The problem is that we are not measuring it well

Artificial intelligence is transforming the work world and business at a speed never seen before. However, this technological advance also raises an important challenge: to measure the real impact of the In productivity. Traditional metrics do not correctly capture the contribution of AI since they only focus on specific process factors, not what economists call Total factors productivity.

In other words, it is not enough that only improve a certain link In the production chain, but the improvement must be given throughout the process, obtaining a greater value of all the resources used in it. Therefore, one of the pending subjects will be Develop adequate tools To measure the productivity that AI contributes in those new parameters.

AI is only a tool. Automation promoted by AI has already begun to make differences in sectors such as financial, technological and manufacturing. These AI tools allow Automize repetitive tasks and release time for activities that require human skills such as strategic decision making. However, that change of time use It is not reflected in productivity because The quality of work is not recorded or the long -term improvement of the company’s strategy.

A study Harvard Business School revealed that generative AI can increase the speed of execution by 25% and individual yield by 43%. However, its impact is not uniform: while some sectors obtain immediate benefits, others do so in the long term.

The “J” of productivity. Historically, technological innovations have followed a pattern known as “Productivity J curve“. Erik Brynjolfsson, economist and teacher of Stanford, explains that this model describes how companies that adopt new technologies in their processes initially experience a Reduction in its productivitythen bounce and rebound it. Some examples were seen with the adoption of the steam machine, electricity or the use of computers, whose positive effects took decades to consolidate.

The generative AI does not escape this pattern. Although some companies may experience growth in their productivity, many face an initial fall due to the need to adapt their internal processes and train their employees, making this increase in productivity applaud over time.

Default AI is a mistake. A proof that AI alone does not represent an improvement in productivity is The studywhich carried out the University of Pennsylvania and Harvard Business School. The study discovered that the use of Chatgpt had an immediate impact on the resolution of certain tasks, while others were resolved in less time when a human without AI intervention was carried out.

Therefore, the productive growth paradigm only because of the fact of impleting AI has a huge asterisk. This growth is conditioned by the type of industry and economic activity that companies develop. The implementation of AI in the field of manufacturing or the primary sector requires greater investment and adaptation time, than for other companies in the services or financial sector.

Time is the best productivity indicator. A Published article in Bloomberg On this subject made a flag of the popular saying “time is gold”. It pointed out that AI automation can be a “throwing weapon” that improves the productivity of some sectors, but sinking that of others. Its author set the Autophagus kiosks They already have some fast food establishments.

With them, the company saves the salary of some ATMs receiving orders and charging them, which increases the productivity of the company. However, this solution makes use of the unpaid time of customers. What if those customers were other companies? Its productivity would be affected since its employees must use their time by making “your orders” and loading with the cost of automation that is saving its supplier.

The new gold mine is not doing more, but less. This approach puts on the table the role of AI in the industry to take advantage of its ability to Automate processes. These processes consume a lot of time on the working hours of workers (such as Amazon already pointed out), so the challenge is to automate them so that workers use that time to improve their products, such and As I pointed out Jensen Huang.

Rather Measure productivity in terms of how much does an employeeperhaps new metrics should be taken that collect factors such as What do you do your working time and How are they improving The product. This is a great change with respect to industrial metrics based on the amount of product produced per hour.

In Xataka | “I have three years of work”: more and more IA managers believe that AI will end up removing the position

Image | Unspash (Remy Gieing)