Over the past year, ERP vendors across the industry have converged on strikingly similar AI messaging: your AI runs on the system you already have, today, with no migration required. It’s a compelling pitch, especially for manufacturers running mission-critical operations on ERP versions they have no near-term plan to replace.

It’s also, in practice, rarely the whole story. Look past the headline claim into the licensing terms, the support documentation, or the fine print of a vendor’s own event announcements, and a consistent pattern emerges: the AI is available, but usually only once you’ve committed to something bigger – a cloud migration, a specific paid tier, or a future release date. This isn’t a criticism of any one vendor. It’s a category-wide pattern worth understanding before you’re in a sales conversation, not after.

Below are six of the claims manufacturers hear most often, what the fine print usually says instead, and the questions worth asking any vendor – including us.

Fact vs. Fiction: 6 Claims Worth a Closer Look

Myth: “AI runs on your current system, no migration needed.”

In the fine print, this usually requires already being on, or committing to, a broader migration journey. It’s rarely sold as a standalone AI purchase – it’s bundled into that larger decision, and activation is often paired with a requirement to redirect a meaningful share of your existing maintenance spend toward the migration itself. In many cases, the AI isn’t really the product being offered – it’s the incentive being used to move a migration decision forward that was already on the vendor’s roadmap for you, not a capability delivered strictly on your own terms.

Myth: “You can stay on your current on-prem/legacy version indefinitely – there’s no pressure to move.”

Several vendors have published or signaled dated schedules for on-prem feature development, after which on-prem versions typically shift to security-patch-only support – sometimes with an extension option available, priced at an increasing annual surcharge. New AI investment is concentrated entirely in the cloud product line, so “no pressure to move” quietly becomes “no new capability unless you move.”

Myth: “You get the full AI suite.”

Marketing often describes a broad, capable agent suite in a single announcement. What’s actually available to provision on day one is frequently a partial or “select” tier, with the fuller capability set scheduled for a later GA date, sometimes months or quarters out. The agent count referenced in press coverage often describes a future state of the roadmap, not what a customer can turn on immediately.

Myth: “AI usage pricing scales simply with what you use.”

Most vendors include a fixed usage allotment with the base subscription. Once that allotment runs out, which is commonly reported as happening faster than expected, overage is frequently priced at 150-200% above the base rate – a cost curve that’s difficult for a growing operation to forecast at the time of purchase, since it isn’t usually the headline number in the pitch.

Myth: “The cost of getting AI live is straightforward, no hidden surprises.”

For heavily customized or highly configured environments, common in manufacturing ERP estates that have been tailored over many years, some vendors’ own implementation guidance openly acknowledges this can mean “significant remediation effort.” That’s a real, often unbudgeted project in its own right, closer to a re-implementation than a simple activation or upgrade.

Myth: “Our AI connects into your ERP natively, out of the box.”

In several cases, what’s actually delivered is a developer toolkit rather than a finished, native connection: the AI runs through a separate cloud or knowledge layer sitting alongside the core application, and connecting that layer back into a customized or on-prem environment is left to the customer or a systems integration partner to build, using their own time, budget, and resources.

A Framework: 6 Questions to Ask Any AI Vendor (Us Included)

None of this means AI-on-ERP claims are false. It means the claim alone isn’t enough information to make a buying decision on. Before you sign anything, ask your vendor – any vendor:

1. Can I buy this as a standalone product, or does it require a bigger migration commitment?
2. Is there a hard deadline forcing me off my current version?
3. Am I getting the full agent set, or a limited subset today?
4. Is there an overage penalty if I go beyond my usage allocation?
5. Is the implementation cost fixed and published, or scoped after I sign?
6. Is this native or bridged – do I have to build the connection myself?

If a vendor can’t answer these clearly, that’s the answer.

Where ChampionAI Stands on All 6

We built ChampionAI to hold up against these questions, because we expect manufacturers to ask them.

• Standalone or bundled? Champions run on any current QAD version today. No migrate-first toll, and no requirement to redirect existing spend toward a broader migration to unlock them.
• Hard deadline to migrate? None. Value is delivered on your timeline, on your systems – migrate later if and when it makes sense for your business, not on ours.
• Full set or subset? You pay for, and get access to, exactly the agentic Champions you need – each clearly scoped and role-based, integrating with your existing systems. Champion Assist, QAD’s embedded AI assistant, works differently: it’s automatically enabled for customers on Adaptive ERP and SRM, as a clearly scoped bundle – not a blurred one.
• Overage penalty? No. Usage is priced at a flat rate whether you’re inside or beyond your allocation.
• Surprise implementation cost? Fixed paths, quoted upfront as part of the SKU – not an open-ended project you discover the cost of after signing.
• Native or bridged? Off-the-shelf Persona Champions, configured to your workflows with pre-built ERP integration. Where workflow extensions are needed, they’re built by a dedicated QAD team on the customer’s behalf, not left for you to assemble.

We’re not asking you to take our word for it. We’re asking you to hold every vendor, including us, to the same six questions.

Want to see what a no-gate approach to AI actually looks like for your operation? Talk to QAD