PTArtificial intelligence is already part of running the school
Spreadsheets, e-mails, price quotes, lesson plans: at a private school in São Paulo, Brazil, the management team now does much of this with the help of AI. The dissertation interviewed ten people on that team to understand what changed in their processes, what still holds things back and who decides in the end. Explained in 12 screens.
A 23-day task now takes 5 minutes, according to the person who does it
At the school studied, sending a mail merge (personalized e-mails to about 300 students) used to be manual work. One person on the IT team built an automation with AI. How long do you think it takes today?
Quotes from section 4.2 (theme 3). These are the interviewees' accounts, not measurements by the study. Bars on a logarithmic scale (each mark is 10 times the previous one), because 23 days and 5 minutes would not fit on an ordinary ruler. For the other accounts, the dissertation does not give the time afterwards.
The dissertation sets out three sizes of change
Not every use of AI changes a process in the same way. Sometimes the work stays the same, only faster; sometimes the way of doing it changes; and sometimes it may even change who does what. Tap each step.
Table 10 and section 2.3.2. The author draws on the Oslo Manual (OECD) for the first level and on Vial (2019) for the second; the third "should be treated as a potential", because the change may not yet be consolidated. Examples: Table 11.
Ten people from a school's management told how they use AI
It is a case study: instead of hearing from many schools, the research looked in depth at just one, a private K-12 school in the city of São Paulo that was already using AI. The author chose people for their experience with the subject, and each answered open questions about adoption, processes, rules, tools and decisions.
Table 6 and section 3. All hold a college degree; 60% are in management roles. The names of the school and of the people do not appear in the dissertation; we show only the groups.
The conversations yielded 163 excerpts and six themes
The author read the transcripts several times and marked, in the MAXQDA software, the excerpts related to the research question. She then grouped the excerpts into subthemes and the subthemes into themes. Tap each layer.
Sections 3 and 4.2: thematic analysis by Braun and Clarke (2006), in six phases. 163 excerpts: total of the code matrix (Figure 5). Subthemes and themes: Table 7. Link to the objectives: Table 14.
Usage rules and difficulties were the most discussed topics
Each row is a topic and each column a person; the height is the number of interview excerpts marked with that topic. Choose a topic.
Figure 5 (MAXQDA code matrix). We show the nine subthemes of themes 1 to 5; the decision ones are on screen 10, based on Table 8. More excerpts mean the topic came up more in conversation, not that it matters more: the author herself says frequency did not decide how relevant the themes were.
The school's Copilot sits alongside ChatGPT and Gemini
Use grew gradually between 2023 and 2024: ChatGPT got the conversation started, and the arrival of Microsoft's Copilot, officially adopted by the school, brought AI into everyday work. Place your bet: how many of the ten people talked about Copilot?
Bars: Figure 5 (excerpts, and people with an excerpt, in each subtheme). Grid: put together by us from the quotes cited in theme 5 and section 4.2; a blank cell only means the quoted passage does not mention the tool. According to Table 13, Copilot is preferred for data security and for being part of the Microsoft 365 the school already uses.
Of twenty uses mapped, fourteen do the same work faster
The author mapped 20 uses of AI, half in administration and half in teaching. Tap one and say what you think: did the work stay the same, only faster, or did the way of doing it change?
Table 11 (the author's classification). Counts are ours. Group numbers and names (P1 to P10) follow the table. Where the text of section 4.5 lists other people for text review and teaching materials, we follow the table.
The main challenges lie in culture, training and ethics
Difficulties were, along with usage rules, the topic with the most excerpts: 28, with quotes from all ten. The author grouped them into three types. Tap each card to read what they said.
Table 9 and quotes from theme 2 (section 4.2). Bars: the "Challenges in adoption" row of Figure 5.
AI helps decide, but the decision stays with the person
Managers use AI in three ways when making decisions. Choose one.
As the interviewees describe it, the person deciding is the one who tells the AI what matters. Choose the criterion:
Our illustration · made-up figuresTable 8 and quotes from theme 6. For the author, AI eases the time and information limits of the person deciding (Herbert Simon's "bounded rationality"), but the final choice remains human, anchored in experience (Mintzberg).
The school protects its data with three layers of rules
A school holds data on children, families and staff. Since the pandemic, a digital governance committee decides what is and is not allowed. Tap each layer.
Table 12 and quotes from theme 4. The numbers of people are those the table lists for each area. Brazil's General Data Protection Law (LGPD, 2018) is the basis of the rules.
The path the author proposes for other schools
The dissertation closes with a three-phase model (Figure 6):
1 · Before: conditions
The team adopts what it sees as useful and easy to use, learns alongside the technology (including how to write good prompts) and works in secure environments, within the data protection law and with an ethics committee.
2 · During: new capabilities
From built-in tools (Microsoft 365, Google Workspace) to agents that carry out tasks on their own. In administration: costs, school assets, data auditing. In teaching: digital literacy, inclusive lesson planning and communication.
3 · After: better decisions
"Hybrid" decision-making: AI processes the data and acts as a sounding board, and the person curates and decides.
What this research does not say
- Just one school. A private school with its own technology structure does not represent Brazil's schools, and lack of resources may leave public schools behind. · according to the author
- Ten people, all from management and technology. Other groups might see AI differently; teachers of other subjects, families and students were not heard. · author and our note
- A snapshot of 2026. AI changes fast; the results hold for this moment. · according to the author
- Time reported, not measured. The time savings come from the quotes; the study did not time the tasks. · our note
Suggested next steps: compare public and private schools, study managers' workload and mental health (some accounts mention a "false sense of being on top of everything"), follow the arrival of agents and discuss authorship and plagiarism in documents made with AI.
Credits
GARCIA, Tatiana de Souza Mendes. O uso da inteligência artificial na inovação de processos de gestão educacional. Master's dissertation (Business Administration), Mackenzie Presbyterian University, São Paulo, 2026. Advisor: Prof. Dr. Gilberto Perez.
Qualitative case study, semi-structured interviews and thematic analysis with MAXQDA. Research supported by CNPq (Brazil's National Council for Scientific and Technological Development). According to the author, Copilot and ChatGPT helped only with proofreading, with no access to the interviews. This version simplifies the work for a general audience.