# ERP Selection in 2026: Where AI Actually Matters > Source: https://netray.co/erp-help/cio-erp-selection-2026-with-ai > Publisher: Netray (netray.co), AI and integration engineering for ERP and manufacturing systems. > Section: ERP Help ## Short answer Treat AI as one evaluation column among many, not the deciding factor: fit, data model, industry depth, and total cost of ownership still decide most ERP selections. Test AI claims live against your own data during the demo, not the vendor's canned dataset, and separate "embedded copilot" marketing from features that ship and work today. ## Kind cio ## System cio ## Erp ERP selection (all vendors) ## Module IT strategy / ERP procurement ## Query how should we evaluate AI when selecting a new ERP in 2026 ## Alternate Queries - ERP selection checklist 2026 AI - how to compare ERP vendors AI capabilities - does the AI copilot matter when picking an ERP - ERP RFP questions for AI features ## Applies To New ERP selection or replacement projects, any vendor, 2025-2026 evaluation cycles ## Steps ### How to weigh AI in an ERP selection process **Items:** Score the core ERP fit first (industry process fit, data model, integration surface) with AI capability as a separate, secondary column, not a tiebreaker that overrides fit. Ask every shortlisted vendor for a written list of which AI features are generally available today versus roadmap/preview, with contract language that only commits to GA features. Bring 3-5 of your own real questions and a messy real report to the demo, and require the vendor to run their copilot against your data live, not a scripted demo dataset. Ask what happens to AI feature pricing after the current promotional period; most vendor copilots bundle a metered consumption charge (credits, calls) that shows up in year two. Check whether the AI features require you to be on the vendor's latest cloud release train; for legacy ERPs still running on-prem, many copilot features are cloud-only and effectively unavailable. Evaluate whether a layered AI approach (grounded Q&A and agents added on top of your chosen ERP, from Netray or a similar vendor-neutral provider) covers the gap if the ERP's native AI is immature. Score data readiness separately: no ERP's AI will perform well on your data if master data, chart of accounts, and item records are inconsistent, regardless of vendor. Get 2-3 reference calls specifically about the AI features, not just the core ERP, since AI is the newest and least proven part of most current ERP roadmaps. ## Sections ### Separate the ERP decision from the AI decision **Body:** Every major ERP vendor now markets an AI copilot: SAP Joule, Oracle Fusion's embedded AI, Microsoft Copilot for Dynamics 365, NetSuite's AI features, Infor's Coleman AI. The pattern in RFP responses is consistent - AI capability slides are polished, but functional coverage is uneven and changes release to release. Treat the core ERP fit (industry depth, transaction volume handling, integration model, cost of ownership over 7-10 years) as the primary decision, and AI as a secondary column that can break a genuine tie but should not override a poor fit. The risk of inverting this is real: teams that picked an ERP partly on AI roadmap promises have ended up with a platform that fits their business worse, while the promised AI features slipped a year or two. AI moves fast enough that a strong core ERP with a thin native AI layer today can be paired with a third-party grounded AI layer later; a wrong core ERP choice is a decade-long problem. ### What to actually test in the demo **Body:** Vendor demo environments are curated. Ask for the copilot to answer questions against your own exported data (a sanitized subset is fine) or, better, a live connection to a sandbox you populate. Bring the kind of question your controller or planner actually asks: "which open sales orders will miss their ship date given current on-hand and open POs" or "show me vendors where our average payment terms don't match the contract terms." Vendors that only look good on pre-scripted demo questions are showing you a demo, not a product. Also test failure behavior: what does the copilot say when it does not know the answer, or when the data needed spans two modules that are not integrated in your instance? A copilot that hallucinates a plausible-looking wrong answer is worse than one that says it cannot find the data - this single behavior tells you more about production-readiness than any feature list. ### The consumption-pricing trap **Body:** Most vendor AI features are priced separately from the core ERP license, usually as a metered add-on tied to API calls, tokens, or "AI credits." Get the current price per unit and a written cap or estimate for your expected usage volume, and ask what happens if usage exceeds the included allotment. Several CIOs have reported the AI add-on becoming a larger unplanned cost than the module license itself once user adoption actually picks up, because usage is hard to forecast before go-live. Build AI consumption cost into the multi-year TCO model in the RFP response requirement, not as a footnote. If a vendor will not commit to a price ceiling or a clear unit price in writing, treat that as a real cost-control risk during contract negotiation, not just a formality. ## Pitfalls - Choosing an ERP primarily on AI roadmap slides rather than shipped, generally available features. - Letting the vendor demo AI against their own curated dataset instead of your real, messy data. - Ignoring that many native copilots are cloud-only and unavailable on the on-prem or older-release version you may actually be buying. - Not asking for a written price per unit of AI consumption before signing. - Assuming AI quality will be uniform across modules; most vendors ship AI first in one or two modules (often service or finance) and lag elsewhere. - Skipping data readiness assessment because AI functionality is graded separately from the core system evaluation. ## Ai Angle Netray's ERPray is designed to be evaluated the opposite way from a vendor's bundled copilot: it is grounded read-only Q&A and agent workflows layered on top of whichever ERP you select, including SyteLine, LN, M3, or NetSuite, so a weak or immature native AI feature on the ERP you pick does not need to block your AI roadmap. In a selection process, this means AI capability can be scored honestly as secondary to core ERP fit, since a grounded layer can be added afterward regardless of which vendor wins. ## Faq ### Should we delay our ERP selection to wait for AI features to mature? **Answer:** Generally no. Core ERP fit and total cost of ownership are decade-long decisions; native AI copilots are improving on a roughly annual release cycle across most vendors. Select on fit and plan to layer AI capability in afterward, either from the vendor's roadmap or a third-party grounded AI tool. ### Is it fair to score vendors on AI features that are still in preview? **Answer:** Score them separately from GA features and weight them much lower, since preview features can slip, change scope, or be discontinued. Ask for the vendor's own GA date commitment in writing before giving significant RFP credit for a preview feature. ### Does AI capability differ meaningfully between cloud and on-prem ERP versions? **Answer:** Yes, substantially for most vendors. Native copilot features are usually built against the vendor's cloud data platform and are often unavailable, delayed, or feature-limited on on-prem or older-release deployments, which matters if your selection includes an on-prem option. ### What is the single best test of a vendor's AI maturity during a demo? **Answer:** Ask it a question it cannot fully answer from the data available and watch what it does. Mature, well-grounded AI says so or asks a clarifying question; immature AI guesses and produces a confident, wrong-looking answer. ## Search terms ERP selection 2026, ERP RFP AI questions, AI copilot ERP comparison, how to evaluate ERP AI features, ERP vendor AI roadmap, AI consumption pricing ERP, ERP selection checklist CIO, AI ready ERP evaluation --- Talk to an engineer: https://netray.co/contact?intent=erp_ai | Machine catalog: https://netray.co/agent/catalog.json