Full disclosure: my company provides production support for JD Edwards, so I profit if you stay. Keep that in mind. I am not writing this to sell you, and I am not staging a debate. I have reached a conclusion and I am trying to find out whether I am seeing it correctly — and you, my peers, are the ones who would know. I sincerely want your read.
So put me in the hot seat. I am your CIO, and we are running JD Edwards EnterpriseOne. Not the latest release, we are a version or two behind, on 9.1 or 9.0 like plenty of shops, and the pressure to do something about it is real. Let me just say where I land and then show my work.
I would not migrate. And I need to be precise about that word, because it is the one everyone abuses. I do not mean I would freeze. If we are behind, I would get us current on 9.2, and getting current is not migrating, it is staying, on a fresher version of the same system we already own and control. When I say I would not migrate, I mean I would not leave the platform: I would not rip out JD Edwards and reimplement the business on Fusion or S/4HANA or another SaaS suite. That is the move the room is pushing, and that is the one I would pass on.
Instead I would get current, host it on a cloud I choose, preferably not one owned by a vendor with an ERP it would rather see me run, protect our data and our exits, spend a year cleaning up our own processes, and compress the run-rate until the system costs a fraction of what it does today. Then I would sit tight and watch what AI actually does, making a big move only if and when the case became undeniable, on our clock, from strength. That is the whole answer, and it is really a five-year plan, not a one-time decision. I have spent four pieces arguing that staying on JD Edwards is a stronger position than anyone gives it credit for. This is where I stop arguing and start choosing.
Get current, clean house, compress, then decide from strength. A five-year plan, not a one-time bet.
Why I would not migrate off
Let me steelman it first, because it is not crazy. Fusion and S/4HANA are real products with real roadmaps and AI already inside, and if my JDE estate were a tangle of unsupported customizations held together by two people about to retire, then staying would carry its own risk and leaving might be the lesser evil. But in most shops the honest math is brutal: leaving is not an upgrade, it is a reimplementation, millions of dollars over eighteen to thirty-six months, and your customizations and integrations do not make the trip.
And remember why the room is pushing you. The vendor on the other side is not being greedy, it is being a fiduciary. Its board answers to shareholders, and the most dependable way to grow the number it owes them is to move you into a cloud it meters and you cannot leave. I wrote a whole piece on that governance machine; the short version is that migrating hands your next twenty years to a company structurally built to enclose you.
That is the setup. What actually keeps me off the path is timing. You would be spending the biggest check of the decade to move onto the application layer, the screens and the workflows, right as AI starts to hollow that layer out, and you would hand over your cloud, your data, and your exits to do it. I do not lean away from migrating because it is expensive. I lean away because it is expensive and it buys the wrong thing at the worst possible moment. Show me a case where the new platform is a genuine force multiplier and not just a nicer hosting bill, and I will listen. I have not seen it yet.
And I have no illusions about where the pressure actually comes from. Yes, somewhere there is a room full of smart people spending real energy on how to corner me into SaaS. But the pressure I answer to is closer to home. It is my board, which hears the SaaS pitch everywhere — the AI story, the innovation story, this benefit and that one — and expects me to have an answer. And it is my own people, who want to be on the latest and greatest to keep their skills current. I get it, but I would tell them the career they should be protecting is mastery of our business and our industry, not fluency in a technology that reinvents itself every few years and resets everyone to beginner.
Here is what I have concluded about the room: they are nearly out of cards. The one real play left is the license audit, the classic lever for shaking an on-prem customer toward the cloud, and it is answerable: I keep my license position clean and reviewed by my own people before Oracle ever asks, so the audit finds a customer with nothing to settle. Past that, the trap only closes if I walk into it, and I am not going to. The one product they never rebuilt as a cloud is the one I run, and that is the single exit they were never able to take away.
Why I would not “do nothing”
This is the popular one, and it is not really a choice, it is a slow way of choosing the worst outcome. Keep the lights on with no plan and the run-rate drifts up quietly, year after year, until the budget is gone, the capacity is gone, and I am getting stampeded into a migration on the vendor's timeline instead of mine. Standing still does not avoid the decision. It makes it for me, badly, later, from weakness. So doing nothing is off my list, not because staying is wrong, but because staying passively is.
Two doors hand the timing to Oracle. The third keeps the clock, the cost, and the call in my hands.
What I would actually do: the five-year play
Which leaves the move I opened with. Let me lay it out as the plan it actually is.
Now, get current and take the wheel. If we are on an older release, step one is not Fusion, it is getting to 9.2. It costs a fraction of leaving, it is in-family, and it puts us on the 2037-and-beyond runway with continuous updates and no forced upgrade. Then I host it on a cloud I choose, and I choose one whose owner has no ERP it wants to move me onto, so I am never handing a single vendor both my application and the ground it stands on. I keep our data where we hold the key, because in an AI world the data is the asset and I am not going to rent access to my own records. What do I mean by that? Read this article to understand the situation brewing in Federal Court.
And let me be consistent here, because I have said publicly, in someone else's thread, that I would not recommend migrating from World to EnterpriseOne — and now here I am telling you to move to 9.2. That is one principle, not two positions. World to E1 is a real migration: a different product wearing an upgrade's name tag, re-implementation money spent crossing from one legacy product to another, right as AI hollows out the layer you would be buying. Getting current on 9.2 is nothing like that — same product, same code line, a fraction of the cost, and it buys the 2037 runway. The rule is one rule: get current where you stand, and do not pay migration prices to cross product lines while the AI question is still open. If I ran World, I would apply it the same way — stay put, hold my license, and lean on third-party support now that Oracle has moved World to sustaining-support-only.
Years one and two, clean house and compress. This is the real work, and it is two jobs running together. The first is cleanup. I clean the basement and the garage: prune the dead customizations, streamline the business processes, clean up the data, clean up the integrations, and document how the place actually runs, so the estate is well-understood instead of tribal knowledge in a few people's heads. That is also my answer to the fair question of who runs this when the veterans retire: a documented estate can be run by a much wider bench — my own hires, a partner, and increasingly AI, which reads documentation better than it reads folklore. And none of it is housekeeping for its own sake. It is how you get genuinely ready for AI, which rewards clean, documented data and punishes a mess.
And cleaning the estate is really about my people, not just the software. Once the place is documented and the knowledge is out of a few heads, I can change what I ask of my team. I do not want them living as task executors, closing tickets and running the same jobs by hand. Their value to me is as business leaders who help me change the business as the business changes, and anything below that is not where they should be spending their time. That makes it my job to mentor them there: raise the change success rate, drive the reactive break-fix time down until it is small, and push the balance of their work into proactive change instead of firefighting. This is not a nice-to-have. It is the capability that makes every objective after it possible.
The second is cost. I drive the operating cost down, hard, every year, until JDE costs a fraction of what it does now.
And here is the quiet gift underneath all of it: because Oracle never rebuilt JD Edwards as SaaS, I still possess the software and control the infrastructure it runs on. And here is the irony that makes it the best card in the deck: the same Oracle now working to herd every customer into a cloud it meters is the Oracle whose own past leadership left this one product in my hands, unconverted and unenclosed.
Whatever you make of the vendor, that decision did JD Edwards owners a real favor — and note what this is not, an abandoned product I am clinging to. Oracle still sells JD Edwards today and supports it to 2037, a current, for-sale system that I also happen to own outright. That ownership is the root of everything on this page.
Because I possess it, I hold my own data, I choose the cloud it runs on, and no one can meter my access or flip a switch. I hold the license and the install, so the system is mine to run whether or not I keep paying Oracle: in SaaS a missed payment is a light switch, but here it is a negotiating position.
The rest follows from that. The 22 percent maintenance bill turns optional instead of mandatory, and third-party support becomes possible at all — firms like Rimini Street and Spinnaker run it at roughly half, more available than the market lets on, though I would weigh that against the loss of vendor updates — security patches included — rather than treat it as a free win. The firms cover that gap with virtual patching, and my CISO signs off on the posture before my CFO signs off on the savings. Every dollar I stop spending just to stand still becomes optionality, a war chest.
Oracle's yesterday quietly defeats Oracle's today.
Around year three, decide with data, not fear. By now I know what AI actually did to my core processes, and whether the destination everyone is selling is real or already obsolete. I would not sign a seven-figure migration while that question was still open. I decide then, from a compressed base, funded and timed on my terms.
Years three to five, live with the payoff. Either I have moved, on my clock and from strength, or I am running lean and free for another decade. Either way it was my call, not the vendor's.
The AI question, and why staying wins it
The one objection worth taking seriously is that JD Edwards will not get the AI Oracle is stuffing into Fusion. Good. That is a reason to stay, not to leave, and it is the part most people get backwards.
The AI that changes your business does not live inside your ERP. It reasons across everything at once, your finance and your operations and your customers, not the sliver trapped in one application. That makes it an enterprise capability, not an ERP feature. So embedded ERP AI is really an upsell for something that never belonged in the ERP to begin with: SAP already meters it by the unit, and Oracle gives it away inside Fusion precisely because it is bait to pull your data into a cloud it controls. Either way you pay, in cash or in custody.
And be clear about what AI does not take. Strip away the screens, the workflows, and the reports, and the ERP's one durable job is the system of record: the ledger, the transactions, the auditable truth of what happened. That job never goes to an LLM, because a ledger must be deterministic — the same entry, the same answer, every time, provable to an auditor — and probabilistic models are built to be almost right, which in a ledger is just wrong. The job is permanent. The incumbent is not, and I hold that thought honestly: if AI ever gets good enough to write the transactional software itself, it will build on clean, documented data that somebody owns. Everything in this plan — the cleanup, the documentation, the data I keep the key to — is how I make sure that somebody is me, and JD Edwards does the system-of-record job as well as anything sold today while I wait to find out.
Stay on JD Edwards, own your data, and AI is never a line item on an ERP invoice. You point the best models in the world at your whole business yourself, on data you hold, without anyone's permission or meter. The system with no AI bolted on is the one that leaves you free to do AI the right way, at the enterprise level, where the value actually is.
Every advantage here is something a migration spends millions to give away. Staying keeps all six.
What would change my mind
I want to be honest about where my lean is soft, because it is a lean, not a religion. Three things would actually move me.
A functional dead end. If the business genuinely outgrew what JD Edwards can do, if we needed a capability the product will never have and it actually moved the P&L, then staying would be stubbornness, not strategy. That is a real reason to move, and no amount of cost compression answers it.
Data I no longer control. If I did not actually own my data today, if I had already let the system get half-enclosed in someone else's cloud, then one of the strongest reasons to stay is already gone, and holding the line is worth less than I am claiming. Picture the version of me who took the vendor's offer a few years back to lift JDE into its cloud and “simplify hosting.” The database, the backups, and the encryption keys now live in the vendor's tenancy, so pulling a complete, usable copy of my own data out runs on the vendor's tools, the vendor's timeline, and maybe the vendor's fees. That CIO has already handed over the leverage I am telling you to protect, and for him the honest math tips the other way.
AI that actually delivers, and not from an ERP. The AI worth uprooting for will not come from an ERP vendor bolting a copilot onto its suite. If it comes at all, it comes from the dedicated AI companies. So if one of the serious AI players offered something that genuinely reinvented how the work gets done, and not just a nicer place to run the same screens, I would give that real consideration and move, gladly, from strength. What I would not do is migrate to another ERP for the AI it claims to have, and I would not pay an Oracle partner ten thousand dollars a month to bolt some JDE-and-AI integration onto the side either, because I can already point the best models in the world at the data I own, right where I am. Both are the same mistake: renting access to intelligence I can get myself, on data I already hold.
But notice what is not on that list. A discount is not on it. A cut enrollment fee, a waived first year, a migration credit Oracle dangles to get me onto the new product, none of that moves me, because it is a sales incentive, not a business case. Nothing is free, and a cheap on-ramp is just lock-in you pay for later, once you are inside and metered. And a capacity crunch is not on it either. If my team cannot keep up, the fix is more hands, more automation, or a partner who runs the work, not the biggest project of the decade dropped on top of the people already underwater.
That is how I see it. That is what my gut says: stay, compress, and decide from strength. But I have been wrong before, and you are closer to your own system than I am. So, what am I missing? Please bounce it back. [email protected]
This is the job we do: get current, keep the cloud and the data in your hands, and show where automation can reduce recurring work. See pricing and terms
John Mathieu is Managing Partner at Allari, which provides ERP production support for JD Edwards, SAP, Oracle, NetSuite and others. OpenBook® shows clients every ticket, hour, and dollar, defensible to the CFO and the board.
