Creating GPT assistants: more analytical capacity and innovation than technology

The challenge: becoming an AI-empowered company

When we started talking about AI-based assistants in our team, the conversation developed around technical concepts: programming, algorithms, the technological complexity of the project, …..

But the reality is very different. Today, the creation of GPT assistants is not a technical challenge, it is a question of understanding, organisation and mental structure – in short, skills are more important than technical knowledge.

A few weeks ago we launched an internal project to improve the efficiency of our processes.

Within the framework of the project, we have worked on the creation of GPT assistants to support us in different areas of the company. With this initiative we pursued the following objectives.

Create a space for our team to resolve all doubts about their day-to-day work in the company. This will reduce the number of queries received by the People team and speed up the response to users, which is now 24×7. Our focus is on efficiency and improving the experience of our employees.
To support consultants in designing proposals aligned with the company’s standards. The objective is to train them in the process, and help them to be more agile in the design of proposals.
Generate a channel of experts on Customer Experience, so that the team can consult all our know-how on the practice and deepen our work methodologies.

The design process of the assistants has been interesting, the first days we went blindly, and then we began to generate learnings in each team, which we have been sharing and enriching the process. These are the main lessons we have learned from the process:

Learning 1. You don’t need to be a technology expert. The barrier to entry is low and the technology is available to everyone.

The first conclusion we came to was that technology was not going to be a problem, as we could communicate with the assistants using natural language. This means that anyone can interact with them without writing a single line of code. If you know how to communicate effectively, you already have the fundamental ability to create your own assistant.

By following simple instructions, anyone can design a wizard tailored to their needs. The process is quite intuitive, and if you need help, there are thousands of manuals and video tutorials on the web that you can consult.

Learning 2. You have to be very clear about what you want to achieve, what your goal is.

This applies to almost anything, think before you act.

In our case, we started to configure the wizards on the user stories, but without being completely clear about how we wanted to achieve, or how the tool worked. This meant that we had to go back several times and start again from scratch, the process was frustrating, but also generated a lot of learning.

So before you start, you need to understand what the user needs and what kind of responses you need to get the wizard to generate. This way you will know if one wizard will be enough for the defined function or if you have to design more than one.

Learning 3. Documentation is the key.

No assistant in the world is going to give good answers if the documentation you train them with is not well developed and structured.

For us, this phase took much longer than the actual configuration of the wizard.

The key here was the ability to organise documentation: collecting, structuring and presenting documentation in an accessible way makes the GPT work efficiently.

And, in addition, we need to be able to synthesise, less is more. It is important to reduce the information, but without losing the meaning. In general, we tend to give them tons of documents with too much information, and that makes their behaviour unpredictable, and far from our expectations.

Conclusion: creating virtual assistants is a skills challenge

Ultimately, creating GPT assistants is not a technological challenge, it is an exercise in analysis, organisation, synthesis, innovation and communication. Anyone with these skills can build an effective assistant without advanced knowledge of AI or programming. Artificial intelligence is not reserved for technology experts, it is within the reach of those who know how to structure and transmit information effectively.

So, if you are thinking of creating your own assistant, I recommend that you follow these tips, and once you have these elements clear, you can configure it, you will see that it is not complicated.

Would you like to give it a try?

To help HR professionals take their first steps in this exciting technology, we invite you to participate in our bootcamp: