Enterprise Platform Services Reimagined: AI in the ERP, CRM, and Core Platform Layer

For thirty years, the ERP and CRM layer has been the enterprise system of record: the place where a transaction is written down after a person decides to make it. AI agents change what that layer is for. When an agent reconciles an invoice, adjusts a forecast, updates a case, or routes an approval, the core platform stops recording decisions and starts making them.

The shift is arriving fast and from inside the products themselves. Gartner predicts that 40% of enterprise applications will be integrated with task-specific AI agents by the end of 2026, up from less than 5% in 2025, and that agentic AI could drive around 30% of enterprise application software revenue by 2035, surpassing $450 billion. SAP, Salesforce, and ServiceNow are each shipping agents into the platforms enterprises already run.

That creates a decision most platform owners have not yet made. Vendor agents arrive configured for the vendor’s view of the process, with access to the vendor’s data. Enterprise processes rarely sit inside one platform, which means the agents that matter most have to work across systems the vendors do not control.

A system of record is designed for accuracy after the fact. Controls sit around data entry, approvals are routed to named people, and audit answers the question of who changed what. The assumption underneath is that a person decided, and the platform captured it.

Agents change where the control has to sit. An agent can execute inside the platform in the time it takes a person to open the record, so the controls that used to wrap human data entry have to be re-pointed at a non-human actor: who authorized this agent, what scope does it hold, what did it read before it acted, and what would have stopped it. Accountability does not move. The process owner still answers for the outcome. What changes is that the platform, rather than a person at a screen, is where that accountability gets enforced.

This is why bolting a conversational interface onto an ERP changes very little. A chat window over a system of record is still a system of record. What changes the economics is an agent that reads context across systems, decides, acts, and leaves a trail that survives an audit.

McKinsey’s research on scaling agentic AI puts a number on how few enterprises have got there. Nearly two-thirds have experimented with agents, and fewer than 10% have scaled them to deliver tangible value. Eight in ten cite data limitations as the roadblock.

Enterprise processes do not respect platform boundaries. Order-to-cash crosses CRM, ERP, and a billing system. Employee onboarding crosses HR, ITSM, and identity. Claims cross a core administration platform, a document store, and a payments rail.

An agent confined to one platform can only automate the part of the process that platform can see. That is why so many vendor-native agents demonstrate well and deliver narrowly. The value sits in the handoffs, and the handoffs are where the context breaks.

McKinsey describes the fix as a semantic layer: a machine-readable definition of what business entities are, how they relate, and what rules govern them, sitting between raw data and the applications that consume it. Without that shared foundation, agents act on incomplete or conflicting interpretations of the same data, and error rates rise with scale. In an ERP and CRM estate, the semantic layer is the thing that lets an agent know that a customer in Salesforce, a payer in the claims system, and an account in SAP are the same organization.

Three decisions determine whether agents in the core platform layer produce compounding value or a portfolio of disconnected pilots.

1. Extend or replace. Whether to build agentic capability on top of the platforms already in place or to move the process onto something new. Replacement programs carry multi-year timelines that outrun the technology cycle they were meant to catch.
2. Consolidate or integrate. Whether to reduce the number of platforms holding the same business entity or to connect what already exists. Agents are more sensitive to fragmentation than dashboards ever were, because they act on what they read.
3. Lifecycle ownership. Who runs upgrades, integrations, and regression once agents are part of the estate. Platforms now ship capability continuously, which makes every release a validation point for agent behavior as well as for user workflows.

Underneath all three sits the same question: where do vendor-native agents end and enterprise-governed agents begin? Each vendor governs its own agents to its own standard, which leaves the surface between them ungoverned unless someone owns it.

The pattern repeats across enterprises. Each platform team adopts the agents their vendor ships. Each set works inside its own boundary. None of them coordinate, no single team owns the surface where they meet, and the audit trail for a cross-platform process is assembled after the fact from three different logs.

A governed orchestration layer is what turns that into a strategy. AAVA™, Ascendion’s agentic AI platform for enterprise software engineering, governs how agents are configured, how they coordinate, how they touch enterprise systems, and how their outputs are validated. It is to enterprise agents what GitHub is to enterprise code: a centralized, governed home for the constellation of agents the business runs on.

The point is not to displace what SAP, Salesforce, or ServiceNow ship. It is to hold the processes that cross them, under one set of controls, with one audit trail.

Ascendion is an AI-native software engineering services company that built its own agentic AI platform and runs it at enterprise scale inside client environments. Services-as-Software is the category Ascendion created and leads, and Engineering to the Power of AI™ is the method. AAVA is the operational system that runs the method in production.

Ascendion’s enterprise platform services cover the full lifecycle rather than the implementation alone. On Salesforce, CRM platform engineering runs from roadmaps and optimization through custom development, enhancements, release management, and day-to-day support, across Agentforce, Sales Cloud, Service Cloud, Marketing Cloud, Revenue Cloud, Data 360, and MuleSoft, with Centers of Excellence built on governance frameworks that tie Salesforce evolution to business strategy.  Ascendion joined the ServiceNow Consulting and Implementation Partner Program in 2025, covering ITSM, ITAM, and ITOM alongside customer service, field service, and source-to-pay workflows. ERP modernization services on SAP run to landscape assessment, structured managed services, and process standardization across business units.

The outcomes show up in the core processes rather than in the platform metrics. A leading bank in the Philippines was running its core banking system alongside a separate business process platform, with talent moving data between them by hand. Ascendion implemented Salesforce Service Cloud across the customer service process and built 40+ integrations connecting it to the core banking and BPM systems, eliminating the manual transfers that were delaying loan, card, and account requests. When the bank subsequently acquired another institution, the same work unified the service agent interface across both.

The ITSM layer shows the same pattern. For a multinational professional services network delivering audit, tax, and advisory services, Ascendion built out the ServiceNow stack for process compliance and traceability, automated catalog task workflows to shorten turnaround, and established event data correlation across Azure Monitor and ServiceNow for predictive rather than reactive support. Ticket volume fell 20%, and new application releases went out faster on the same platform.

At scale, the same discipline runs in regulated environments. For a US healthcare payer, 650+ AAVA agents serve 39 million Americans across all 50 states with zero downtime. For a top-five global bank, Ascendion delivered $300M in annual savings while improving service quality for more than 10 million customers, including $20M in immediate savings on customer service AI.