AI in the Workplace: What Employers Should Put in Place Before Deploying AI Tools
In most African businesses, artificial intelligence did not arrive through a board decision. It arrived through employees. A relationship manager started drafting client emails with a chatbot. A finance analyst began summarising board packs with an AI assistant. A recruiter discovered that a screening tool could shortlist five hundred applications in an afternoon. By the time leadership asked whether the company had an AI strategy, its staff were already using a dozen AI tools, most of them on personal accounts and free tiers.
That pattern, which is common from Lagos to Nairobi to Johannesburg, creates a specific set of legal and commercial risks. The question for employers is no longer whether to allow AI, but how to bring existing use under control and deploy new tools on terms the business can defend. This article sets out what employers should put in place before, or in many cases immediately after, AI tools enter the workplace.
The commercial trigger is usually an incident
Most businesses formalise their approach to workplace AI after something goes wrong, or nearly goes wrong. An employee pastes a client’s confidential term sheet into a public chatbot. A draft contract generated by AI contains a clause that does not exist in Nigerian law. A candidate rejected by an automated screening tool asks why. A client asks, in a tender questionnaire, whether the business uses AI in delivering services and how it manages the risk.
Each of these is manageable, but each is easier to manage if the business already has a framework. The elements below are what that framework should contain.
1. Know what is already being used
The first step is an inventory. Employers should find out which AI tools staff are using, for what purposes, on which accounts, and with what data. This is less about discipline than about visibility. A business cannot manage risk it cannot see.
The inventory should distinguish between enterprise tools procured by the company, which come with negotiated or at least reviewed terms, and consumer tools used on personal accounts, which usually come with terms that permit the provider to retain and use inputs. The second category is where most of the confidentiality and data protection risk sits.
2. Adopt an acceptable use policy
An AI acceptable use policy should be short, practical and specific to the business. It should cover:
- Approved tools: which tools may be used, and on which accounts. Employees should use company-provisioned accounts with appropriate data settings rather than personal accounts.
- Prohibited inputs: confidential client information, personal data, trade secrets, source code and non-public financial information should not be entered into tools that have not been approved for that data.
- Human review: AI output used in client deliverables, regulatory filings, contracts or decisions about people must be reviewed by a qualified person who takes responsibility for it.
- Disclosure: when AI use must be disclosed to clients, regulators or counterparties, and how.
- Prohibited uses: for example, fully automated decisions about hiring, termination or credit without human involvement.
The policy should be incorporated into employment terms or the employee handbook so that breaches can be addressed through the disciplinary process.
3. Protect confidential information and trade secrets
The most immediate risk from uncontrolled AI use is leakage. Information entered into a consumer AI tool may be retained by the provider, reviewed by its staff, or used to train future models, depending on the terms and settings. For a professional services firm, a bank or a technology company, that can amount to a breach of client confidentiality obligations and, in the case of trade secrets, may undermine the business’s ability to claim that it took reasonable steps to keep the information confidential.
Employers should review client contracts for confidentiality and data handling obligations that may restrict AI use, configure enterprise tools so that inputs are not used for training, and train staff on what counts as confidential.
4. Address data protection obligations
When AI tools process personal data, whether about customers, employees or job applicants, data protection law applies in full. In Nigeria, the Nigeria Data Protection Act 2023 requires a lawful basis for processing, compliance with purpose limitation and data minimisation principles, and appropriate security. The General Application and Implementation Directive issued by the Nigeria Data Protection Commission in 2025 reinforces expectations around data protection impact assessments for high-risk processing, which will often include the deployment of new technologies on personal data.
Three points deserve particular attention:
- Automated decision-making: The Nigeria Data Protection Act gives individuals rights in relation to decisions based solely on automated processing that produce legal or similarly significant effects. Similar rights exist under South Africa’s Protection of Personal Information Act and Kenya’s Data Protection Act. Using an AI tool to reject job applicants, flag employees for termination or set pay without meaningful human involvement engages these rights.
- Vendors as processors: An AI vendor processing personal data on the employer’s behalf is a data processor, and the relationship must be documented with appropriate terms on security, sub-processing, cross-border transfers and deletion.
- Cross-border transfers: Most AI tools process data outside Africa. Transfers of personal data out of Nigeria must satisfy the Act’s transfer rules, and the same is true in other African jurisdictions with transfer restrictions.
5. Manage employment law risks
AI deployment raises employment questions that go beyond data protection.
Recruitment and bias: AI screening tools can replicate bias present in their training data, disadvantaging candidates by gender, ethnicity, university, location or other characteristics. Employers remain responsible for discriminatory outcomes even if a vendor’s tool produced them. Before deploying a screening tool, the employer should ask the vendor how the tool has been tested for bias and should retain human review of shortlisting decisions.
Monitoring: AI-enabled productivity monitoring can be intrusive. Employers should be transparent with staff about what is monitored and why, ensure that monitoring is proportionate, and reflect it in employment documentation and privacy notices.
Changes to roles: Where AI changes job content or reduces headcount, employers must follow the applicable procedures. In Nigeria, redundancy affecting workers covered by the Labour Act requires notice to trade unions or workers’ representatives and observance of the statutory principles on selection and payment. Employers in other jurisdictions face comparable consultation requirements. AI-driven restructuring that ignores these procedures invites claims.
Ownership of AI-assisted work: Employment contracts should make clear that work product created with AI assistance in the course of employment belongs to the employer, and that employees must follow the AI policy in producing it.
6. Get the vendor contract right
Enterprise AI tools should be procured with the same care as any other technology that touches sensitive data. The contract should address whether inputs and outputs are used for training, data location, security standards, confidentiality, ownership of outputs, intellectual property indemnities, liability, and exit. We cover AI procurement in more detail in a separate piece, but the central point for employers is that the terms of a consumer AI service are rarely appropriate for business use.
7. Consider sector-specific expectations
Regulated businesses should check whether their regulators have issued guidance or expectations on AI, model risk, outsourcing or technology risk management. Banks and payment service providers in Nigeria operate under Central Bank of Nigeria frameworks on risk management, cybersecurity and outsourcing that will apply to AI tools used in their operations. Capital market operators, insurers and healthcare providers face similar considerations. Businesses that serve European clients or process data about European residents should also be aware that the European Union’s AI Act imposes obligations on certain uses, including some employment-related applications, with extraterritorial reach.
8. Train people, then revisit
A policy that no one has read does not reduce risk. Employers should train staff on the policy, the approved tools and the reasons behind the rules, and should revisit the framework regularly. The tools are changing quickly, and a framework adopted today will need updating within months.
The regulatory direction
Nigeria does not yet have a dedicated AI statute. The government has published a National Artificial Intelligence Strategy, and legislative proposals, including a bill to establish a national body for AI regulation, have been introduced in the National Assembly. Kenya, Rwanda, South Africa and Egypt have each published AI strategies or policy frameworks. For now, the legal framework for workplace AI in most African markets is the existing law on data protection, employment, confidentiality and consumer protection, applied to a new technology.
Add a Comment
Your email address will not be published.We'd love to help you!
Let us know the needs of your business, and we will pinpoint the best-suited solution to fulfill them.
