce

Salesforce's response to the agentic AI moment has been one of the most aggressive commercial pivots in enterprise software in recent memory. Agentforce — its AI agent platform — launched in late 2024 and has moved faster than most expected, reaching over 9,500 paid deals and crossing $500 million in annualized revenue within its first year. The company has also built a significant data layer through Data Cloud and the Informatica acquisition, which gives it a more complete picture of enterprise customer data than it has ever had before.

The underlying logic makes sense: if CRM processes will increasingly be executed by agents rather than users, Salesforce wants to be both the data source those agents query and the orchestration layer through which they act. The challenge — and it's one Salesforce has been transparent about — is that moving from per-seat subscription pricing to consumption-based agent pricing is a non-trivial transition. It creates near-term uncertainty even when the long-term direction is clearly right.

On the moat: The CRM data asset is genuinely difficult to replicate — decades of customer relationship history is the kind of context that agents need to be useful. The pricing model transition is the variable worth watching closely.

SAP

SAP often gets characterized as a legacy ERP vendor navigating a difficult cloud transition — and there's some truth to that. But the agentic AI conversation actually surfaces something interesting about SAP's position that doesn't get enough attention: the company processes more enterprise business transactions than almost anyone else on the planet, and that data — financial records, procurement flows, supply chain positions, workforce data — is exactly what AI agents need to function meaningfully in regulated enterprise environments.

SAP's cloud business grew 26% in constant currency in FY2025, and Business AI was included in two-thirds of its cloud order entries by year-end — which suggests genuine customer adoption rather than just marketing attachment. Its Joule platform, which allows customers to build custom AI agents trained on their own SAP data, is a credible attempt to make SAP the governance layer for enterprise agents rather than just a data source for them. The company's "apps, data, AI" framing — the idea that no LLM provider has access to the depth of business context that SAP's installed base generates — is worth taking seriously.

On the moat: SAP's workflow complexity moat — built around global compliance, multi-jurisdictional operations, and decades of embedded process logic — is one of the more durable ones in this group. The legacy architecture is a real challenge, but the underlying data depth is genuinely hard to compete with.

Oracle

Oracle's story in 2025 was less about its application portfolio and more about a single, large bet: that Oracle Cloud Infrastructure would become the preferred compute substrate for hyperscale AI workloads. OCI infrastructure revenue grew over 50% in FY2025, GPU-related revenue grew substantially faster, and the company's contracted revenue pipeline reached extraordinary levels — driven by multi-year AI infrastructure commitments that extend well beyond its traditional ERP customer base.

The application side of Oracle's business — Fusion ERP, NetSuite, its industry cloud portfolio — is a separate conversation. These platforms carry strong switching costs and deep process complexity, but they face the same questions about agentic readiness that any large enterprise application suite faces. Oracle's AI Agent Studio is a step toward addressing that, but the pace of agentic feature adoption in ERP historically lags the pace of announcement.

On the moat: Oracle's most defensible position may actually be its database heritage and OCI infrastructure rather than its application layer — agents need reliable, auditable data stores, and Oracle's database reputation is hard-earned over decades. The infrastructure bet is large and still being validated.

Adobe

Adobe's situation is one of the more nuanced in this group. On the surface, a company whose business is built on creative software looks vulnerable to AI — and the market has reflected that concern, with significant stock pressure through 2025 driven by fears that generative AI tools will commoditize the lower end of creative work. Those concerns are not entirely wrong.

But Adobe has a structural asset that gets less attention than it deserves: its Firefly generative AI models are trained on licensed content with clear commercial provenance, which means enterprises using Firefly for marketing production have legal IP protection that generic AI models cannot offer. In regulated industries and large marketing organizations, that matters more than most people outside those environments appreciate. Adobe's GenStudio platform crossed $1 billion in ARR in FY2025, growing over 25% year-over-year, which suggests real enterprise adoption of the content supply chain vision.

On the moat: The commercially safe IP moat is more defensible than Adobe's recent stock performance suggests. The risk is at the commodity tier of creative work — individual productivity tools face real competition. The enterprise content governance layer feels more durable.

Intuit

Intuit is one of the companies in this group that I find genuinely compelling to watch in the agentic era — not because it's the most talked about, but because its structural position is quietly very strong. The combination of tax data, small business financial data, and consumer credit data that Intuit holds across TurboTax, QuickBooks, and Credit Karma is an irreplaceable asset for any AI agent trying to manage personal or business finances intelligently.

What's particularly interesting is Intuit's explicit "AI plus Human Intelligence" model — the idea that AI handles the routine work (roughly 80% of data categorization and entry) while a network of 13,000 human tax and financial experts handles the edge cases and provides the accountability layer that regulated financial domains require. Over 3 million customers engaged with Intuit's autonomous AI agents in FY2025, with repeat engagement rates above 85%. TurboTax Live revenue grew 47% for the year. The hybrid model appears to be working, and it's a template that other companies in regulated domains are watching closely.

On the moat: The financial data moat runs deep, and the regulatory complexity of tax and compliance creates a natural barrier that general-purpose AI cannot easily cross. The AI plus human hybrid feels like the right architecture for this domain — not a transitional concession, but a durable structural choice.

ServiceNow

ServiceNow occupies a genuinely interesting strategic position right now. It began as an IT service management platform — the system enterprises use to manage service requests, incidents, and change management — and has expanded into HR, customer service, legal, and risk operations. What makes ServiceNow's position relevant to the agentic AI conversation is its Configuration Management Database: the authoritative map of enterprise IT assets, their relationships, and their histories. For any AI agent to manage IT operations autonomously, it needs to know what exists in the environment. ServiceNow spent two decades building that knowledge base.

In February 2026, ServiceNow announced its Autonomous Workforce initiative — AI agents designed to handle IT service requests end-to-end, with early customer deployments reporting significant reductions in manual resolution time. The architecture is worth noting: ServiceNow describes its approach as combining probabilistic AI reasoning with deterministic workflow execution — meaning agents can think flexibly but act within governance guardrails. That's a meaningful distinction for enterprises that are understandably cautious about letting AI make irreversible decisions without human oversight.

On the moat: The CMDB and governance workflow layer is a genuine moat — it reflects two decades of enterprise IT complexity that cannot be shortcut. The question is whether ServiceNow can successfully extend that governance model to become the orchestration layer for the broader agent workforce, not just IT operations.

Shopify

Shopify's relationship with agentic AI is structurally different from every other company in this group — and it may be the most strategically elegant response we've seen. The fundamental question for Shopify was this: as AI agents become buyers — shopping conversationally through ChatGPT, Google Gemini, or Microsoft Copilot rather than browsing websites — does commerce route through or around Shopify's infrastructure? Shopify's answer has been to make routing around them essentially impractical.

The Universal Commerce Protocol, co-developed with Google and now supported by Wayfair, Walmart, Target, Visa, Stripe, and PayPal, establishes Shopify as the standard for how AI agents connect with merchant storefronts and complete purchases. AI-driven traffic to Shopify stores was up 7x in 2025, and orders attributed to AI searches grew significantly through the year. The company has been direct about the commercial logic: when a transaction starts in an AI conversation and completes through Shopify's checkout, the economics for merchants are the same as any other channel. Shopify captures its take regardless of where discovery happened.

On the moat: Shopify is positioning itself as infrastructure rather than application — the payment rails and commerce logic layer that AI agents must use to complete transactions. That's a more durable position than being the interface through which humans shop, and it's a thoughtful response to a genuine structural shift.

Workday

Workday's situation is one of the more watched stories in enterprise software right now. The company manages HR and financial operations for over 65% of the Fortune 500, processes over a trillion transactions annually, and holds workforce and financial data that is genuinely difficult to replicate. And yet its stock declined roughly 50% between mid-2024 and early 2026, as investors began repricing the risk that agentic AI reduces the number of human users who need to actively operate the platform — which directly challenges a per-seat revenue model.

The company's response — Workday Illuminate agents, the Agent System of Record for managing enterprise AI agent fleets, and a new consumption-based Flex Credits pricing model — is directionally coherent. Co-founder Aneel Bhusri returned as CEO in February 2026, and the framing on the earnings call was notably candid: the transition from seat-based to agent-based economics is real, vendors who benefit from Workday's data without compensating for that access will face new friction, and Workday intends to position itself as the governance layer for the emerging agent workforce the same way it became the system of record for the human workforce.

On the moat: The HR and financial data moat is among the strongest in this group — the depth of workforce context Workday holds is a prerequisite for any agent trying to make meaningful people or money decisions. The transition from seat-based to consumption-based pricing is the execution challenge that will define the next chapter.

Snowflake

Snowflake is the company in this group whose business model is most directly aligned with the agentic AI transition — not just resilient to it, but structurally accelerated by it. Its consumption-based data cloud means that more AI agent activity equals more data queries equals more Snowflake revenue. The platform does not depend on human users logging in to generate revenue. In an era where agents are becoming the primary interface for enterprise data, that is a meaningful structural advantage.

Snowflake's AI workloads are growing fast — AI-related activity influenced roughly 50% of new bookings through FY2026, and the company reached a $100 million AI revenue run rate ahead of its own internal projections. Partnerships with both OpenAI and Anthropic — each involving $200 million, multi-year commitments — ensure that leading AI models are natively available within Snowflake's governed data environment, which addresses one of the key concerns enterprises have about running AI on sensitive data. The company signed the largest deal in its history during this period, exceeding $400 million.

On the moat: As a data infrastructure layer rather than an application, Snowflake sits in an interesting place: agents need reliable, governed data to function, and that's precisely what Snowflake provides. The competitive pressure from Databricks and hyperscaler-native data services is real, but the consumption model's alignment with agentic AI growth is a structural tailwind that most of the other companies in this group would welcome.

Key Takeaways

Looking across these ten companies, a few things stand out which are worth mentioning.

The companies that seem most comfortable with the agentic transition are the ones that were never primarily selling an interface in the first place. Snowflake sells data infrastructure. Intuit sells financial expertise backed by irreplaceable data. ServiceNow sells workflow governance. Shopify sells commerce infrastructure. These companies have something underneath the UI that agents actually need — and that changes the conversation entirely.

The companies navigating the most uncertainty are the ones whose primary value proposition has historically been the richness and usability of their application experience. That's not a death sentence — some of the most sophisticated AI strategies in this group belong to companies in exactly that position. But it does mean the next few years require genuine reinvention, not just incremental AI feature additions.

What's also striking is how many of these companies are converging on similar strategic moves: building agent platforms, shifting toward consumption-based pricing, investing in data governance layers, and positioning themselves as orchestration infrastructure rather than standalone applications. The architectures are starting to rhyme, which means differentiation will increasingly come from depth of data, richness of workflow logic, and the trust that regulated enterprises place in specific vendors for specific domains.

Next, I'll zoom out from individual companies to ask the bigger structural question: when the dust settles on this transition, which categories of players actually capture the most value? The hyperscalers, the LLM providers, the vertical AI native challengers, and the data aggregators all have legitimate claims. How will those claims resolve, and what does all this mean for the enterprise software market as we've known it?

‍