Every day there is news about Artificial Intelligence, the advantages it represents and its spectacular forecasts. The constant launch of new products by technology companies confirms this. AI is the technology that has been introduced the fastest and most users use it as a tool to help them in their tasks. However, we have the contradiction that very few companies are achieving significant cost savings from AI that they can justify in their income statement.
A recent MIT study indicates that only 7% of companies have managed to translate AI into cost savings, the majority have not gone beyond carrying out pilot projects, with better or worse results. In Spain the results are lower. A majority of users (more than 88%) use AI as a tool, but its use is limited to searching for information, making summaries and comparisons, or helping with presentations. These uses are saving hours of work in companies, but they do not translate into real savings since there is a lack of a concrete plan on how to apply the hours saved, nor do they consider the remaining tasks of the job.
MIT reports also indicate that AI models are sufficiently advanced to be able to replace 12% of the tasks currently performed in companies. Nobody doubts the usefulness and capabilities of AI anymore and most CEOs consider that the viability of the company depends on its adoption.
Why then are companies not realizing the savings that AI represents?
Possibly the first answer is that cutting-edge advances are rapid but changes in the economy and society are slower. Humans need time to assimilate and adopt a new technology en masse, understand what it can mean and accept it generally.
Many of the pilot projects carried out by the companies did not have transformational objectives. In many cases, it was about achieving short-term results and the most spectacular achievements possible that would allow the internal sale of the new technology and raising awareness of its importance. However, they did not have the capacity to scale, nor did they seek to achieve significant savings and, sometimes, they had not even thought about how to measure the savings.
To achieve the results we must move from the experimentation phase to the transformation or reinvention phase. The implementation of AI requires a business project that can generate savings in the income statement. This project requires sustained investments, the clear identification of the areas of the company where AI can be most effective and the decision of senior management to carry out real transformations through interactive cycles that feed back with the results obtained. These processes, like any change management process, require governance, measurement and monitoring of results, specific training in the areas of change and, above all, leadership. You have to know and combine the technological possibilities of AI with the strategy and manage a complex change process.
Technology departments are the fastest to implement AI and achieve measurable results. The main reason is that AI tool providers are applying them internally and technology and software development are their main jobs. In these areas it is no longer about being a leader, but rather about not being left behind in the accelerated transformation and improved productivity that is occurring with the introduction of AI in software development. The change and productivity improvements are greater than those achieved with the already very important DeVOps improvements that have occurred in previous years.
As in any strategic project, it is essential to measure the costs incurred and the savings achieved in each of the costs.. AI projects involve tool licenses and subscriptions that may increase in subsequent years. These possible situations must be considered and an architecture used to reduce dependence on suppliers.
MIT reports indicate that AI models are advanced to be able to replace 12% of the tasks performed in companies
Regardless of whether they are software development tasks, or any other type of tasksit is crucial to be able to measure the result and quality of work. This will allow us to know if we are achieving the expected results and plan the next improvement cycle.. It is also necessary to be able to compare ourselves with similar facilities. It may be the case that we are very satisfied because we are achieving productivity improvements of 20% and we do not see that our competition is achieving improvements of 40% or greater.
A valid alternative is to establish a system that allows productivity to be quantified. If we talk about software development, we must measure the quality of the software objectively. We have to know if the project delivery time has decreased, if the results correspond to the needs and specifications, and verify that the quality of the software has improved and the associated costs have been optimized. These metrics are crucial for making decisions about how to advance software development and being able to define cycles of incremental improvements.
Another fundamental resource is benchmarking, which makes it possible to compare the performance of equipment, technologies and processes in relation to industry standards or with historical results of the organization itself. In this context, Benchmarking provides a clear view of how AI implementation is impacting productivity.
Julián Gómez Bejarano, Chief Digital Officer LedaMC
