The short answer: your core enterprise applications are not going away anytime soon, but how you use and pay for them will change.

Industry analysts consistently point to a long transition, not a sudden collapse:

• Core systems remain: Forrester notes that core applications will stay in place for the foreseeable future, with only “erosion around the edges.” Executive teams are not making decisions to rip out CRM or ERP wholesale.
• Timeline is measured in decades: It could take decades for AI agents to fully take over many enterprise workloads. In the meantime, humans will stay in the loop and autonomy will be introduced gradually.
• Incumbents are adapting: Major vendors (Microsoft, Oracle, SAP, Salesforce, ServiceNow, etc.) are embedding agents into their platforms and building unified data layers, which positions them to pivot into the AI era rather than be displaced by it.

What will change is how you extract value from your existing stack:

• You’ll be highly incentivized to add agentic AI to current platforms to get more value from multimillion-dollar investments.
• Some simple point products (e.g., basic workflow or lightweight project tools) are likely to disappear faster because they’re easy to replicate with agents.
• Highly specialized, vertical applications (e.g., EHR systems like Epic and Cerner, or tools like Procore in construction) are better insulated because they embed deep domain expertise and integrations.

In practice, you should plan for a staged evolution: keep your core systems, layer agentic capabilities on top, and selectively retire or consolidate simpler tools that agents can easily replace.

Agentic AI is set to rethink how software is priced, shifting focus from access to actual work done.

Several trends are emerging:

• From seats to outcomes: Bain & Co. notes that when an agent replaces a human task, customers will expect to pay based on outcomes, not log-ons. The core idea: stop charging for access and start charging for completed work.
• Seat-based pricing becomes obsolete: IDC predicts that by 2028, pure seat-based pricing will be obsolete, and 70% of software vendors will have refactored pricing around new value metrics such as consumption, outcomes, or organizational capability.
• Early movers: Vendors like Intercom and Salesforce are already experimenting with these models.

For CIOs, this opens up new levers to manage cost and value:

• Hybrid licensing strategies: You might keep a base of traditional seats (e.g., 80–90 out of 100) and convert the rest to consumption- or outcome-based pricing, especially for agent-driven use cases.
• Flexible tiers and “flex” options: Expect vendors to introduce more granular tiers that bundle agentic capabilities, consumption pools, or outcome-based elements.
• Third-party optimization: Consultants and service providers may offer to manage your agentic AI implementations and charge a percentage of the outcomes they help you achieve.

Importantly, AI agents themselves can help you optimize spend:

• Agents can analyze consumption and usage patterns across your application portfolio.
• You can use those insights to renegotiate contracts, right-size licenses, and reduce underutilized capacity.

Actionably, you should start modeling scenarios where a portion of your portfolio moves from seat-based to consumption/outcome-based pricing, and use AI-driven analytics to support those negotiations.

Agentic AI is poised to reimagine the enterprise stack in three key ways: data unification, new development patterns, and an agentic interface layer.

1. Blurring traditional software categories

• AI agents don’t care whether data lives in CRM, ERP, ITSM, or elsewhere—they just need access.
• Vendors are responding by merging platforms and unifying data:
  • Oracle: integrated cloud ERP + CRM plus a managed agentic platform.
  • Microsoft: ERP and CRM under Dynamics 365, plus industry-specific agentic offerings using smaller, cost-effective language models (SLMs).
  • SAP: combining Signavio (process management), LeanIX (enterprise architecture), and its Joule AI agent into a cohesive system.
  • Salesforce: merging Mulesoft (integration/automation), Data360 (customer data), and Agentforce (AI agents).
  • ServiceNow: acquiring agentic AI platform Moveworks and pushing into CRM territory.

2. Vibe coding and end-user-built agents

• Vibe coding uses AI agents to build software from natural language prompts, extending low-code/no-code.
• End users can ask tools like ChatGPT, Gemini, Claude, Cursor, or GitHub Copilot to create apps that sit outside traditional CRM/ERP boundaries.
• This is particularly disruptive for point solutions (small, narrow apps). It’s less likely to replace systems that manage entire customer or supply chain databases in the near term.
• Organizations with limited technical maturity may struggle to safely deploy mission-critical agents, so adoption will vary by capability and risk appetite.

3. Agentic orchestration as the new interface

• Analysts expect the primary user interface to become agentic and conversational, not a traditional SaaS dashboard.
• IDC highlights complexity as the Achilles heel of SaaS: each app has its own UI and learning curve, often used sporadically and inefficiently.
• In the emerging model:
  • Users interact with agent-driven interfaces that perform tasks across multiple systems.
  • An orchestration agent routes work to the right engine—LLM, SLM, or RPA—based on efficiency, cost, and even energy usage.
  • Traditional CRM/ERP/ITSM systems still exist, but are increasingly hidden behind this orchestration layer.

For CIOs, this means planning for a stack where:

• Data platforms are unified and accessible to agents.
• Category boundaries (CRM vs. ERP vs. ITSM) matter less to end users than the workflows agents can execute.
• You make a strategic choice: rely on incumbent vendors for orchestration, or bring in disruptors like OpenAI, Anthropic, Palantir, UiPath, and others to provide that layer.

In the near term, a pragmatic approach is to pilot agentic orchestration on top of existing systems, experiment with vibe coding for non-critical point solutions, and push vendors toward more open, unified data architectures.