Companies will be hard-pressed to find traditional ERP providers that supports on-prem anymore. Others, like Acumatica, offer a Private Cloud license.
It will be interesting to see how the market reacts. Someone may feed business rules/compliances into a model and create code for a certain vertical or even custom to a single business. If enough companies feel over a barrel with their SaaS provider that can’t provide value, quality or reliability, I’m sure some enterprising person will redefine how enterprise systems are delivered.
I still hope Epicor does change their mind and offer container or Microsoft Azure customer hosted options but at the same time, the more I experiment with AI models and workflows, the more I understand a reluctance for ERP and other enterprise systems to want to continue offering customer-controlled stacks.
The whole ‘headless-ERP’ concept and concerns, while initially sounding absurd, does seem more and more likely as an option for many operations.
Even outside of directly using AI for business processes, I think if most of us are honest about it, if we could have a strong Kinetic core, but custom generated UXs for our primary operations for our specific business processes and needs that don’t require as many ERP seats (i.e. lower ERP costs), at a similar or lower total cost, we would/will.
Funny you should mention the word “container”. Just got an email that 2026.100.5 hits Pilot June 7 with Production scheduled for July 10…at the same time as the Linux container upgrade.
We use Gen AI in our business’s daily and at no point has anything needed to be on the cloud apart form the Gen AI itself (and we are looking to host a smaller model on prem either way). It’s rest endpoint to rest endpoint. AI doesn’t care where it is as long as it can get to it. Their is fundamentally no different to my REST API endpoints than a cloud instance of Kinetic.
Just smells of the vendor putting themselves over the customer.
It seems like enough of a legitimate risk, I’m surprised there hasn’t been any players announcing a completely new framework that supports an AI-first customization scheme. AI could totally slash costs and make hosting on prem easier, but they are so afraid of losing what they have they will lock things down more and more. I expect it will continue to the point of additional API restrictions/cost.
Software companies seem to believe making software has gotten a lot easier based on firing software developers and ripping away control from customers. Seems like to me that will leave angry ERP devs and angry customers, making them very vulnerable to seeding their own replacement. (Which they will try to buy of course..)
Completely agreed. I have had similar experiences, except the vast majority of my AI usage and testing has been using local models. It is incredibly easy to interact with Kinetic via the REST APIs… which can quickly lead to the questions, and risk for Epicor and all other similar vendors, around why companies would even purchase and or keep additional expensive ERP seats for many users when we can now so easily just build external custom UXs and services for specific business processes? It has always been a challenge for ERP vendors, but AI completely changes the cost and practicality factors.
I think that many of the major ERP players are all trying to claim to be the companies with that framework, with SalesForce and SAP being very loud about it, except in Cloud deployments for control… but are yet to figure out how they can possibly balance the books and current owners’ revenue and profit expectations.
While our topic here is Epicor/Kinetic, it is really not too different to the same challenges many other organizations have, do, and will face regarding major changes and challenges with automations, new technologies, potential mass redundancies, and or a need for dramatically more efficient and effective business models to be viable and successful in the future. It is all almost always a lot easier said than done, particularly when it comes to the people element, and it seems inevitable that there will be mass disruption either way and that there is not a lot any of us can do except be as prepared as we can be.
PS: Whether, or not, there is an AI corporate/shares bubble vs a technology capabilities/shift bubble I think are two very different things. The capabilities of even just small local LLMs (i.e. Qwen 3.6/Gemma4) and agentic tools (i.e. Hermes Agent) are already pretty incredible.
We chose Vantage 8 because of its class leading SOA and that lineage has flown trough and improved over the years. “Flexibility at the point of delivery” is what we wanted and what we got. Everything from a program that literally just moves things between two warehouse to full blown SFDC.
Docker already did this. Having AI tell you how to get going but epicor in docker, I suspect it is something like this on your linux box:
docker pull epicor/kinetic
docker run -d
–name epicor-kinetic-app
–restart unless-stopped
-p 443:8443
-e KINETIC_ENV=PROD
-e DB_SERVER=mssql.whereveritis.local
-e DB_NAME=ERP_Prod
-e DB_USER=epicor_svc
-e DB_PASSWORD_FILE=/run/secrets/db_pass
-v /opt/epicor/config:/app/config:ro
-v /opt/epicor/customisations:/app/Server/Customizations
-v /opt/epicor/reports:/app/Server/Reports
-v /opt/epicor/logs:/var/log/epicor
-v /opt/epicor/certs:/app/certs:ro
–memory 16g
–cpus 4
epicor/kinetic-app:2024.2.x # ← this tag does not exist publicly
docker has been the single biggest shift in onprem software deployment in a decade. We are using it ramping up services that would have taken weeks to get doing before. Plus all free which helps
They had a session on this actually 2 sessions. They are working on a solution to address a lot of these concerns.
Targeted to start getting some early adopters next year.
Full on prem support through 2028.100 and then 2 years of regular support through December 31t 2029.
The details haven’t been given but they did say they are looking at containers (or something like it) and other solutions to allow both the flexibility we need while allowing us to take advantage of all the other offerings like Prism etc.
So yes it wasn’t in the keynote but they had 2 sessions dedicated to it one of them with the entire Epicor leadership answering questions the entire time.
I don’t think it was recorded, but this is what I got out of that session:
There is no specific timeline for fedramp med/high and other compliance requirements but they are working on it and updates will be communicated in the future. If public/gov cloud solutions don’t meet any customer’s compliance requirements, that customer can remain on active support even after December 31 2029 until Epicor can provide a compliant solution.
There might be some kind of hybrid solution coming but there are no details. Elsewhere in the conference it was conveyed that on premise containers will definitely not be provided so there seems to be a lot of confusion about this.
Extensive multi-hour maintenance windows will be going away for public cloud customers after the linux migration this summer. Gov cloud will move to linux next year sometime.
In 2027 public cloud customers will get to delay upgrades for a lot longer. The specific question asked was if the major release was in March, would a public cloud customer be able to delay that upgrade until the following February when internal teams would have bandwidth for testing. Epicor confirmed this would be possible.,
I want to believe it. I’d like to see stability in the platform. No more maintenance windows, no more random SSRStoppages, all that. We all would.
EDIT - speaking of maintenance windows, this just hit my inbox. Production outage this weekend…don’t recall seeing a previous announcement.
Kinetic - Americas customers are scheduled for infrastructure maintenance on database servers.
From : Saturday, May 23, 10:00 PM Local Datacenter Time
To : Sunday 24, 01:00 AM Local Datacenter Time.
During the maintenance window, application, printing functionality and read-only databases will not be accessible. All-clear notification will be posted after completion.
You will probably not find a more cloud fan-boi than I am. The allure for me has been unlimited compute and redundancy. However, I think we’re running up against Jervon’s paradox. The more we made compute available, the more demand it has created. AI is currently eating the cloud and all the infrastructure it can. People are not able to spin up resources today and are asked to move to other regions or even to local resources. In the meantime, communities are fighting new data centers and electricity is getting more expensive because of high demand. The token economy is upon us. Users are running out of their allotments daily now.
ERP providers (not just Epicor) are adding fuel to the fire by forcing users into the cloud while promoting AI, which does not help the situation. How does a manufacturing company mitigate the risk of their systems becoming unresponsive?
One obvious solution is to provide a method to utilize local resources/private clouds as a backup in case the hyperscalers brown out. ERP providers who don’t provide such an option are putting their customers, and their reputations, at risk.
Time is of the essence for many. The 2028-30 deadlines are forcing decisions to be made soon. ERP searches and implementations take time. Some companies have been on Epicor/Kinetic for years and still have not fully implemented the system. If there is no way to guarantee access to our production systems, companies have to find a plan to cover that eventuality.
Same with my company where I first got into Epicor (Vantage 8.0) back in 2006 and why I stay with Epicor through 4 other companies. It’s a great platform and I’m looking forward to seeing what’s next for on-prem folks.
I was happy to hear both of these too this year. We’d like to do the major upgrades during our slow season but not every company has the same slow season so this will be great for us SaaS customers.
Agreed, it’s coming to a head soon-ish on if the cloud can keep up as demand continues to spike. The cost of AI tokens are shooting up fast as well as other costs as the economy is struggling. Will be interesting to see where things head in the coming few years.
With local small LLMs becoming more capable I think that should free up some SaaS resource while also making it even more feasible for on-prem people (and I use that in the holistic sense not just ERP) to spin up their own server with a capable LLM.
I was happy to hear both of these too this year. We’d like to do the major upgrades during our slow season but not every company has the same slow season so this will be great for us SaaS customers.