Artificial intelligence at work: who's really in charge?

Artificial intelligence agents are entering planning and purchasing. For managers, the challenge is to decide quickly without losing control. Imagine you are running a small distribution business in Libreville. An artificial intelligence (AI) tool alerts you: in three weeks, a highly demanded product is likely to run out. Should you order more, raise the price, look for another supplier, or accept a...

July 1, 2026

Imagine you run a small distribution company in Libreville. An artificial intelligence (AI) tool alerts you: in three weeks, a high-demand product is at risk of running out. Should you order more, increase the price, look for another supplier, or accept a temporary stockout? The answer isn't just about technology.

Another scene: in a hospital, the procurement team receives an automated proposal to replace a soon-to-be unavailable medication. The idea seems useful. But who approves it: the pharmacy, the financial department, senior management, or compliance? That's the real issue for managers.

What is it, concretely?

An agent AI is an artificial intelligence system that follows an objective and chains multiple steps to help a team act. Instead of just providing an answer, it can analyze sales, inventory, supplier lead times, margins, and risks, then propose a plan. For a manager, it's comparable to a control assistant: it shows several options with their possible consequences. For example, ordering more reduces the risk of stockouts but ties up cash. The main challenge, therefore, is not to replace the decision-maker but to organize clear rules. With human validation, reliable data, and written logs, the tool becomes a support, not an autopilot.

Concrete case: what to do and what not to do

Let's take a hypothetical case study: Libreville Medical Services, a small and medium-sized enterprise (SME) in Gabon that distributes medical equipment to private clinics and pharmacies. It sells gloves, syringes, blood pressure monitors, and small equipment. Its problem is concrete: stockouts are recurring, import lead times often change, and the sales department is losing sales.

The temptation is strong to install an AI agent that monitors sales, inventory, and suppliers, then automatically proposes orders. The idea can improve responsiveness. But rules are still missing: who approves, what data is allowed, how are errors tracked, and who is responsible if a recommendation is expensive for the company?

What can be done

Limit the tool to recommendations: it prepares an order, but does not send it alone.
Name a business manager according to the subject: purchasing, finance, sales, or general management.
Set simple thresholds, for example, purchase validation at 500,000 FCFA and management for amounts over 5 million FCFA.
Keep track of each recommendation: data used, proposed scenario, person who validated.
Train teams to question the tool, especially when the numbers seem inconsistent.

What not to do

Let AI automatically command without written procedure.
Connect all company data without sorting sensitive information.
Confusing speed with quality: a quick response may rely on incomplete data.
Remove the manager's role, even though they remain responsible for the final decision.
Measure only the time saved, without tracking avoided downtime, errors, and cash flow.

Questions to Ask Before Acting

Which decision do we want to improve: inventory, pricing, recruitment, scheduling, budget, or customer relations?
Should AI recommend, prepare, or execute an action?
Who validates the final decision if it has a financial, human, or legal impact?
Are the data used reliable, recent, and authorized?
Can we simply explain why the tool proposes this decision?
What do we do in case of an error: correction, alert, suspension, or audit?
Are teams able to challenge a recommendation instead of accepting it reflexively?

UNIVGA Viewpoint

For Gabon and Francophone Africa, the issue goes beyond purchasing software. UNESCO describes an AI adoption that is still emerging in Gabon, with progress in the private and quasipublic sectors. This is precisely the time to train managers capable of governing these tools, before their use becomes widespread without a method. Reading a recommendation, asking for proof, arbitrating between cost, timeline, risk, and human impact: these skills are becoming central to management.

Data sovereignty must also become a management reflex. A bank, university, SME, or hospital cannot transmit client, patient, student, or supplier data to any service. Hosting, transfers, confidentiality, and individual rights must be verified. In Gabon, the National Commission for the Protection of Personal Data (CNPDCP) and law no. 001-2011 provide a basis to be aware of.

UNIVGA's position is simple: no AI without human responsibility. A good leader does not delegate blindly. They organize validation, document choices, train teams, and are willing to slow down a decision when risk demands it. This is how AI can support business performance without undermining trust.

Sources

To go further

21 August 2026

Can business software steal your sensitive data?

A flaw in a business software can grant access to plans, contracts, or client data. Here’s how to respond without limiting yourself…

Read the article

19 August 2026

Can AI enhance the reliability of accident investigations?

An AI can identify weaknesses in an accident report, but the decision remains human. Here is a simple method for safety at…

Read the article

17 August 2026

Digital cadastre: how to avoid land errors?

A Colorado example sheds light on land digitization: registering markers online can reduce disputes, if the evidence remains…

Read the article

Artificial intelligence at work: who's really in charge?

AI at work: how to succeed without disorganizing?

AI in business: who really keeps your data?

Excel with AI: how to maintain reliable results?

Professional Certificate in Governance and CSR — CSR Director

What is it, concretely?

Concrete case: what to do and what not to do

Questions to Ask Before Acting

UNIVGA Viewpoint

Sources

Can business software steal your sensitive data?

Can AI enhance the reliability of accident investigations?

Digital cadastre: how to avoid land errors?

Cookies and Privacy

Set cookies

Professional Certificate in Operational Cybersecurity

Shopping cart

AI at work: how to succeed without disorganizing?

AI in business: who really keeps your data?

Excel with AI: how to maintain reliable results?

Professional Certificate in Governance and CSR — CSR Director

What is it, concretely?

Concrete case: what to do and what not to do

Questions to Ask Before Acting

UNIVGA Viewpoint

Sources

Related Articles.

Can business software steal your sensitive data?

Can AI enhance the reliability of accident investigations?

Digital cadastre: how to avoid land errors?

Shopping cart