Most of you who already use an AI assistant have probably noticed, just like I did, that you often have to give the same prompt over and over again for almost identical tasks. And this quickly becomes tiring and counterproductive.
If we have repetitive tasks that we regularly perform with an AI assistant, there are options that can streamline the process and save time. In this article I present two such options: one simple, fast, and compatible with most AI models, and another that requires some setup but then delivers a high level of automation.
Optimization with Reusable Context Blocks
These context blocks help customize how the AI assistant responds, allowing us to define a structure or style once and then reuse it anytime by simply pasting it into the chat.
Instead of manually writing the instructions every time about how we want to receive the answers, we define them once and keep them handy.
Context blocks are AI model agnostic, meaning they can be used with any AI assistant, regardless of provider, as long as they’re already configured and saved somewhere.
These blocks can cover various nuances we want to set at the beginning of a conversation:
- preferred response style (structured, narrative, etc.);
- the format in which we want to receive the information;
- details about us, so the assistant can personalize the answers;
- information about a personal or professional project, so that when the chat starts, the necessary context is already prepared.
This way, we no longer need to explain every time that we’re working on project X, which needs to do Y, and that we want it to perform certain specific tasks. You get the idea.
GPTs
In Claude they’re called Projects, in Copilot they’re called Agents, and in ChatGPT they’re called GPTs.
Important: Copilot agents should not be confused with ChatGPT agents, which I discussed in a separate article. Copilot agents are natively integrated into many apps in the Microsoft ecosystem – for example, when we configure a Copilot agent, we can grant it access to certain locations or resources in SharePoint.
From here on we’ll focus on ChatGPT’s GPTs, which offer a general way to customize an assistant to perform a repetitive task.
A GPT is essentially a configurable bot with a specific role that can be reused whenever we need it, without having to provide the full prompt every time. Once configured, it automatically performs what it was set up to do, and we only feed it the data to process.
How to configure GPTs in ChatGPT
There are two configuration methods:
1. Advanced mode
In this mode we manually fill in all the fields in the template — suitable for users who know exactly what they want to get from that GPT.
2. Assisted mode
Here we write directly in the chat what we want to achieve, and the AI fills in the required fields for us. Along the way, the assistant asks questions to better understand the purpose and role of the GPT.
This assisted mode is, in a way, the equivalent of asking it to generate a personalized prompt.
After configuration, the GPT appears in the left sidebar. We can open it directly from there or call it from a normal conversation using @.
When we use @, a list appears with all GPTs created in the My GPTs section, from which we can select it.
Practical example: translating blog articles
A few weeks ago, we added an English section to the blog, where we translate the articles from the Romanian section. For this process I created a GPT specialized in Romanian-to-English translations, Gadgetisimo Translator, which I used to quickly produce the translations.
Such a GPT can be:
- kept private,
- shared with other people, who can use it in their own accounts while keeping their conversations private,
- or published in the public ChatGPT library, just as we can also use GPTs created by others.
On the blog we used Gadgetisimo Translator to translate each article into English after finishing the Romanian version. I said used because in the meantime we further automated the process, so the translation is now done at the press of a button — still with the help of AI.
