If this is you, read this email.

Because in this email, I'll give you my 2 cents on where I think the AI indsutry is heading, what to focus on, and what to let go.

AI Agents

These have been around for a while, but recently, new and new companies are popping up with (semi) autonomous agent solutions.

Like Devin, an AI Software Engineer that can write and fix code for you, build and deploy apps, learn new tools and even train new AI models.

They have raised a $21 million Series-A led by Founders Fund.

In a nutshell, these AI Agents are usually an LLM wrapped in powerful software, giving it access to data, the internet, other tools, etc.

The LLM gets data from somewhere, does something with it, and gives you back the output in an INTEGRATED way.

Most of the times its not one AI, but multiple LLMs, sometimes even different models talking to each other, superwising their own work and helping each other.

My suggestion: try them, use them, but if you are alone, don't build them as a software as a service company.

Especially seeing how OpenAI develops things, its entirely possible that GPT-5 will have functions that allow it to be onboardod to a workspace, learn how a company operates and then assist the employees with tasks.

The Custom GPT Actions were a hint that OpenAI is pushing into this direction. The Microsoft Copilot is also a hint towards this.

User interfaces like the ChatGPT chat interface will start disappearing and we will be communicating with AI tools where we already communicate with our teams. On Slack, via email or even a phone call.

We're not far from having AI agents in our "Favorites", calling them, giving them tasks and they can do it on their own because they will have access to our data.

Which brings me to the second part:

Software Integrations

This is a big one in my opinion. Integrating LLMs into the tools we already use.

This is the Zapier, Make, Bubble etc. No-code bot building we keep talking about.

Things happen digitally in most businesses now.

Data is created digitally all the time.

Must create a new meeting for a project?

As a human, I know the context of a project, who is involved, what the goal is overall, what the goal of the meeting is, and I can send out a calendar invite.

An AI can do that too, if I can integrate it to our project management system. The more moving parts the project has, the more difficult the integration.

This is basically a step down from the AI Agents.

And these integrations are now possible.

For example, with Zapier Central, you can create bots that have access to Live Data.

For example, you can connect a Google Sheet into an AI bot, talk to it, ask it to update rows, do calculations, and you don't need to worry about updating the Sheet, as you or anyone in your team adds data to it, the AI bot will also see it.

Zapier has more than 6,000+ integrations + you can have an HTTP request or a Webhook for those apps that don't have native integrations yet.

Make currently has 1,739 native app integrations, and there are also integrations you can buy from 3rd party developers (plus HTTP and Webhook of course)

And these are the apps we use every day. Notion, Airtable, Slack, Stripe, Quickbooks, Gmail, GDrive, Dropbox, ConvertKit, Mailchimp, SamCart, Clickfunnels, Meta, and so on, I could keep going forever.

Basically the template is:

IF [something happens] in [any app], THEN [do something] in [any app], THEN [do something else] in [any app], etc.

For example:

IF [I upload a new video] to [Dropbox], then [transcribe it] with [AssemblyAI], then [summarize it] with [GPT-4], then [save both as separate Google Docs] in [my Google Drive folder].

All this without writing a single line of code. And this is nothing new, no code apps have been around for a LONG time. I remember when I built my first online business in 2019, I was using Zapier right left and center.

The transformation is that now you can integrate AI into these workflows. Before LLMs, you had to hard code everything.

Every automated email fit a template. Now every automated email can be unique to a person receiving it. And that's life changing.

My suggestion: ABUSE THIS POWER, learn no-code tools like Zapier or Make by building stupid little funny useful or trash automations for yourself.

For god's sake go have some fun!!

Like our Rainmaker Slack bot that sends a funny GenZ Slang message EVERY TIME someone makes a purchase:

And it always brings a smile to my face when I read these goofy messages.

So yeah, don't forget to have fun while you make things.

The big reason why I highly encourage you to build these no-code apps for yourself and learn this technology is because they are LLM agnostic, meaning if GPT-5 comes out, you can just swap a dropdown from "gpt-4-turbo" to "gpt-5" and everything works the same with updated output quality.

So no new AI model announcement will destroy your automations. On the contrary, it will make them better.

And the third thing I wanted to talk about was vectorized knowledge base solutions, but I see this email became lengthy anyways, plus my laptop is at 2%, plus I really want to watch the replays of the Australian F1 Grand Prix practice sessions, plus I'm exhausted and my back hurts from carrying this company the whole week.

I appreciate you for reading the emails I write, for sending your feedback, questions and comments. Keep them coming, I read and respond to all of them.

Have a nice weekend, talk soon!

Dave

P.S: Want to master LLMs and integrate them into no-code apps? Click here and I'll show you how to do it.

‍