The greatest gap in AI today is between GPT and app.

There is a moderate learning curve in using foundational generative AI, things like learning how to prompt. This is where a lot of people start their journey, and it’s a great place to start. You can do a lot and get a lot done with little more than a prompt and a browser.

There is a moderate but slightly steeper curve in building self‑contained mini‑apps for AI – what I call things like Gems, GPTs, Claude Projects, etc. They still require relatively little technical skill, and can dramatically increase the scope and scale of what you can do with AI.

And then there’s the great, giant chasm between GPT/Gem and full automation. Yes, tools like Zapier, Make, and others can bridge that gap to some degree, but it’s still got vast hurdles to overcome.

Let’s look at a simple example. Suppose you want to programmatically access your Gmail account so that an automation can routinely process your email without you there. There are two kinds of authentication for this task, OAuth (the thing that makes a popup in your web browser asking if you want to sign in with your Google account) and Service Account – a special server‑to‑server version of authentication that doesn’t need you to sign in so often.

Setting up either one of these, especially the latter, is where the hurdles begin. Here’s a short list:

• To set up a service account, the Gmail API has to be enabled to a service account

• Which means you need a service account

• Which means you need IAM and Admin to provision it

• Which means you need a Google Cloud account, with billing turned on

And it is that technical barrier that’s preventing people from making the leap to full automations. It’s not an AI problem – this is an infrastructure problem. Some companies are starting to solve for it with things like connectors and other related services, but those are typically limited to one ecosystem or the most popular services.

Will this get better? Yes, unquestionably. As more companies pivot to the use of AI, providers have stronger and stronger incentives to make their applications AI‑ready, or be left behind and lose business to competitors. But those people who have technically‑minded team members who can bridge those gaps earlier will succeed faster and scale more.

Photo by me of the new I‑95 interchange on the Mass Pike.

#AI #GenerativeAI #GenAI #ChatGPT #ArtificialIntelligence #LargeLanguageModels #MachineLearning #IntelligenceRevolution