What’s Ahead - Alien Processes, Domains, and Data Models

Excerpt from the chapter on Modeling Business Processes and Domains

This section closes out my upcoming chapter on Modeling Business Processes and Domains. While obviously a speculative piece, I can’t help but think about what’s next, and it warrants inclusion in this chapter as a consideration of what’s coming at all of us. What happens to business processes, domains, and data models? I speculate that AI will change these in ways that will seem alien to us. It’s still early days, though. The cool thing is, we’ll soon find out as agents become more integrated into every organization.

The whole chapter will be available very soon for paid subscribers.

Thanks,

Joe

---

The game of Go was thought to be unbeatable by AI due to its massive search space, with more possible moves/states than atoms in the universe. Then, in 2015, AlphaGo burst onto the scene and beat the best human Go players in the world, first Fan Hui and then Lee Sedol. World-class Go player Mok Jin-seok remarked about AlphaGo, “At first, it was hard to understand, and I almost felt like I was playing against an alien.” I had similar feelings when seeing AlphaGo decimate its human opponents - we’re entering a different phase of humans and machines where things will feel increasingly strange and alien.

At the time I write this, we’re in a similar AlphaGo moment for processes, domains, and data modeling. It’s still very early days for organizations deploying AI agents. While today’s LLMs and agents are prone to hallucinating and have limitations with memory and context, these shortcomings are being improved at warp speed. Non-LLM methods and approaches for AI are also being explored. The temptation to deploy agents that can work nonstop 24/7 is valid and something many organizations are pursuing. And with hundreds, thousands, or millions of agents running within and across organizations, AI models will have ample opportunities for improvement.

We often talk about human tacit knowledge, the “gut feeling,” and “it’s all up here in my head,” know-how, and tribal wisdom that people acquire through years of experience. It’s the reason processes function even when the documentation is a disaster. While this may seem like science fiction today, it’s worth considering how AI agents might eventually learn enough about an organization to begin creating their own processes and domains. Might agents eventually develop their own form of “machine tacit” knowledge, where agents run thousands or millions of cycles, and find optimizations that are statistically perfect but narratively and physically incomprehensible to humans. Much like those Go players, we will soon find ourselves operating alongside “alien” processes and domains that we didn’t design and can’t easily articulate or understand.

This shifts the entire burden to data modeling. No matter what, data is still a raw material, and a data model will be required. It might not be a traditional data model as we think of it, but it’s still a data model. The ways we’ve modeled data so far might change as a result of the use cases, processes, and domains in which agents operate or create on their own. Rather than primarily modeling “what” happened as we do today, in a world where AI invents processes and domains, a data model might become more about “why” and “how.”

What new data models and formats might AI create to serve its purposes? Today, metadata provides the structural glue for providing this context. AI might also invent new data formats for its purposes. We’re moving from a world of human-readable tables, optimized storage formats for apps and analytics, and metadata. The new world might be AI-generated high-dimensional vector spaces or dynamic ontologies that agents modify in real-time to suit a specific task. Again, this is science fiction today, but I expect we’ll see indications soon.

Much as Go players feel they are playing against an alien, we should be open to the possibility that agents might develop their own processes and domains, and build and maintain data models that seem alien to us. Mixed Model Arts now refers to a hybrid world where human-defined domains, processes, and data models must coexist with those invented by machines.