Mind Readings: IT and Infrastructure in Generative AI

In today's episode, let's talk about the plumbing. Well, not literal plumbing, but the pieces that surround generative AI. Large language models like the ones that power tools like ChatGPT and generative AI models are insanely capable tools. They they seem magical because they're good at language and we communicate in language all the time. We almost every task we do uses language in some capacity.

And that in turn means that we think of these things as magic. We think that they they could do anything because we do almost everything with language in some fashion. But there are a lot of things we do that are not language. Data processing, copying and pasting, math, executing code, although the code itself is a language, storing data, retrieving data, organizing data. These are all really important tasks that are that are not language based.

Language may be involved in them, but the fundamental task itself is not language. Painting is not a language. All these tasks, like data processing, like execution of code, like math, are critical tasks that AI needs. AI needs access to them. And that's important because in our rush to try to incorporate AI and everything, generative AI in particular, in everything, we're forgetting that we need this, these other infrastructure pieces.

For example, suppose you have a SQL database that contains your customer information, and you want to use that information to build, I don't know, sales enablement emails with generative AI, like with ChatGPT. You cannot ask ChatGPT, uh, just, hey, can you connect to my database and write me a sales email for it? It will not work. Right. You you can't ask any model to do that right now.

Your SQL database has to have a port open, hopefully a secured one, for code to interface with. You need to have code that can speak SQL when whatever variant your database speaks, which ChatGPT and tools like it can help write, but then you need a service, an API service to call to it, or a WebSocket service, or something to connect to it. And then you need to make that API call open AI compatible, open API compatible, so that a tool like ChatGPT in the custom GPTs can talk to it. And when you're done, or when it's done with that, you probably should have some way of getting the results out of a tool like ChatGPT back into your sales system. So imagine you're in your CRM and or you're in your your, yeah, you're in your CRM, you're in your Salesforce automation, your marketing automation, and you want to do sales enablement using generative AI.

There are a lot of pieces, a lot of pieces that you need to connect to that system. Very few of those pieces are things that AI can do by itself. Almost all of them are actually IT. They're IT combat systems, right? They're the plumbing, they're the infrastructure that you need to connect the data from one place to another, get that data flowing like water, so that a generative system can use it.

And people forget that. Now there are plenty of services that can connect to uh some of these parts, tools like Zapier or Flow or Integrately or Pably or Workado or whatever. But it's still plumbing, and those aren't automatic either. You still need someone from IT to help you rig all that stuff together. The point of this is to be cautious with our expectations around generative AI.

That these things are not magic wands. These things are not they're not genies that can fulfill any wish. They're word prediction engines. And they still need connections to other stuff. If you want to unlock the true value and power of generative AI, you need to connect it to your data.

And to do that, you need the plumbing and the infrastructure to be in good working order. If it's not, you need to get that fixed up as soon as possible because a competitor who has better plumbing than you gets data to their AI systems faster than you and gets results into market uh sooner than you. And that's bad. So go fix your plumbing. Go make sure that IT is on the same page as marketing or sales or finance or whoever's using generative AI in your in your company, in your system, and make sure everyone has the data that they need to feed to generative AI.

That's gonna do it for this episode. Thanks for tuning in. Talk to you next time. If you enjoyed this video, please hit the like button, subscribe to my channel if you haven't already, and if you want to know when new videos are available, hit the bell button to be notified as soon as new content is live.