There would seem to be a first, silent victim in the generative artificial intelligence revolution. Where the algorithm takes hold, the entry of young people into the world of work is interrupted. The extent of the fracture is measured by a study by Harvard University which monitored the trajectories of 62 million employed people. The verdict is conclusive: the digitalisation of basic processes has caused a drop of between 9 and 10 percent in the employment of entry-level profiles. A dynamic that risks sterilizing field training, replacing junior tasks with the efficiency of algorithms. The employment of seniors, on the other hand, would remain stable.
The problem, however, is not just about numbers. If many basic tasks are automated, the way tomorrow’s professionals will be trained will also change. And many skills risk being lost.
This is why understanding how to use AI in the company comes before choosing which solution to adopt. Datlas Group, an Italian SME that develops artificial intelligence solutions for banks, insurance companies and public administration, has sought a response in the field to this delicate topic. It organized an internal challenge with mixed teams of juniors and seniors, with no rules on the use of AI. It emerged that there are those who use it passively to have their work done and those who use it virtuously to be accompanied: to understand the problem, verify their reasoning, delve deeper. Francesco Paolo SchiavoneHead of Innovation & Development of Datlas Group, explains to us that the right path is the second one. A way to use AI that allows people to reskill and learn faster, not stop doing so.
Who is Datlas and why does the topic concern her closely
Datlas operates in the outsourcing of business processes, focused on the development of solutions for the management and valorisation of data.
Customers entrust it with entire chains of activities: verification, control and management of incoming data and documents (such as contracts, loan requests, reimbursement procedures, the onboarding of a university student). A sector once dominated by printing, today driven by digitalisation. Artificial intelligence, here, is nothing new: the Group has been using it for almost ten years to extract information from documents. It is precisely this experience that posed a complex underlying question. For Schiavonein fact, «automating (in a negative sense) means delegating the knowledge of those processes to AI and automatically losing it in the company».
How to use artificial intelligence in companies
The method is clear: try to establish in advance the role of people and how to be helped (and not absorbed) by automation. Relying on the idea (too often taken for granted) that “after AI people will do much more strategic and more impactful things”, without knowing which ones, inevitably leads to a dangerous and difficult to control drift.
For this reason, Datlas has experimented with man-machine coexistence also in customer support. The operators who were initially engaged in verifying tax codes, in this case, were able to acquire skills on insurance practices thanks to the support of AI. In fact, the system limited itself to checking the documents and providing suggestions, explaining why a case could be liquidated or not, while the final decision remained with the operator. In this way, AI has not replaced human work, but has helped operators develop new skills and grow professionally.
What emerged from the internal challenge within the company
The experiment was created to update internal company rules on the use of AI. Teams of two or three people, coming from different areas and seniority, worked on real and daily tasks for the company with full freedom of digital tools. Halfway through the challenge, some juniors stopped to reflect on this automation of their work. The reason, he says Schiavonethey explained it like this: «I’m going very fast on the task, but I’m not understanding how I’m doing some of the complicated things I do, even if I do them well». From there, they stopped having the code generated by the AI, and asked it for explanations, step by step, to understand how to move autonomously.
A winning behavior that Datlas wants to extend to all employees, exploiting artificial help without losing the fundamental contribution of human knowledge. The company has thus created a version of its multi-agent tool reserved for juniors, in which the AI evaluates and explains how to work, and does not carry out the work directly. Beyond the specific area of programming, “I don’t want to automate the decision, but I want to automate the data processing and the preparation of the scenarios that lead me to make a decision”, summarizes Schiavone.
Because generational comparison is still important
One of the surprises came from the more experienced developers. In programming, they were the ones who delegated the operational parts more to artificial intelligence, with an approach that can be summarized in one sentence: “You write the code to me, I’ll check it in review and send it into production.” The architectural comparison, however, remained open with the AI but above all with the more junior profiles, and this brought out solutions in the seniors to make work more efficient by breaking their deep-rooted patterns. AI, in this case, has therefore also favored an exchange of skills between generations. A dialogue that, however Schiavoneremains extraordinarily precious.
Buy a custom-made product, not a model
The challenge for companies today is to understand how and which AI tools to use. For less experienced companies, the Head of Innovation & Development of Datlas Group suggests distinguishing between the purchase of a model and that of a product. «It is not the AI model that matters today, but the set of tools that the chosen product makes available to the models and the business». The choice, therefore, must start from the concrete needs of the company and the activities that you want to improve, not from the simple availability of a model. Many companies activate a license only because they already use other software from the same vendor. Without critical review, they risk poor presentations or half-read files. Key performance factors such as conversation memory also depend on the product chosen.
The same principle applies to the organization of work. If for juniors AI represents an “enabler” of the apprenticeship, at the top a more conscious use of the skills already acquired is needed, especially in new settings such as working in hybrid teams. «The individual person must know how to be the manager of the AI team in front of him».
In the Datlas tool, for example, five artificial agents and five people worked together and discussed how to manage some tasks, in a real working group. From this experiment it emerged that in a similar team of cohabitation between man and machine, progress is slower, but the result is significantly better in terms of quality and risk prediction. This is why Datlas is experimenting across the board with this new way of working, training its people on the languages and inputs to work best with internal systems.
Because, remember Schiavone: «Learning to provide the right inputs also means learning which ones not to provide: what you include in a prompt is also the company’s asset. Understanding how to align communication between people and machines is anything but trivial.”
