Quick prediction. AI as we know it is done. Done as in: cooked, not finished. The next step, and my prediction is: to progress the impact of AI and business optimization, we cannot just stick and thrust AI onto existing platforms and data. We need to retrofit our current systems and data sharing to be suitable and usable to reuse and AI.
Funny story: this too was our advice at the dawn of the age of internet advertisements, SEO, and accessibility, respectively 25, 15 and 10 years ago. Nobody cared, so let’s see what actually happens.
So, the mashup is back!
Because the real AI innovation for the coming period lies in how legacy software and platforms share their data, logic, and structure to usability for robots, AND each other! That’s what will actually make AI useful, because it gives AI the chance to make structure and networks of choices and relationships explicit and imagine new ones based on the needs of people. For those in the know: first we need some API‑ and schema‑driven process mining and process optimization, for and with AI.
This step will make possible the leap from leaving breadcrumbs to using Google Maps, from gossip channel to journalism, from faking to making, from having a computer and having access to the internet. Some might call this a ‘world model’, a meta view on several AI models, but I think it’s more. To me it’s like creating a connection between what the world actually does and how AI currently sees it, including the relationships. Especially relationships!
And that structure, that grid, schema, and meta view, those informal but content‑relevant connections are the crux to being truly useful. These threads do exist inside the AI data models, but they’re not being fed to them. The models just statistically guestimate, powered by a healthy dose of smart math. Right now, all the underlying systems that provide AI are not truly connected to the infrastructure, the APIs, and the data layers. As a result, AI is guessing based on how we report and talk about our interactions.
We should now provide a map of how services like SalesForce, SAP, and Booking work. Not to mention the plethora of logistics, health, production, admin, legislative and research data streams which can not only inform AI, but could also result in a good think in sharing data, mashing it up and creating new services.
We could go even further: map out how all those processes, datas and interactions are connected to the rest of the world, to other systems. That’s how we move from a statistically meh guess, without any original thought to a genuinely valuable service where new insights can feed our imagination until the far future. That’s how we move from searching to finding, from automatic to automagical.
When Legacy Systems Start Talking: The Hidden Grid That Can Transform AI
