The old selection model is partially obsolete
The traditional ERP selection process evaluates functional fit, TCO over 5-7 years, technical architecture, vendor viability, and implementation partner. All still matter. But two new factors now dominate the real competitive differentiation between platforms:
AI-readiness of your data. All four major vendors have now shipped AI copilots that can answer questions, produce insights, and automate workflows — but only on the data you give them, at the quality level you maintain. The platform is now partially a function of how well you will actually be able to use these capabilities, which is a function of your data governance posture.
AI cost over time. AI features are priced on top of core licences — per-user, per-query, or consumption-based. Cost models that ignore this will understate TCO by 20-40% over five years. Finance teams doing ERP business cases need to model AI consumption the same way they model cloud run-rate.
What each major platform actually does with AI
All four major platforms (Dynamics 365, Oracle Fusion, SAP S/4HANA, plus our own VE ERP) now offer broadly comparable AI capabilities — embedded copilots, natural-language querying of operational data, anomaly detection, forecasting, and document extraction. The differences are in depth, licence models, and data-governance posture.
Microsoft Copilot is the most mature end-user-facing experience, benefiting from deep integration with Microsoft 365. Oracle and SAP have been more aggressive on specific domain models (finance, supply chain) and on enterprise-grade governance. VE ERP focuses on practical automation for mid-market operators — less depth per feature, more coverage across the workflows where AI actually saves time.
The selection implication: AI is no longer a differentiator in the way a "Is the vendor investing in AI?" question would suggest. Every major vendor is. The real question is which platform's AI approach best matches your data maturity and operating model.
Where implementation risk has moved
Historically, ERP implementation risk concentrated in customisation and integration work — the places where the packaged platform met the client-specific reality. That risk has not disappeared, but a new risk category now sits alongside it: AI adoption risk.
AI features typically fail to deliver value for one of three reasons: the underlying data is too inconsistent for the models to be useful; the workflows that would actually benefit from AI are not the ones the platform's AI is trained on; or the operational change needed to actually use the AI is not budgeted in the implementation programme.
Selection processes that ignore these three risks will be surprised 12 months post-go-live when AI feature adoption is 15% rather than the 60% the business case assumed. The fix is straightforward: include AI data readiness, AI workflow fit, and AI change management as explicit selection criteria.
The practical recommendation
For mid-market enterprises (€20-200M revenue) reviewing ERP platforms in 2026, we make three concrete recommendations:
1. Do a data-readiness assessment before the vendor selection starts. If your master data is inconsistent, the AI capabilities of any platform will underperform. Data readiness is the foundation, not a subsequent remediation.
2. Evaluate AI cost over 5 years at realistic consumption levels — not the vendor's launch pricing. Include AI query costs, copilot licensing, and the compute for model fine-tuning if relevant.
3. Evaluate the implementation partner's AI delivery experience separately from their core platform credentials. Shipping AI to production is a separate skill set from shipping ERP to go-live — and many partners have not yet built the muscle.