ERX: the next generation of enterprise software

Everyone has a friend they keep meaning to see. "Let's plan dinner," you say, and you mean it. Then you say it again six months later, when you happen to run into them. The intention is always real, but the dinner rarely happens. A plan that never turns into action is just a nice idea.

That is why the phrase "enterprise resource planning" has never quite sat right with me. Words matter. Plenty of vendors are talking about what they plan to do with AI agents. Far fewer are doing it. 

When Gartner introduced enterprise resource execution (ERX), the next generation of intelligent and autonomous ERP, it gave a name to a shift already underway. The name is new. The work is not.

Gartner describes ERX as the next generation of ERP: a system that not only records what happened, but acts on what is happening. Where traditional ERP is a system of record, ERX builds agentic AI, governance, and real-time data into its core, so it can sense, decide and execute.

From system of record to system of execution: The move from ERP to ERX

For more than a generation, the job of enterprise software was to remember. Record the transaction, close the books, store the order, report what happened. That level of functionality has been enough for a long time.

But is it still enough? Expectations are changing. Fast. A system of record can tell you the line went down last week. A system of execution sees the shortage coming, secures a supplier that still meets the date, and keeps the line running. The value shifts from merely explaining the past to shaping what comes next: problems pre-empted, plans that adapt in real time, and processes that get more efficient on their own.

Make no mistake - the system of record is not going away. If anything, it matters more, because every action an agent now makes inherently takes its direction from that system. Getting 80 per cent of the right answers to prompts is often fine. Getting 80 per cent of the right actions posted to your ledger or released to your floor is not.

This difference is where the paths split. Exactly what directions are your system of record giving? For some vendors, ERX means bolting an agentic layer onto a generic core, with agents and governance perched on top. It may look fantastic in a demonstration, but that generic core system has already limited how precise the agents can be, how safely they can act, and how cleanly they integrate with the rest of the stack.

The true embodiment of ERX starts a layer deeper, with a natively integrated core system of record running on a data fabric built for AI, with semantic meaning, industry processes and a knowledge graph built in. After all, the agentic layer is only as good as the core beneath it.

Building the agentic enterprise

Becoming an agentic enterprise is not about a technology upgrade. It's about strategically transforming how your organisation uses its data and automates its decisions. 

A successful transformation requires AI systems that understand the realities of specific industries: the processes, data, regulations and terminology. General-purpose language models can reason about an industry, but executing business-critical decisions requires much deeper operational context.

You can't deliver that with clever prompting. It comes from industry-specific data models built into agents that understand the processes and priorities of a particular industry. That is the data model, not prompt engineering.

Even the most capable agents are only as good as what they can reach and direct. When organisations orchestrate across a disconnected stack held together with custom integrations, the outcome is a weak ERX, no matter how capable the agents appeared in the demonstration. 

Ultimately, trust is what will determine whether autonomous AI succeeds in the enterprise. You wouldn't sign a contract you couldn't read, so why would you allow agents whose actions you can't audit?

That is why governance cannot be an afterthought. Every AI-driven action should be auditable, traceable and compliant by default, with organisations setting the limits on what an agent can decide on its own. Human oversight should remain built into critical workflows, particularly where financial, operational or compliance decisions are concerned.

For the agentic enterprise, trustworthy AI is not a nice-to-have. It is the price of entry for allowing AI to act, and the line between an agent you demo and one you trust to run the business.

AI-powered by industry context

The demise of ERP makes great headlines. The notion of "do or die" to compel buyers to abandon logic and jump is exactly the tone being set by many newcomers to the AI space. Headlines proclaiming "ERP is dead" suggest organisations can point agents at a data lake or buy a general-purpose AI platform and skip the industry system entirely.

It is a compelling story, and parts of it are true. AI is dissolving some of the advantages traditional vendors have leaned on for years, and new entrants will carve out real space at the edges. But there is a ceiling. As Gartner notes, AI may remove many of the traditional barriers that ERP vendors have enjoyed, but it is unlikely to achieve the higher levels of autonomy on its own.

AI agents need the full power of enterprise systems behind them when acting at enterprise scale or making decisions that move money or halt operations. An appealing sales pitch may create agents quickly, but it leaves them lacking deep decision-making insight and context when it comes time to make critical business decisions.

The road to autonomous execution runs through industry context. Meaningful, deep context cannot be bought, downloaded, bolted on or prompted into existence. It is the foundation for becoming an agentic enterprise, and the catalyst for how quickly you can journey through the stages to ERX.