Alice's methodology and structure optimise and scale preventive analysis and continuous auditing efforts. The use of Artificial Intelligence and robotic automation (RPA) generates inputs for timely interventions in public procurement. Data from various sources is collected automatically: Compras.gov.br (mandatory use in the Federal Government), Licitações-e (Banco do Brasil's procurement system) and Diário Oficial da União (DOU). Data analysis techniques consolidate around 40 risk typologies. Personalised alerts are sent to more than 500 users of the CGU, other Executive bodies, the Judiciary, the Public Prosecutor's Office and the State and Municipal Comptrollers' Offices. As well as optimising work, prioritising audits that represent greater financial sums and risks, it expands the capillarity of actors involved in improving the efficiency of spending, preventing errors and fraud in acquisitions for the operation of public bodies, policies and services.

In 2016, Alice was shared with the Federal Court of Auditors (TCU), which began to develop and use it independently, making evolutions that resulted in improvements. In 2018, the CGU incorporated the TCU's developments and resumed improving it. In 2021, it partnered with the Ministry of Management to integrate Alice into Compras.gov.br and issue alerts to managers during procurement planning. Judiciary and Public Prosecutor's Office use Alice to stop identified irregularities

Alice assists public managers in the early identification of inconsistencies, supports internal auditors in the detection of risks, and benefits Control Bodies in the prevention and containment of irregularities. It benefits Brazilian citizens, seeking the correct allocation of public funds. Alice generates alerts for more than 500 users who are crucial in the preventive analysis of purchases: CGU auditors and auditors from the Executive and Judiciary branches, the Public Prosecutor's Office and state/municipal comptrollers.

More than R$9.7 billion in bids were suspended or cancelled between 2019 and 2022 due to Alice alerts. Also, according to CGU's calculation methodology, more than R$1.3 billion were accounted for as financial benefits in the period from 2019 to 2022. These measurements were made by consolidating records from the Alice and e-aud systems. The results and impacts observed so far contribute to the compliance of tenders and the proper use of public resources. For the future, we expect to continue identifying irregular practices, expanding and modernising the audit trails, and increasingly contributing to improving public management and preventing significant financial losses.

We faced several challenges. Initially, there were restrictions on accessing data from Compras.gov.br due to the captcha mechanism. This was initially solved by hiring human services, and then with an automatic solution. Restrictions on the use of software tools were circumvented by auditors using their own computers until they obtained the necessary authorisations. The shortage of NLP tools was overcome by developing our own algorithm. It was necessary to opt for economical alternatives: in this way, the CGU's own auditors implemented the tool, and all the technological infrastructure already available was utilised. The partnership with MGI also ensured access to essential data.

Crucial elements for Alice's success: 1) Integration into Routines: The incorporation of Alice into e-aud streamlined the sending of alerts to auditors, integrating it into CGU's routines; 2) Speed and Preventive Nature: The speed and preventive nature of the analyses increased the auditors' perception of Alice's value, encouraging its use; 3) Interministerial Collaboration: Collaboration with the MGI expanded access to accurate and reliable data; 4) Qualified Technical Staff: The existence of qualified technical staff with technological expertise and in government procurement risks was crucial; 5) Senior Management Support: The support of CGU's senior management boosted the consolidation of Alice as a solution with an impact on public management.

Shortly after being launched at CGU, Alice was shared with the Federal Court of Auditors (TCU), which began to develop and use it independently. As it grew in size, other bodies from other branches of the Brazilian state also became interested in the tool and in receiving its alerts, so that units of the judiciary and the Public Prosecutor's Office were registered with the solution. Finally, when it was realised that Alice was collecting and analysing data on purchases by states and municipalities, but was not handled by CGU (it is not its responsibility), a strategic project was also set up to make it possible for these spheres to receive reports and thus contribute to the public sector at a national level.

Alice's career at CGU offers valuable lessons. Flexibility and innovation were essential in overcoming technical and bureaucratic obstacles. Strategic partnerships have revealed the value of sharing knowledge for collective progress. The incorporation of the solution into CGU routines emphasised the importance of efficient integrations. The qualified and motivated technical team was a major differentiator. The support of senior management showed that committed leadership drives innovation. The replicability of the solution to different spheres highlighted its transformative potential. Finally, the experience reinforced that collaborative approaches, adaptability and strategic vision drive innovative practices in the public sector.

It is worth highlighting the initial efforts of the team that conceived and implemented Alice, as they faced technological limitations and limitations in their knowledge of data science and artificial intelligence, at a time when these techniques were not very widespread. Even with a precarious environment, they demonstrated creativity and resilience to achieve the noble goal of delivering public policies to society with integrity. This initial journey shaped innovation and its positive impact.