Excel with AI: how to maintain reliable results?

Microsoft is withdrawing a Copilot function from Excel before its wide launch. For Certiport candidates, the message is clear: mastering the basics remains crucial. Imagine that the student services of a university want to automatically classify 2,000 student comments in Excel. Artificial intelligence (AI) promises to identify themes, summarize complaints, and produce a clean table in minutes. Good idea, as long as you don’t...

The Editorial of the UNIVGA Review 24 August 2026

Imagine that the student services of a university want to automatically classify 2,000 student comments in Excel. Artificial intelligence (AI) promises to identify themes, summarize complaints, and produce a clean table in minutes.

Good idea, as long as you don't confuse speed with reliability. Microsoft announces the removal of the =COPILOT() function in Excel starting September 14, 2026. For students preparing for Certiport certifications, particularly Microsoft Office Specialist (MOS), this news serves as a simple reminder: a good tool does not replace a method.

Rio Tinto accelerates five mining startups, the proof remains to be seen

1 October 2026

Offshore: Oxy and Valaris awarded, HSE AI must remain human

24 September 2026

MOL solar and battery branch, the cost remains to be proven

What is it, concretely?

The =COPILOT() function was a tested formula in Excel to ask AI to produce a response directly in a cell. Instead of writing only a classic formula like SUM() or a lookup in a table, the user could formulate a request in plain language, such as summarizing reviews or classifying comments. Microsoft indicates that this function, linked to preview programs, will be removed, but Copilot will remain available in Excel via the sidebar. The key point is trust: AI assistance can speed up analysis, but it can also make mistakes. Rest assured, this does not make Excel less useful; it makes your skills even more important.

Concrete case: what to do and what not to do

In Libreville, the student services of Omar Bongo University receive a satisfaction questionnaire filled out by 2,000 students. The responses cover courses, internships, guidance, payments, and administrative reception. The staff wants to quickly produce a report for management.

The temptation is strong: to use an AI function in Excel to classify each comment by theme, then generate graphs. The problem is not the idea, but what is lacking around it: a privacy rule, a control sample, human validation, and true mastery of Excel. Without this, a quick summary can become a fragile report.

What can be done

Clean the file before any analysis: clear columns, duplicates removed, unnecessary data deleted.
Hide personal information when it is not necessary for the analysis.
Test the AI on a small sample before processing the entire file.
Compare multiple AI responses with a classification done by a person.
Use classic formulas for exact calculations: totals, averages, filters, percentages.

What not to do

Treat an AI response as an automatic truth for an administrative decision.
Copy names, grades, or sensitive information without a clear protection framework.
Replace human verification with a summary produced in seconds.
Sign an official report based solely on an uncontrolled classification.
Abandon learning Excel on the pretext that an assistant can help.

Questions to Ask Before Acting

Do the file's data contain names, grades, salaries, health information, or client data?
Should the expected result be exactly 100%, or should it only provide a trend to explore?
Who will control a sample of the responses produced by the AI before dissemination?
Do important calculations rely on reliable Excel formulas rather than a generated response?
Has a responsible person validated the final file and the conclusions of the report?
Can a colleague understand the method used by opening the file in six months?
Are users trained in both Excel functions and the limitations of AI tools?

UNIVGA Viewpoint

For Gabonese students and French-speaking Africans preparing for Certiport certifications, this news is very concrete. AI features are evolving rapidly, sometimes even before they are widely launched. The skills that last are more stable: structuring a sheet, choosing the right formula, building a readable table, verifying a result, and explaining one's method to a supervisor.

In a university, a small and medium-sized enterprise (SME), a bank, or a hospital in Gabon, Excel often contains sensitive data. Gabonese law n°025/2023, relating to the protection of personal data, reminds us that this information cannot be handled lightly. Therefore, before using AI assistance, rules must be established: what data is used, who accesses it, who validates it, what is kept as a record?

UNIVGA's position is clear: training in AI, yes, but on a solid foundation. Microsoft Office Specialist, Adobe, and IT Specialist certifications are not just lines on a CV. They provide useful professional reflexes for working with modern tools without losing control over decisions, numbers, and documents.

Sources

To go further

17 September 2026

Sustainable wood: AFi promotes traceability towards proof

AFi publishes guidance on the traceability of wood supply chains. For African exporters, the certificate is no longer sufficient without verifiable data.

Read the article

10 September 2026

The United Kingdom warns about hidden AI, audit risk

The British NCSC warns about undeclared AI in companies. A key issue for auditing, compliance, data, and governance.

Read the article

4 September 2026

Copyright: Washington supports OpenAI, open risk

Washington supports OpenAI against the New York Times on AI training. For African lawyers, liability remains to be organized.

Read the article

Excel with AI: how to maintain reliable results?

Rio Tinto accelerates five mining startups, the proof remains to be seen

Offshore: Oxy and Valaris awarded, HSE AI must remain human

MOL solar and battery branch, the cost remains to be proven

What is it, concretely?

Concrete case: what to do and what not to do

Questions to Ask Before Acting

UNIVGA Viewpoint

Sources

Sustainable wood: AFi promotes traceability towards proof

The United Kingdom warns about hidden AI, audit risk

Copyright: Washington supports OpenAI, open risk

Cookies and Privacy

Set cookies

Professional Certificate in Mastering Diplomatic Protocol and Communication

Shopping cart

Rio Tinto accelerates five mining startups, the proof remains to be seen

Offshore: Oxy and Valaris awarded, HSE AI must remain human

MOL solar and battery branch, the cost remains to be proven

What is it, concretely?

Concrete case: what to do and what not to do

Questions to Ask Before Acting

UNIVGA Viewpoint

Sources

Related Articles.

Sustainable wood: AFi promotes traceability towards proof

The United Kingdom warns about hidden AI, audit risk

Copyright: Washington supports OpenAI, open risk

Shopping cart