No, they said in the WI EUG meeting that it [Knowledge Assistant] will be available to all Epicor clients both on-prem and SaaS. Knowledge assistant is only a cloud based tool.
1 Like
What is “it”? I’m only talking about the knowledge assistant, not sure what else we are discussing here.
it = knowledge assistant.
Knowledge assistant is a cloud based tool, it’s hosted in the cloud.
BUT, both on-prem and cloud customers can use it, “it” being knowledge assistant, a cloud-based tool.
I think that’s what @Randy is saying.
5 Likes
That isn’t what I was told.
3 Likes
Yeah I have no clue, I was just clarifying what I understood from Randy.
1 Like
Thank you @utaylor for clarifying.
@aosemwengie1 Why would Epicor keep a cloud-based tool away from on-Prem customers when there is no technical reason? Makes no sense and would only cause animosity between customers & Epicor.
I don’t know but I would guess there is a technical reason.
Knowledge Assistant is basically a custom trained version of ChatGPT with Epicor’s user guides and other documents. There is no technical reason for Epicor to block any of their customers from using it [Knowledge Assistant].
@bconner did a presentation overview of how they trained the tool at the last WI EUG.
3 Likes
There’s information online about Knowledge Assistant:
But, yes, to confirm @aosemwengie1 statement, we’re being told it is (currently?) only available on Cloud. I can’t talk to their plans on whether it will or won’t be available on-prem eventually.
My guess is that since it is “powered by Microsoft Azure OpenAI with ChatGPT-4”, it may require a browser pipeline to access all of Epicor’s documentation. On-Prem installations may or may not support that based on your installations/set-up.
Can’t say it won’t open up to on-prem (with additional set-up)… but it makes sense that “cloud” users can get there easily since cloud users are already living in that space.
8 Likes
Hello @dcamlin and everyone,
This is my first post to EpiUsers but I wanted to chime in. Epicor’s Prism, Knowledge Assistant, and related LLM-based AI solutions are only available on cloud infrastructure.
We understand that this shift in prioritizing cloud may be disappointing, especially for those who have been eagerly anticipating an on-premises version. You may also be wondering why. It would be reasonable to ask, for example, “Epicor releases other cloud-based services to on-prem, why not its AI-based technology?” AI and LLM advancements are moving very quickly. New techniques and methods are published every day. Consequently, Epicor is furiously experimenting on multiple fronts to bring the best of these watershed-moment innovations to you in a safe and productive way. While it is technically not impossible to extend Epicor’s LLM-based products to on-prem, it would be a considerable effort to build secure edge connectivity to a large volume of data for LLMs. Even after that effort, however, we believe it would not be the ideal experience, security, or performance. Moreover, we can provide a superior product and iterate on it much faster in the cloud by collocating it with a secure, managed environment where your data lives (Epicor cloud).
Below are some detailed reasons why ERP use cases for LLMs are better in the cloud:
1. Continuous AI Model Updates and Improvements
- Hosting Prism in the cloud allows Epicor to seamlessly deploy the latest AI models, updates, and enhancements without requiring action from customers. This ensures you always have access to the most advanced AI capabilities. Prism’s CICD release cadence will be regular and frequent, as opposed to the two major release cycles of its core ERP products. A managed cloud environment is ideal for CICD releases of AI innovation.
2. Enhanced Scalability and Performance
- The cloud environment dynamically scales resources based on demand, ensuring optimal performance even during peak usage without the need for costly hardware upgrades on your end.
3. Rapid Deployment and Time-to-Value
- Cloud-based Prism can be deployed quickly without extensive on-premises setup, allowing you to start realizing the benefits of AI more rapidly and stay ahead in a competitive market.
4. Access to Advanced AI Capabilities
- The cloud enables the use of advanced AI technologies that require significant computational power, such as deep learning and natural language processing, which may not be feasible with on-premises infrastructure.
5. Future-Proofing Technology Investments
- Cloud platforms are continuously updated with the latest technologies and innovations. Deploying Prism in the cloud ensures your investment remains current and competitive without the need for frequent on-premises upgrades.
6. High Availability and Disaster Recovery
- Cloud services offer built-in high availability and disaster recovery solutions, ensuring continuous access to Prism with minimal downtime and robust data backup capabilities.
7. Controlling Costs
- Beyond the quality of output, trying to run Prism on-premises could cause a cost overrun. This puts Epicor and our customers at risk. In the Epicor cloud we enable monitoring services to throttle the compute if its exceeding standard thresholds.
8. Mitigating Risk
- Epicor is committed to resilient, secure scalable cloud services for all of our solutions. AI is newer technology and therefore higher risk. Epicor is pursuing responsible use and working with Microsoft and other providers to optimize for the Epicor cloud to avoid misuse.
When you are ready to take advantage of the efficiency that Epicor’s LLM-based products offer, we look forward to assisting you on your journey to cloud.
Michael Atkisson, Ph.D.
Director of Product Management, AI
Epicor Software Corporation
17 Likes
Thanks for the explanation Michael and welcome to EpiUsers. We look forward to learning and hearing more from you. We are a friendly (yet passionate bunch) so expect some constructive feedback and push back on some of this stuff. Just ask @timshuwy 
But we really appreciate you taking the time to explain your reasoning behind this decisions.
14 Likes
One thing I will push back is on the difference between Knowledge Assistant and Prism. While I could get behind the points regarding data transfer and data ownership and security with Epicor Prism
I would argue that Knowledge Assistant is just looking at the Epicor Help Documentation, KB Articles and Training you’ve provided it to be better at troubleshooting / assisting users with Epicor questions. I do not understand based on your points why that particular product can’t be made available on Premise.
Unless I am missing something Knowledge Assistant doesn’t use your (our) database to help answer usability questions on the software it is (unless you are doing other direction) a fairly generic Epicor centric chat bot that can help users use and understand the product better no?
If its a question of cost, it could be upsold to us and even do it based on usage one could imagine.
Collaborate for example is available to on premise customers even though it is a SaaS product
13 Likes
And frankly, I thought that ServiceNow’s AI offering should be able to do that now since it’s holding all of that data and is already in the cloud. Just curious, why write our own?
At some point, vectorization will come to SQL Server on prem (already in Azure SQL, Postgres, etc). Are there thoughts about taking advantage of this for improved search capabilities?
And EDA, and Grow, and Avalara, …
Completely agree with the advantages of the cloud. I’m kind of a jerk about it here. But some companies run Kinetic in Azure on their own and get all these benefits. Would they also be able to use Prism?
CI/CD is also nice for users of said cloud services to make regular and frequent customizations that are easier to upgrade. We want what Epicor uses. Just sayin’.
7 Likes
This is a fantastic idea. I have already been using Chat-GPT for some help in epicor… BUT it typically references old pre-kinetic data that isn’t necessarily accurate any longer. The thought of just loading in the user guide hadn’t occurred to me. I’m definitely going to try that.
1 Like
Right, I don’t necessarily need AI to do my job - I want it to explain my users’ jobs to them so I don’t have to.
No proprietary data needed there.
It would be cute, though.
AI: Enter a part number on the sales order line. For example, part 755555 for a TICO PRO-SPOTTER TERMINAL TRACTOR.
3 Likes
And that’s what Prism will do (as I understand it) but that is separate from Knowledge Assistant i think
2 Likes
3 Likes
5 Likes
One more question:
For cost savings, some companies are utilizing less-demanding small language models for lower powered devices for similar performance. Will Epicor consider the same if it makes sense?
3 Likes
Frankly I just want to be able to ask Eddy the Epicare AI “At what version did the enter key stop totalling the sales order, and is there any resolution?”
That would be a great start.
6 Likes