Model-as-a-Service (MaaS) - This is the best I could come up with for a vision I have for how AI will be adopted and leveraged in the proverbial "Enterprise."
It feels right at a concept and as a term, but there are also some very familiar traps that I see along the way that need to be considered as well.
If you close your eyes and think about how we as humans understand our world, we build models in our heads to explain what we see, how we feel, and how we believe things work.
Commercial enterprises can also be decomposed into various "models" such as business models, operating models, revenue models, etc.
And when you break down all of the hype that we're seeing and hearing around AI, most of the discussions center around "models" as well...Large Language Models (LLM), Machine Learning (ML) models, and other model-based concepts.
Models, models everywhere!
So where does this eventually lead to?
I think OpenAI is directionally correct in their thinking with ChatGPT Plugins.
I do believe this is the basis of the delivery and commercial models (more models!) that will emerge with AI services moving forward.
Rather than nuggets of intellectual property / value being hidden away in bloated applications or 7+ figure management consulting projects, the deep knowledge of individuals and organizations will be captured and made available in the forms of abstracted and distributable models.
Whether they take the form of plugins or something else completely, these models will be integrated with existing model architectures to form the collective knowledge and capabilities of humans, machines, and organizations comprised of both.
Thus my "Model-as-a-Service" label.
But having just lived through the Web 2.0 / Cloud / SaaS era, we are seeing that enterprises are struggling to maintain structures and flows that align to value because of the overhead and chaos created by the complexity of the tech sprawl that resulted from the unchecked (viral) distribution of SaaS applications and tools.
So what should we consider in our future state enterprise architectures to implement and leverage the incredible power of AI models without falling prey to the complexity trap that we fell into with SaaS?
Do you agree with the vision of model integration and architecture being a potential future state, or do you see things differently?
Who else has been thinking about these things?