AI won’t eat software—it will reshape systems of engagement

There’s a growing narrative that AI will “eat” enterprise software. The idea is that agents will ultimately automate and demolish SaaS application and security companies.

It’s a compelling headline.

It’s also wrong.

Agents will create value across the software stack. AI and agents are a tsunami of innovation coming fast to the technology industry. If you don’t surf the wave, it could demolish you. Yet we’re finding that many of the best SaaS and security companies –Salesforce, ServiceNow, Workday, SAP, Oracle, Palo Alto Networks, CrowdStrike, and even Cohesity – are taking advantage of these new GenAI and agentic solutions to reshape their futures fundamentally. In this agentic shift, the software companies that own either systems of record (in the case of applications) or systems of protection (in the case of security), along with the data gravity that comes with their large customer base, critical business logic, and workflows, are well-positioned to ride this incredible AI/agentic wave to new heights.

We're already seeing many of these software companies embracing agentic solutions, in many cases in close concert with NVIDIA, OpenAI, or Anthropic. One good example is ServiceNow’s collaboration with Anthropic to help customers build AI-powered applications integrated with ServiceNow. Another example is our own collaboration at Cohesity with NVIDIA to build a leading RAG-based application for secondary backup data on top of NVIDIA NIM and Cohesity’s work with NVIDIA on NEMOClaw.

The core of applications – systems of record for enterprise assets

It’s first important to understand, in very basic terms, what enterprise applications are typically made up of.

Every enterprise application at its core consists of:

• A CRUD database (Create, Read, Update, Delete) as a system of record
• Sophisticated application logic
• Workflows and analytics

For example, SAP is a system of record for financial assets, Workday for people assets, and Salesforce for all customer assets. The crown jewel of Salesforce is the customer master, which organizes a schema of all interactions around a customer, and you build your customer knowledge graph on top of it.

AI agents don’t replace these systems—they depend on them. You can’t build the system of record, the application logic associated with it, and all the data gravity that comes with it, overnight with vibe coding. However, each of these companies, if they’re smart, is leveraging agents to accelerate innovation and transform workflows, alerts, and analytics on top of their apps. They are turbocharging what I describe as the new systems of engagement for an AI/agentic world.

We use ServiceNow, Salesforce, Workday, and Oracle NetSuite among our key SaaS applications internally at Cohesity. They are strategic partners of Cohesity, and we fully expect to leverage their innovations in the agentic world, driving the linkage between systems of record and systems of engagement. For example, ServiceNow is a configuration management database (CMDB) that maintains a well-organized schema of a company’s IT assets, enabling elegant IT workflow operations on top. They’re now a system of action that ties the system of record to the systems of engagement in the work they are doing at the AI/agentic layer.

The essence of security software – systems of protection for IT assets

Like the applications world, the security industry includes 5,000+ players. The biggest players, like Palo Alto Networks and CrowdStrike, are platforms within platforms. But they all started with a mission to be a “system of protection” for a key IT asset: 

• Palo Alto Networks for networks and firewalls
• CrowdStrike for endpoints
• Zscaler for secure access service edge
• Okta for identity
• Google (with Wiz and other assets) for cloud security
• Cohesity for data security

All these companies have built sophisticated capabilities in their domains that agents can’t replace. Palo Alto’s deep knowledge of firewalls, CrowdStrike’s expertise in endpoints, or Cohesity’s expertise in data security and recovery can’t be vibe coded easily or quickly by an agent. Furthermore, each of these companies has extensive telemetry in its systems, creating data gravity that makes its systems smarter and much harder to recreate.  

I recently asked Anthropic’s Claude Code to build a working prototype of Cohesity. Our platform represents millions of lines of code, years of engineering, and a deep understanding of enterprise data, immutable file systems, sophisticated AI-based security algorithms, and more. What it came back with wasn’t code in any meaningful sense, not something you could take and deploy, or even really call a product. It didn’t give me Cohesity. But it gave me a map of how you might start thinking about building something like our cyber resilience platform.

That said, we can’t rest on our laurels. If we’re smart, just like our peers in application software, we’ll leverage agents in our innovation. This will turbocharge the new systems of engagement for an AI and agentic world.

The real shift – agents turbocharge new systems of engagement/action

This leads us to where the action is going to be – the new systems of engagement and action in an agentic world, where workflows, alerts, and analytics help us make better, faster decisions with AI. The focus is on outcomes, not transactional events.

Every application and security software company includes a control-plane layer of workflows, alerts, and analytics that agents will completely transform. That’s where the new value accrues and must be fundamentally rethought.