Appian (APPN) CEO and founder Matt Calkins describes how he thinks people and companies should be using AI.
First of all, I love AI. I think it’s it’s going to change the world and we should all use it, but we need to use it safely. It’s important that we use it particularly for the the processes and the actions that can’t afford a mistake. You’ve got to use it without a mistake. It needs help and software can provide that help. It needs a deterministic framework because AI is fundamentally probabilistic and you need to be more reliable than that. It needs access to data because AI is only as smart as the data you serve it, so you should have access across the enterprise. It needs governance so you can track it and understand what it’s doing and improve it later on. And frankly, tokens are too expensive. So we need to allocate the right work to the right workers so that we can save cost.
The generative AI can now write code really, really well. So how does low code fit into the modern paradigm?
Uh, generative AI is a revolution in the way we’re all going to develop applications. It’s going to completely change things and natural language development is the future. However, when I say that to people, sometimes they just naturally assume that when AI builds an application, it’s going to build a stack of code, a unique stack of code. And I say, don’t do that. I call it a lonewear. You don’t want a lonewear because then when something goes wrong, there’s there’s no community, there’s no consultant you can call, there’s no help desk, there’s no executive commitment, there’s no company backing it up, you’re totally alone. That will not do for the most important applications. So you shouldn’t build alone ware. Instead, you should use AI, you should do natural language development, but you should build an application on an existing platform so that there’s a community and there’s support, and there’s help, and there’s upgrade patches and and you’re covered. So I think that the world is actually going to prefer an equilibrium where AI accelerates your ability to build an application, but when you do, it’s on a platform and thus safer. So that’s the that’s the sweet spot that we’re aiming toward.
How much of an Appian application can AI actually build today? So, can I describe for instance a business process that I want to uh have AI create um or have created, how much is AI doing that? How much is it participating in that?
It is astonishing how much AI can do at this point. However, I want to caution everybody, even if AI can make something right off the bat that blows your mind and it will, you still want to be able to iterate on it. You still want to be able to carefully audit everything AI has done, be sure the interface is right, the data structure is right, the roles, the responsibilities, the behavior. You want to check all that and then carefully iterate because no matter how good and fast AI is and it will impress you, it will still need to be adjusted over time. So you want to keep that high level of control on your application, but I’ll tell you right now, it’ll amaze you today.
