From system of record to system of intelligence

ERP systems have been the backbone of enterprise operations for more than six decades. They started as glorified ledgers, ways to record what had happened. Over time they grew to encompass HR, procurement, supply chain, manufacturing, and finance, becoming the single source of truth for the entire business. The problem was that “truth” still had to be laboriously extracted, interpreted, and acted upon by humans.

That’s changing fast. Gartner now predicts that by 2030, over half of routine ERP tasks will be autonomously executed by AI – across finance, supply chain, and HR. The research firm also projects that by 2027, 62% of ERP application spending will incorporate AI capabilities, up from just 14% in 2024. The direction of travel is unambiguous.

But what does that shift actually mean in practice, and where does the “death of ERP” narrative go wrong?

The death of the front end (not the ERP)

Microsoft CEO Satya Nadella stirred the pot not long ago when he suggested that the traditional SaaS model may be “dead” – that AI agents, not software interfaces, will become the primary gateway to enterprise tasks. The sentiment rippled through the industry.

He has a point, but it’s a narrower point than it first appears. What AI is likely to kill is not ERP itself, but the front end of ERP, the dense, menu-driven interfaces that require specialist training and months of onboarding. The dashboards nobody fully understands. The reports that take three analysts and a consultant to interpret. Those are the things being disrupted.

Underneath that interface layer, the core of the ERP – the data model, the business logic, the transactional engine – becomes more valuable, not less. As one McKinsey analysis put it, enterprises deploying AI agents are not replacing their systems of record; they’re building orchestration layers on top of them. The ERP increasingly becomes the trusted database layer that AI draws from, writes to, and reasons about.

This is a profound architectural shift. ERP moves from a system you interact with to a system that works on your behalf, surfacing the right information, triggering the right actions, and flagging anomalies before they become problems.

Why determinism is actually an asset

Here’s a nuance often lost in the noise surrounding AI’s enterprise potential: AI is genuinely bad at being deterministic. Large language models are probabilistic by nature – they’re excellent at reasoning, synthesis, and communication, but they can’t be trusted to consistently execute a financial close or a procurement workflow with zero error.

ERP systems, by contrast, are built to be deterministic. They execute the same process the same way, every time. That’s not a limitation, it’s a feature, and increasingly a competitive differentiator. A “reality check” is occurring among those who recognise that LLMs lack the consistency required for critical business processes in regulated industries.

The result is a natural division of labour: AI handles interpretation, recommendation, and communication; ERP handles execution, audit trail, and data integrity. These systems don’t compete, they complete each other.

SAP has already seen this, evolving its Joule assistant from a copilot into an autonomous agent capable of building and deploying AI skills on top of the existing ERP core. Microsoft’s Copilot is on the same trajectory. In both cases, the underlying ERP isn’t being replaced, it’s being given a voice.

The modular ERP and the end of monoliths

Perhaps the most significant structural shift underway isn’t AI at all – it’s how systems are built, enabled by AI.

The monolithic ERP, where a single vendor’s software governs every process from purchase order to payroll, is giving way to something more modular. A cloud-native financial core at the centre, with best-of-breed applications plugging into it as business needs change. The rigidly customised, impossible-to-upgrade behemoths that tormented organisations for decades are being replaced by something genuinely agile.

AI is the connective tissue that makes this work. When your inventory system, your CRM, and your financial platform speak different languages, AI acts as the interpreter, synthesising data across systems, filling gaps, and presenting a coherent picture without requiring years of integration work. The emergence of user-friendly AI tools is already enabling businesses to extract value from their ERP ecosystems without needing deep technical expertise at every step.

This democratisation matters. It shifts power away from the small group of specialist users who “know the system” and towards everyone who needs to make decisions. Every user, as one analysis put it, becomes a power user.

The real risk: underutilisation

If there’s a cautionary note in all of this, it’s not that AI will destroy ERP, it’s that many organisations will invest in AI-enhanced ERP and fail to capture the value.

The barriers are well-documented: poor data quality, lack of internal ownership, insufficient change management, and resistance to new ways of working. AlixPartners found that while AI tools can deliver 20–30% productivity gains in software development, most organisations are struggling to translate those gains into strategic business value. The technology is moving faster than the organisational capability to absorb it.

This is the real challenge for the coming years. Not whether AI will work (it increasingly does) but whether businesses can build the governance, data foundations, and change readiness to benefit from it. ERP implementations have always been as much about people as about technology. That hasn’t changed.

The opportunity in the disruption

So: is AI the death of the software industry? Not exactly. It is, however, the death of some software; specifically, the software that exists to compensate for the limitations of older systems, to perform tasks humans should never have been doing manually, and to present data in ways that require human interpretation at every step.

The software that survives, and thrives, will be the software that knows what it’s for. The ERP systems that endure will be the ones that lean into their deterministic strengths: auditability, integrity, consistency. They’ll shed their clunky interfaces and become quieter, more powerful, and more embedded in the fabric of daily operations. Less visible. More essential.

That’s not a gloomy future. That’s what progress looks like.

The ERP isn’t dying. It’s finally growing up.