For decades, enterprise technology advice was remarkably consistent: buy your ERP system, don’t build it.

The reasoning was straightforward. Enterprise Resource Planning systems contain decades of accumulated operational knowledge covering finance, procurement, manufacturing, inventory, compliance, reporting, tax, auditability and governance. So simple business software is actually an extraordinarily complex operational framework.
Most organisations that attempted to build ERP internally eventually discovered they were not simply writing software; they were reinventing the wheel by recreating business logic already used by platforms such as SAP, Oracle, and Microsoft.

The historical risks were substantial. Gartner estimates that between 55% and 75% of ERP projects fail to meet their original objectives. Those figures reinforce a long-standing reality: ERP failures are rarely caused solely by poor technology. The problem is actually that organisations seriously underestimate the complexity hidden inside operational processes.

AI Begins to Rekindle the Discussion

Generative AI, AI-assisted coding, low-code development and agentic workflows are dramatically reducing the cost of software development and customisation. Internal tools that once required large engineering teams can now be prototyped rapidly, while AI accelerates integration, testing, documentation and workflow automation.

The result? Organisations are reconsidering a question that would have seemed risky only a few years ago: Should they still buy large monolithic ERP systems, or is it now more viable to build their operational software in-house?

The answer is nuanced. AI does not eliminate the need for ERP, but it is changing where the boundary between ‘buy’ and ‘build’ should sit.
Historically, the argument for buying ERP systems was overwhelming.

Established vendors provided standardised workflows, mature governance, compliance frameworks, auditability, security models, integrations, support ecosystems, and upgrade pathways that individual organisations would struggle to replicate internally. Financial systems alone must accommodate changing tax rules, multi-currency accounting, segregation of duties, retention policies and increasingly complex regulatory requirements.

Which meant buying ERP reduced operational risk. Companies could focus on configuring processes rather than engineering foundational systems from scratch.

At the same time, fully customised ERP projects often failed for predictable reasons:

• Many organisations underestimated the complexity hidden inside operational workflows.
• A procurement process that initially appeared simple quickly expanded into approval hierarchies, supplier management, exception handling, audit trails, contract logic and dozens of integrations.
• Over time, internally developed systems accumulated technical debt and became dependent on small groups of developers with highly concentrated knowledge.

Those risks are still very valid ones.

In fact, AI may increase the importance of operational discipline rather than reduce it.

McKinsey research found that approximately 65% of advanced planning and enterprise optimisation programmes fail to achieve expected ROI, with poor data management among the leading causes. AI systems depend heavily on structured data, process consistency, and operational governance. Organisations with fragmented systems and inconsistent data definitions may therefore struggle to scale AI effectively, regardless of how advanced their models become.

That challenge is already visible across the market. Research from BCG found that 74% of organisations struggle to move AI beyond proof-of-concept deployments, while separate industry studies suggest only around a quarter of businesses have successfully scaled AI across the enterprise.

In other words, AI is not replacing the need for operational foundations. It is increasing their value.

What AI Changes are the Economics of Customisation

AI-assisted development significantly reduces the friction involved in creating operational tooling. Engineering teams can now build integrations, reporting layers, dashboards, workflow automation and user interfaces much faster than even five years ago. However, AI-driven development also introduces new governance and maintenance considerations. Research into AI-assisted software development has shown that while coding productivity increases, experienced developers can face growing review and maintenance overhead as systems expand.

This matters because many organisations do not actually want to replace their ERP core. They want to escape the rigidity and cost of excessive ERP customisation.

Historically, customisation within large ERP platforms was notoriously expensive. Companies often found themselves dependent on implementation partners and resellers who charged substantial fees for relatively small workflow changes. In many cases, businesses adapted their operations to fit the software rather than adapting the software to fit the business.

AI provides a viable solution to that problem. As development costs fall, organisations can increasingly build lightweight operational layers around a stable ERP core. Instead of heavily modifying the underlying system of record, businesses can create AI-assisted workflow tools that sit above it.

This is where the conversation becomes less about ‘build versus buy’ and more about making sure you choose the right foundation to build on.

The future is unlikely to belong to either fully bespoke ERP systems or to entirely closed enterprise suites. Instead, the advantage increasingly shifts toward open, extensible ERP platforms that allow organisations to customise intelligently without rebuilding core governance and compliance functionality.

This is especially important as AI development converges around open developer ecosystems. Most modern AI frameworks, orchestration platforms, automation libraries and data engineering tools are heavily centred around modern, open architectures. That means ERP platforms built on these architectures are becoming strategically attractive, because they integrate more naturally into emerging AI environments.

Open ecosystems also increasingly make economic sense. McKinsey’s 2025 Open Source AI Survey found that 60% of technology leaders viewed open-source AI models as having lower implementation costs than proprietary alternatives, but also found that user adoption lagged behind proprietary systems.

An area where AI can see lower implementation costs is in labour. Traditional ERP implementations will often have additional layers between those implementing and those developing the software. As AI accelerates software iteration and workflow customisation, direct collaboration with the platform’s owners and architects becomes increasingly valuable.

Organisations need partners who understand both the underlying codebase and the realities of AI-driven operational change, not simply implementation methodologies.
This does not mean every company should build its own ERP from scratch. For most organisations, that remains unnecessary and risky. Finance, compliance, governance and core transactional systems are still better served by mature platforms with proven operational resilience.

However, AI is making it increasingly practical to customise the areas where businesses genuinely differentiate. These include:

• operational workflows,
• scheduling,
• forecasting,
• manufacturing optimisation,
• logistics,
• customer operations,
• and decision support.

The Future is Hybrid

So the emerging model is a hybrid one.

ERP remains the system of record, while AI becomes the orchestration and operational layer sitting above it. Businesses standardise commodity functions, while customising the workflows that create competitive advantage.

AI has not resolved the ERP build-versus-buy debate. It has simply changed the economics.

The winners are unlikely to be organisations that attempt to rebuild ERP entirely from scratch, or those that lock themselves into rigid proprietary ecosystems. Instead, the advantage will increasingly belong to companies that combine stable ERP foundations with open architectures, AI-compatible development environments, and the flexibility to rapidly evolve operational workflows over time.

The future of ERP may not be fully bought or fully built. But it may well be open, composable – and AI-augmented.

Ready for what comes next?

The future of ERP isn’t about choosing between stability and flexibility. It’s about having a strong operational foundation that can adapt as new technologies emerge. Discover how Enapps provides a connected ERP platform designed to evolve alongside your business, including new ways to access and work with ERP data through AI.

Explore AI and MCP integrations
Book a demo

References

1. Gartner ERP failure statistics (via Rand Group)
2. McKinsey – Data management best practices for APS deployments
3. BCG AI scaling statistics (via Midfield Group summary)
4. TechRadar – AI adoption and scaling research
5. McKinsey Open Source AI Survey 2025
6. ArXiv – AI-assisted software development productivity research