This web app uses cookies to compile statistic information of our users visits. By continuing to browse the site you are agreeing to our use of cookies. If you wish you may change your preference or read about cookies

November 28, 2023, Pedro Trillo

ChatGPT is a year old. How did the job change?

The first anniversary of ChatGPT is approaching. Its launch was on November 30, 2022, and some aspects of our work have changed radically, while others have not; in the following article, I intend to analyze precisely the actual impact on the work. And, as the song says, how have we changed?

YouGov conducted a user study that is already several months old. According to these data, since January 2023, 17% of people between the ages of 30 and 44 have used ChatGPT to generate text. The percentage of people between 18 and 29 decreases to 15%. Approximately 9% of people aged 45-64 and 5% of people 65 years of age.

We understand these statistics in their context, in an environment of experimental and individual use, in a recreational type of use, we could call it. These a priori data may seem insufficient about the media noise we know. Still, a phenomenon very similar to the Internet adoption level is happening.

Remember who and why you used the Internet in 1996, 1998, or 2000, and reflect 25 years later on what we do not do in our lives and jobs that are not directly or indirectly related to our use of the Internet, then project in the next 20 years, what will be the level of adoption of generative AI? … how many areas of our lives and work will it impact us?

The fact that state-of-the-art technology is very advanced and there is constant innovation on the market does not mean that the mass adoption of technology by the vast majority of society is going at the same speed. In fact, as with any other technology or significant advancement, it occurs at two speeds: the speed at which new technological developments are released and the speed with which society adopts and integrates them into its day-to-day life.

For those who predicted at the beginning of the year that the startups that we rely on the surface layer of OpenAI’s LLMs (large language models) had the days counted, the prophecy has not been fulfilled. The result, fortunately, is the opposite in most cases, except for the exceptions of startups that have contributed very little value above the OpenAI API layer; in these cases, there has been implacable cannibalization.

But in general, it is a good year of growth for the rest of the startups of generative AI, which will cost us more and more to differentiate, but the market is vast.

If by the end of 2023, only 16% of society is generating text with GPTs recurringly, we must enter our startups and the other 84% in the coming years. There are reasons to be optimistic and many more to be rational with knowledge from within the market in which we live.

However, survey data on actual implementation in the company differ radically. ResumeBuilder surveyed in February 2023 to 1,000 US business leaders to determine how many companies are currently using or intending to use ChatGPT.

Nearly half of the companies surveyed (49%) confirmed they already use ChatGPT, and another 30% plan to do so soon. Thanks to this technology, 25% of companies using it have saved over $75,000. Most ChatGPT users (93%) want to increase their use in the future. Furthermore, 90% of business leaders believe that knowing how to use ChatGPT is a valuable competence for job seekers.

Then, the data radically change according to the context, as I always recommend in my articles; understand these data as indicative, not as absolute truths; the truth will be in the middle. In any case, the most serious and detailed study I have found on the future impact of this technology on work was conducted by OpenAI earlier this year.

It follows that 80% of the U.S. workforce could have at least 10% of their work tasks affected by the introduction of GTCs, while around 19% of workers could see at least 50% of their tasks impacted.

I like this study because it starts from a more realistic basis, understanding that the impact should be evaluated, not roughly by industries or professions, but by the tasks you perform within your occupation.

The nature of modern knowledge-based jobs (or white-neck jobs) is more complex than it seems; it requires task resolution; yes, a percentage of these tasks can already be done with AI, but it also requires coordination, communication, management, problem conceptualization, risk analysis, critical thinking, etc.

Then we have to fold very fine for the moment, and there is no doubt that after the launch that OpenAI has made of specialized agents (GPTs) working in individual silos, there will be a point at which we think about how to make them communicate with each other? And the next step will be to coordinate.

You will not have to do 100 prompts for a particular project. Otherwise, you will send one to agent 1, will divide it into tasks towards agent 2, who will proceed to distribute those tasks between various agents, the 3, 4, etc, and at no time will there be human intervention, then the part of communication and coordination that we do in our work, will also do the AI agents between them.

This concept, I call chain agents, will be the logical step towards where the GenAI industry is going in the short and medium term. The question will be whether the AGI will appear in the way (Artificial General Intelligence). It doesn’t matter if they’re text bots or ultra-realistic AI avatars like Meta’s, the carcass of the outside; it does not matter; the key will be that they communicate and coordinate. Here comes a new giant disruptive step.

Impact of generative AI by task.

The impact matrix below is about analyzing the increase in productivity that has occurred. I like it to be expressed in those terms, in increases in productiveness, not in replacement of workers. The X-axis shows the grouping of everyday tasks in virtually any business sector. In the Y-axis, the various industries appear at the general level.

Generative AI Impact Matrix by task and industry sectors.

Today, according to the data, we can say that marketing and sales are the ones that are taking the most advantage of these advances. Generative AI is causing furor for content creation and sales automation; its application is straightforward and natural, and the results are obtained immediately.

The second group of tasks relates to software engineering, and those who have programmed it are the ones who are taking the most advantage of it for its intensive use in the creation of high technology.

The third group would be customer care, and the fourth would be product and research tasks. The technology sector, banking, and telecommunications are the ones that are beginning to experience the most advantages in the implementation of this technology.

In the customer service sector, it was profiled as the first profession that would disappear. Still, according to my personal experience of years ago, it does not matter if you have an old bot, a modern bot with GPT-4 integrated, who has all the knowledge of your company, or does not have it, the end customer when he decides to contact the company, does not want any automation, wants to talk to a person who solves the problem, wants a human…

On the other hand, Harvard conducted a productivity study on 18 essential consulting tasks and randomly selected 758 consultants from the Boston Consulting Group (BCG), who were assigned access to GPT-4 (the green and red group of the chart) in the case of the red were consultants who had previously received training on how to use GPT-4.

And they were compared to another group that didn’t use Generative Artificial Intelligence. The conclusion is that those consultants who used AI completed 12.2% more tasks on average and 25.1% faster. Quality improved by 40%. Those who took the most advantage of the experiment were middle or low-profile employees, and in high-profile employees, it improved their productivity, but not as much as in the previous ones.

Impact study of generative AI in consultant BCG.

Sectors that generative AI is replacing.

In this section, I reviewed an old blog post from March 2020; The 3rd Disruptive Generation is Already Generative, in this article, I commented that the first professions impacted by generative AI would be the freelancers who work online on platforms like Fiverr; this type of worker offers unitary services of particular tasks, for example, I write you an optimized SEO article for your blog, I create a video reel for Instagram, I record you a voice in off for the video, I make you an excel with research data on the Internet, etc.

As they are unique microtareas on demand, they are solved very easily with generative AI that works in unit silos and with software as a service that integrates this technology. The impact is tremendous in this sector, which has dropped drastically in the number of jobs contracted, and the price has also decreased logically.

Impact study of ChatGPT on online self-employed workers – Source Financial Times.

Whether online, in these marketplaces of freelancers, or offline, my advice for this professional is that you sell the entire project, not sell the unit and individual services within a larger project, which can be done with AI in a matter of minutes.

Those who hire you because they do not know their company probably continue to do so because they are not hiring the supplier itself but your experience, knowledge, and judgment for that project; those who employ you for lack of time in their business even though they know, will go directly to generative AI tools to obtain these services.

In my view, the way to avoid losses is to offer the whole package, which requires coordination of areas of knowledge and communication with several people.

Other professions impacted are editors, screenwriters, translators, administrators, film producers, legal assistants, graphic designers, journalists, etc.

Regarding code programmers, the demand before introducing this technology was very high and continues to be high; studies say they help them in 40% of their programming tasks, but in the other 60%, the knowledge of these professionals remains essential.

In minimally serious and customized projects, which are not no code with prefabricated modules, I do not think there will be a high impact of job loss in this profession in the short term. We will see what happens in the medium and long term because a new generation of compelling tools is being cooked to build the project code directly from the design deliverables and be able to deploy it from the same tool to the development environments if it can be said, that today, the productivity has increased dramatically in this sector.

The 4 phases of entrepreneurial adoption with generative AI.

During the year, I have held punctual conversations with mainly small and medium-sized enterprises, from 20 to 30 employees, asking myself, Pedro, how do I integrate generative AI in my company?… I have identified patterns in all of them, and the first point before integrating this technology into your operations comes by understanding what you did in the previous phase of the one we came from, namely, How was the data culture working in your company? How did you capture that data? What is the state of the data?

The answers have been very disparate; my advice is that you start with the previous step; otherwise, the tragedy is masked as you try to run to implement GenAI in a hurry without having in your company a culture of data implemented or the processes are not well defined.

Without going into details, the implementation project of GenAI in your small business should follow these four fundamental phases:

Exploration–At this point, I take it for granted. Still, you can start with concept tests and familiarization with ChatGPT, create prompts tailored to your business, test several different departments of your company, get some pre-training, etc.

Driving-A a few months ago it was much more complex; a few years ago, it was much more, but with the recent release of OpenAI GPTs, there is no excuse anymore; in a matter of minutes, you can create your specialized GPT for the specific and personalized knowledge of your company, here the critical point will be in the quality of the data you have, more than the quantity. Well used, this bot can make you very productive, reducing task execution times, lowering costs, and increasing quality.

Scaled-At this stage would come the automation of customer support operations; you can organize, separate, and relate all your business documentation to create different GPTs for each case of use of each department. Finally, it starts with automatic content generation for marketing, sales, and products and installs this philosophy in employees.

Innovation-With all the knowledge acquired in the previous stages and with some pilot projects of an internal character, it is time to create products to the outside for your customers. Find your use cases and start testing them with your end customers. First, think about how you are going to monetize. This point is not trivial, so as not to cannibalize your main activity…

Generative AI adoption journey.

Generative AI and the future of work.

Generative AI and the future of work.

On the working model of the future present, we are already experiencing it somehow; at the start, you interact with general knowledge models such as ChatGPT, you access customized bots with proprietary data of your company, you use other software or web apps of AI generative of a specific character in your workflows, and in the future (not far away seems to be) will come General Artificial Intelligence (IAG) or general AI.

Then, it helps you manage and generate information, research, automate, innovate, and make data-based decisions. In this work scenario, the layers of security and ethics become more meaningful the more significant the company is and when the volume of internal data being handled is high.

At the end of your work, you divide the tasks into two groups: the tactical or execution tasks, already done by the AI and the heavy work. As I commented at the beginning of the year article, The meaning of work in transformer times, this new mode of work inevitably pushes you towards high-level work of a strategic character, such as constructing new services and products, opening new markets, and designing new lines. For this task, we have built a specialized application for you in Vizologi based on generative AI so that you can imagine and generate more business for yourself.

Conclusion.

The most crucial variable of modern economies is based on productivity, not the number of people employed, and the concept of human capital within the mute economy, when you have synthetic artificial intelligence close to zero cost and very high levels of automation.

An economist and social activist, Jeremy Rifkin, relates economics to thermodynamics to explain productivity.

He sees economic productivity as a thermodynamic process transforming energy and resources into goods and services. According to Rifkin, the economy is inefficient from a thermodynamic point of view. It proposes that we are in the process of the Third Industrial Revolution, which aims to be much more efficient than the previous ones, driven by new information technologies and renewable energies.

When the cost of intelligence and energy approaches zero, it will be the turning point towards the so-called age of abundance that technicians love so much. But we are not at that point yet, much less at the beginning of the path, where Generative Artificial Intelligence has been the flame that has lit the spring of a combustion process that has been going on for three decades.

The economy is based on how productive the process of creating a good or service is, no matter whether there is a machine behind it or a human; even though there will be significant job losses, the macro economy will theoretically grow a lot by the simple fact that the critical factor, productivity will increase dramatically in the next ten years, then we have to think about how to manage that macro-economic surplus in profits to be able to alleviate the shortage of work that will be.

The risk is real; if this surplus is concentrated only in the four hands of the American big tech if they don’t return that surplus to society in some way, they would also run the risk of collapsing in a massive unemployment scenario, regardless of size, no company is interested in unemployed or a severe crisis, as consumption falls and sales fall.

In 1900, 36% of the American working population worked in agriculture. That figure is 1.2% today, and unemployment is <4%. AI will take away jobs but create new ones; although I am not clear that more new professions will be created than destroyed, I have my serious doubts. Today’s best thing you can do is learn to work closely with AI in your field.

CEO at Vizologi | Website

Pedro Trillo is a tech entrepreneur, telecommunications engineer, founder of the startup Vizologi, specialist in Generative Artificial Intelligence and business strategy, technologist, and author of several essays on technology.

Share:
FacebookTwitterLinkedInPinterest