Home
Dynamics 365
Business Central
It’s the end of the year and I’m thinking about AI…

It’s the end of the year and I’m thinking about AI…

Stefano DemilianiBusiness Central1 month ago4 Views

I usually use the short holiday working period between Christmas and the new year for fully recharging my batteries and to relax my mind. This is the period where I usually slow down the usage of my computers and where I increase my favourite sport activities (running and cycling, more cycling now). But when the evening comes I need to admit that I’m not able to avoid spending some minutes reading news on socials or on my favourite IT channels.

And in the last days I’m reading and thinking a lot about AI and its latest evolutions…

There’s a sentence coming from the CEO of NVIDIA that I listened months ago in a conference that I want to report:

Generative AI is the single most significant platform transition in computing history. In the past 40 years, nothing has been this big. It is bigger than PC, it is bigger than mobile, and it is going to be bigger than the internet by far.

And this is a continuous transition…

Only until few months ago, for me the usage of AI in real-world applications followed mainly this pattern:

Select an AI provider (OpenAI, Azure OpenAI, Google, …).
Deploy an AI model in the cloud.
Create applications that interacts in different ways (accordingly to the business case) with the deployed AI model.
Review, track and monitor the AI usage costs.

But now the AI world is changing again…

In the ever-evolving world of artificial intelligence, open-source AI models and private AI solutions are making significant strides. These advancements are not only pushing the boundaries of what AI can achieve but also ensuring that privacy and security are at the forefront.

Open-source AI models have become a cornerstone for innovation, providing developers and researchers with the tools they need to create cutting-edge applications without the constraints of proprietary software.

I’ve worked a lot in the last months with open-source AI models in real-world applications and they have really opened my mind. LLaMa 3.1 for example is still impressing me with its performance and versatility and I’m currently still using it in many scenarios. And every day there’s something new in this field. Mistral AI for example is a company that focuses on providing open and portable generative AI models for developers and businesses. They offer a range of models that can be customized and deployed in various environments, ensuring flexibility and independence from cloud providers. And they all work really good…

And between all this ocean of open source AI news, few days ago DeepSeek (a Chinese AI firm) released the new DeepSeek-V3 model that breaks all the previous limits and represents a significant new advancement in the field of open-source AI models, offering both high performance and big cost-efficiency.

My first think of these days is the following: I think it’s now the time to take open-source models seriously…

Together with the open-source AI revolution, there’s another aspect that in the last months consumed a lot of my time: private AI, alias the possibility to host your own AI models (LLMs) and infrastructure in your local network.

Running AI models locally on your own networks (and devices) has several advantages, including enhanced privacy, control over data and the ability to operate without relying on internet connectivity. Plus, you can have zero cost for AI usage!

Can a locally-hosted LLM be a valuable choice for a company investing on using AI for their business tasks?

This is my second cause of thoughts these days.

I think that there are many scenarios where an LLM model hosted in the cloud is the better (and the go-to) choice (it’s difficult or impossible to host extremely powerful LLMs in local resources because they will require extremely high GPUs, CPUs and memory resources).

But when analyzing the AI solutions that often customers require, I often see that the requested tasks can be performed also by using open source and locally-hosted LLMs. So why not using them? And this is exactly where I’m working today… I see (and I have) lots of scenarios where a locally-hosted LLM is a great choice (yes, also for AI tasks connected to Dynamics 365 Business Central).

If you have a good hardware you can run your AI solution fully offline (or make your local AI cluster like I did in this pic some weeks ago):

Why not taking also this advantage?

I’ve personally tested lots of scenarios where using a locally-hosted LLM fits perfectly. Just to spoil one of those scenarios, here is a fully local AI agent monitoring camera images coming from different warehouse locations, determining the filling status of each locations and then changing the way to handle pickings in Dynamics 365 Business Central:

Conclusion

I don’t want to bore you too much with my thoughts on these topics, it’s the end of the year and it’s better to think about relax, panettone and sparkling wine. But if you are fascinated or immersed in these topics, I think sharing is good for everyone.

Before closing, I want to share some conclusions I’ve done (thinking internally in my company and also talking with a large ecosystem of partners). Companies (expecially in the ERP / Business Central world) are doing baby steps in AI adoption. I think that:

Succeeding with AI will depend more on people’s adoption than the technology itself.
People lack the knowledge and skills to effectively implement and scale AI across the enterprise. And this must change soon…
Organizations and partners must work on an AI strategy for their business and they should have courage to take advantage of technologies that change faster than people can adapt.
Having an “AI tier” in their solution should be considered the biggest goal for the new year for every organization.

I’m extremely curious to know your opinions on these topics, so feel free to share with me your thoughts or your experiences (or what you would like to know more) using comments here or on my socials. They can help a lot (and thanks for that ).

Ok, now that I have shared my thoughts with you I can more peacefully go and toast the new year.
My best wishes to everyone of you for a wonderful and happy New Year.