← Video Library · ChatGPT & GPTs
Creating Your Own AI Assistants on the OpenAI Platform
Published November 8, 2023 · 420 views on YouTube
Key takeaways
- The OpenAI Playground lets you create multiple named AI assistants, each configured for a different task such as AI optimization audits, narrative work, or business growth.
- You build an assistant by setting its instructions to whatever recipe, workflow, or series of prompts you'd otherwise repeat manually.
- Assistants can be fine-tuned with your own uploaded knowledge and data so responses are grounded in your specific business content.
- Testing an assistant in the Playground replaces the old process of manually copying and pasting a prompt recipe into ChatGPT or Claude one-on-one.
- A single assistant workflow, such as an AI workflow optimization audit, can take a user's inputs and generate structured insights and opportunities automatically.
How do you create your own AI assistant on the OpenAI platform?
You need an account with access to the OpenAI Playground. From there you create a new assistant, give it a name, select a model, and write instructions containing whatever recipe or workflow you want it to follow. You can then test the assistant directly inside the Playground before using it on real inputs.
What can you configure for each assistant?
Each assistant has a name, a selected model (the video uses GPT-3.5 Turbo), and a set of instructions where you place any recipe, workflow, or series of prompts you want the assistant to execute. Multiple assistants can be created side by side for different purposes, such as an AI optimization audit, a narrative-focused assistant, or a business growth assistant. Assistants can also be fine-tuned by uploading your own knowledge and data so their responses are grounded in your specific business content.
How do you test an assistant with real input?
- Open the assistant in the Playground and confirm its name, model, and instructions are set.
- Provide a real example input, such as a business submission describing a current workflow (for example, a pharmaceutical company describing its content briefing, research, client feedback, and quality-check process).
- Click to add the input and run it against the assistant.
- Watch the assistant work through the instructions using the context provided, producing analysis such as identified pain points and workflow opportunities.
- Review the output, such as suggested content-creation automation, predictive analysis, or AI copywriting opportunities.
Why does this change how you use AI assistants day to day?
Before assistants, the process was manually copying a specific prompt recipe into ChatGPT or Claude and running it one-on-one each time. With assistants, that recipe lives permanently in the instructions field, so you simply add new input and run it. This also opens the door to having one assistant hand off to another, for example an assistant that surfaces an opportunity handing off to a second assistant that turns it into a script or additional content.
Full video transcript
Hey, Darby here with Gen University, and stick with me for the next three minutes or so. I’ll give you a very quick walkthrough of how you can start to create your own AI assistants within the OpenAI platform. You’re going to need an account and access to the OpenAI Playground to see what you see here on my screen.
With the recent updates from OpenAI and the next stage of where GPT is going, you’re now able to create AI assistants that can operate tasks for you. In multiple assistants, as you can see here, I’ve got one for AI optimization, one for narrative, one for business growth, and a number of different blueprints. We just started making these with the announcements starting November 7th, so we’re still playing around with how to get these assistants to talk to each other. There are a lot of really great things we’re going to see these accomplish within your organization, and it’s very easy to get things set up.
There’s a lot you can do in terms of fine-tuning, uploading your own knowledge and data to tune these assistants on your specific content, on what you’re doing within your business, for a better frame of reference. Then you can test it very simply inside the Playground. Here I have my assistant, with the name of this assistant being an AI optimization audit, and I’ve got GPT-3.5 Turbo selected.
This is a workflow I’ve created to answer basically a formula: giving people insights as to where AI can help with auditing and improving their existing operations. I’ve got this over here in the instructions, so if you have any sort of recipe, workflow, or series of prompts, you can simply create those here in the instructions.
As an example, if you’re on our newsletter, I sent an email out this week about doing a free AI workflow optimization audit. People submitted about a 30-word example of their business — one described a pharmaceutical company developing content for healthcare practitioners and patients, and a workflow that ensures accurate, engaging articles through initial briefing, publication prioritization, research, client feedback, and quality checks. That’s one example among a bunch of responses in our email inbox.
The timing was perfect for these assistants to come out, because before this I was going into either ChatGPT or Claude, copying and pasting in a specific recipe, and running it one-on-one. Now I can simply click add to the run, and you’ll see ChatGPT and the assistant going through all of these commands, using the user’s inputs — current workflow, content creation, briefing, research, client feedback — what’s the pain point, time-consuming research, manual revisions, and quality checks. The workflow takes that data, applies its analysis based on the context provided, and accelerates the ideation process for what the opportunity is: content creation automation, predictive analysis, AI copywriting for scalable content production.
There are some ways to start, but you can go much deeper into these ideas. As we continue to build and test and figure out the best use cases for these AI assistants and our own custom GPTs, there are a lot of ways we’re going to be able to take this initial workflow and have one assistant talk to another assistant — one that helps find an idea and creates a script or more content around a specific opportunity. This is just one example; you could use this for emails, for marketing copy, for data analysis inside your business — basically whatever GPT can do, you can customize these assistants to help you do on a repetitive level much faster.
That’s going to lead into a lot more automations and opportunities to integrate AI into things you’re already doing, without the copy and paste, streamlining and consolidating all the places you go back to for creating content within GPT. I’m excited about where workflows are going, and we’ll be creating more content and demonstrations on different use cases for AI assistants within the OpenAI platform. Make sure you subscribe to us here on YouTube, click like on this video, and catch us on the other side. All right, we’ll see you in the next video. Keep coming back, Darby from Gen AI U. See you on the other side.