If you’re a New Zealand CFO or operations leader, you’ve almost certainly seen the headlines. “AI will replace your ERP.” “Vibe-code your own business system in a weekend.” “The age of enterprise software is over.” LinkedIn is full of it. So are the vendor pitch decks that used to promise “digital transformation” and now promise “autonomous enterprise.”
The noise is loud enough that the question is worth taking seriously. If a competent person can now describe a piece of software to an AI in plain English and watch it get built in minutes, does the traditional ERP still have a place?
The short answer is yes — but not in the form you’ve always known it. The longer answer is the useful one, because it tells you what to do with your ERP decisions over the next two to three years.
The real shift underway isn’t AI replacing ERP. It’s AI being embedded into ERP, to the point where the experience starts to feel very different even though the foundations are the same.
Gartner projects that by 2027, 62% of ERP application spending will include AI capabilities, up from 14% in 2024. That’s an enormous jump, and it tells you what the market is actually buying. It isn’t custom-built AI tools replacing finance systems. It’s finance systems with AI built into every workflow.
At the same time, Gartner predicts that up to 40% of enterprise applications will include integrated task-specific AI agents by the end of 2026, up from less than five per cent today. These aren’t chatbots. They’re agents capable of executing end-to-end workflows — matching transactions, flagging exceptions, drafting narratives, surfacing insights — inside the applications people already use.
Look at what the vendors are actually shipping and you see the same pattern. NetSuite’s 2026.1 release added eight AI features in one go: an Intelligent Close Manager, generative-AI bank transaction matching, AI assistants for reconciliation and planning, and a new connector service that lets external AI platforms (including Claude and ChatGPT) query NetSuite data securely using natural language. MYOB Acumatica’s 2025 R2 release shipped a new AI Assistant for natural-language queries, automated anomaly detection, AI-generated summaries, and an AI Studio framework for embedding custom AI workflows into business events.
None of this is “replacing” the ERP. It’s changing what the ERP can do and how people use it.
Still, it’s worth steelmanning the opposite view. The case goes something like this: if AI can now generate working software from a conversation, businesses will stop buying expensive, complex, integrated systems and simply assemble their own. Why pay for a single source of truth when every team can vibe-code their own niche tool?
The evidence on its face looks compelling. Harvard’s Graduate School of Education now teaches vibe coding to educators, noting how quickly capable tools like v0, Replit, Claude Code and Figma Make can stand up a functioning application. When developers with ten or more years of experience can build production-grade applications in weeks instead of months — and when business users without any coding background can generate usable tools from a few prompts — the logic does start to wobble.
Three things break the argument at anything bigger than a side project, and all three matter especially for the New Zealand mid-market.
The source-of-truth problem. An ERP exists, fundamentally, to hold a single, reconciled, auditable version of what happened in the business — what was sold, what was paid, what’s in stock, what was made, what it cost. The reason we don’t run companies on spreadsheets isn’t that spreadsheets can’t do maths; it’s that you can’t run a business where sales, finance and operations each hold their own version of reality. Generating more little apps on the side doesn’t solve that problem. It makes it worse.
The compliance problem. GDPR, NIS2, BCBS 239, the Cyber Resilience Act — and on this side of the world, our own Privacy Act 2020, IRD audit requirements and FMA expectations — all require traceability of how data is processed, stored and transformed. A generated script from a chat with an AI does not document its transformations, does not produce lineage, and cannot demonstrate compliance during an audit. “The AI built it” is not a compliance answer, and regulators aren’t accepting it.
The security problem. A large-scale scan by Escape.tech of 5,600 publicly deployed vibe-coded applications found 2,000 highly critical vulnerabilities, 400 exposed secrets (including API keys and access tokens) and 175 instances of exposed personal information. Georgetown CSET research found that around 45 per cent of AI-generated code contains exploitable vulnerabilities. These aren’t pet projects — they’re live production apps. You do not run your general ledger on that.
Even the people selling vibe coding tools acknowledge this. Salesforce’s own framing is useful: vibe coding works brilliantly in the “green zone” of prototypes and UI, but the “red zone” of core business logic and data layers is where developer expertise — and the structure of a proper enterprise system — remains essential. Industry analysts at diginomica put it well: vibe coding is best thought of as an accelerated sketchpad, not a system of record.
If you’re running a legacy or on-premise today, or you’re weighing up MYOB Acumatica or Oracle NetSuite, or you’re thinking about when to migrate — the implication of all this is not that you should wait for AI to make the problem go away. It won’t. The implication is the opposite.
The ERP isn’t going anywhere. But the ERP experience is about to change more in the next three years than it has in the past fifteen. Users will stop clicking through forms and start asking the system questions. Month-end close will shift from chasing spreadsheets to reviewing AI-suggested exceptions. Approvals will be triggered by agents monitoring your data in real time. The keyboard-and-mouse ERP of 2010 is going to feel very old, very fast.
That means your ERP choice today needs to be made with one eye on what the platform is doing with AI — not just what features it has now. You want a vendor that’s embedding AI inside the core, publishing an open architecture for external AI tools to connect in, and moving quickly enough that you’re not locked into a dying interaction model.
The risk isn’t that AI replaces your ERP. The risk is that you pick an ERP that can’t keep up with where the experience is going.
Our view is that the independent advisors who talk honestly about this will do better in this market than the ones who hype it. AI isn’t coming to sell you a new product — it’s already reshaping the products you’d buy anyway. Your job as a leader is to pick the right foundation, at the right time, with the right partner to see you through the transition.
If you’re weighing up the move off legacy/on-premise ERP, sizing up MYOB Acumatica against Oracle NetSuite, or you just want a straight read on where the market is going, we’d welcome a conversation. No pitch — just a grown-up chat about your specific situation.
1. Gartner Predicts 2026: The Future of ERP (overview) — RamBase.
https://www.rambase.com/resources/blog/gartner-predicts-2026-the-future-of-erp
2. Gartner Predicts 40% of Enterprise Apps Will Feature Task-Specific AI Agents by 2026 — Gartner Newsroom.
3. NetSuite 2026.1 Features New AI Close and Cash Management, AI Agents for EPM — Project Salsa.
https://projectsalsa.co.nz/blog/netsuite-2026.1-ai-fortified-financial-management
https://projectsalsa.co.nz/blog/netsuite-2026.1-ai-optimised-operations
https://projectsalsa.co.nz/blog/netsuite-2026.1-better-ai-outcomes-through-unified-data
4. Acumatica 2025 R2: Advancing AI, Usability, and Industry Innovation — Avanza Solutions.
5. Vibe Coding Security Risks: Enterprise Guide 2026 — BeyondScale.
https://beyondscale.tech/blog/vibe-coding-security-risks-enterprise
6. Can Vibe Coding Handle Your Organization’s Data? — Tale of Data.
https://www.taleofdata.com/blog/vibe-coding-enterprise-data
7. How To Use Vibe Coding Safely in the Enterprise — The New Stack / Salesforce.
https://thenewstack.io/how-to-use-vibe-coding-safely-in-the-enterprise/
8. Does Vibe Coding Have a Place in the Enterprise? Yes and No — diginomica.
https://diginomica.com/vibe-coding-place-enterprise-yes-and-no
9. ‘Vibe Coding’ May Offer Insight Into Our AI Future — Harvard Gazette.
https://news.harvard.edu/gazette/story/2026/04/vibe-coding-may-offer-insight-into-our-ai-future