Most federal government bodies in Brazil fail to use AI efficiently

The majority of federal government organizations in Brazil are still in the very early stages of using artificial intelligence (AI), according to a study carried out by the Federal Court of Accounts (TCU).

The aim of the TCU research was to understand the use of AI by the federal administration, identify the associated risks, and inform the evaluation of the Brazilian Artificial Intelligence Strategy.

Using a scale ranging from 1 to 4, the study has found that 38% of the federal government bodies in Brazil are at Level 0 of maturity in terms of AI use. This means they are not using the technology. Some 33.5% of the organizations are at Level 1, meaning internal discussions around AI have started but are still on a speculative basis.

“In those cases, it was found that the simple lack of knowledge regarding the opportunities and benefits that can be achieved with the use of these technologies was an impediment, which suggests a gap that runs through the limitations in the workforce and reaches the leaders of these organizations,” the report noted.

The TCU research also found that 28.5% of all federal government bodies are at Levels 2, 3, or 4 in terms of AI maturity. Most of the institutions within this group (17.1%) are carrying out pilots or more elaborate proofs of concept. AI projects are in production in around 8% of the organizations, while the most advanced initiatives are in place at 3.4% of the Brazilian federal government bodies.

Criticisms to the national AI plan

The TCU report also contained a number of criticisms in relation to the national AI plan. Published in April 2021, the strategy was intended to guide actions around research, innovation and the development of AI and related technologies to tackle the country’s greatest challenges, as well as ethics.

The report found that the plan does not have a set of specific, measurable and achievable objectives, to be reached within a set period of time. “In practice [the plan] can be considered as guidance only, which are more open by nature,” the TCU report pointed out.

Moreover, the report noted the strategy lacks an initial reference stage, and that the indicators and targets for performance evaluation are not clear. “Without baseline values about the reality to be improved, it will not be possible to develop any relevant analysis of the results achieved,” the TCU report noted, adding the strategy only presents general data on the use of AI in Brazil, without defining what is intended to be measured or prioritized.

In addition, the report noted there are flaws in the logical model set out in the national AI plan and how the plan impacts the identified problems or prioritized opportunities, since there is no discrimination of inputs, activities, products, impacts, and results.

The publication of the AI strategy was preceded by a process of over a year since the launch of the consultation to gather input for the plan in late 2019, after a period of engagement with AI consulting firms and an international benchmarking process. According to the Brazilian government, the consultation lasted until March 2020 and more than 1,000 contributions were received.

According to then minister of Science, Technology and Innovation, Marcos Pontes, the publication was “the fulfillment of a dream” and a big step forward for Brazil. Pontes said the Brazilian government considers AI to be “essential” for the development of many other technologies, such as innovations around the internet of things approach.