An employee uses an approved AI tool to prepare a campaign brief and submits the finished work hours ahead of schedule. In an appraisal built around deadlines, that speed earns recognition. To assess the employee’s contribution, their manager also needs to understand how they arrived at the campaign idea and whether they checked the material before it reached customers.
That closer examination becomes important as AI takes on more of the work involved in producing a finished assignment. A polished report gives a manager limited insight into the reasoning behind its recommendations, while the time recorded for a task omits the problems an employee caught before submission. Decisions about pay and promotion depend on how well the appraisal captures those contributions.
Nigerian employers should revise their performance criteria wherever AI changes how employees carry out their responsibilities. The extent of that change should follow the work itself, with clear expectations about what employees are accountable for and the evidence managers will use to assess them.
How AI Changes the Evidence Used to Assess Performance
Much of the appeal of generative AI comes from how quickly it produces something people can work with. Employees can organise research or draft correspondence within minutes, leaving more time to examine the material and decide how to use it. As this becomes part of a role, managers need to understand what happens between the first output and the work reaching its intended audience.
Research involving 758 Boston Consulting Group consultants shows why that process deserves attention. On suitable assignments, participants using GPT-4 completed tasks 25.1% faster and produced higher-quality work than participants without AI. On a complex managerial task outside the technology’s capabilities, AI-assisted participants were 19 percentage points less likely to reach the correct answer. The experiment, conducted in 2023 and published in Organization Science in 2026, demonstrated how sharply results could vary with the demands of an assignment.
For an employer reviewing AI-assisted work, those findings make the employee’s decisions during an assignment worth examining. A financial report requires someone to establish whether the underlying figures support its conclusions. When the evidence is incomplete, the employee needs to recognise the gap and determine what further work is necessary before management acts.
Within a contact centre, the consequences become visible through the customer’s experience. An agent using suggested responses still has to understand the complaint and check that the proposed answer follows company policy. A billing dispute that recurs the following week gives the manager useful evidence about the original handling of the case, particularly when the record explains what the agent knew at the time.
Performance measures should capture enough of that experience to establish whether the work achieved its purpose. Handling time can sit alongside repeat enquiries, with supervisors examining a sample of completed cases to understand the quality of the response. That gives an employer a firmer basis for recognising an agent’s contribution and identifying where the service needs improvement.
Setting Performance Targets that Account for AI Assistance
Once AI begins improving delivery, employers face a practical question about what a reasonable target should become. Previous output levels provide a starting point, although managers also need evidence about how the tool affects employees with different levels of experience.
A study of 5,172 customer-support agents, published in The Quarterly Journal of Economics in 2025, found that access to a generative AI assistant increased issues resolved per hour by 15% on average. Less experienced and lower-skilled agents recorded gains of about 30%. Researchers also found evidence that the tool helped employees learn, giving newer staff access to practices associated with more capable colleagues.
An employer introducing similar assistance should allow those differences to inform target-setting. A substantial improvement after training provides evidence that an employee is learning to work effectively with the tool. The manager must also examine the assignments involved, including the complexity of the cases allocated to experienced colleagues.
Before raising targets, the business needs to establish the effort required to produce work that is ready for use. Checking an AI-generated analysis takes time, as does correcting a draft that contains unsupported claims. Colleagues who receive the work can help explain whether faster delivery has reduced their workload or created additional checking further along the process.
The use of time released by AI also belongs in the performance conversation. When a marketing team can prepare initial material more quickly, its manager should agree how that capacity will support the campaign, perhaps through deeper customer research or closer testing of the message. Employees then have a clear basis for deciding where to direct their effort, and the review can examine what that additional work achieved.
Recognising Professional Judgement in Employee Appraisals
As employers reset expectations, they need language that makes critical thinking and professional judgement observable. Broad competencies such as “analytical ability” become more useful when managers explain what they mean within an employee’s actual responsibilities. Completed assignments provide material for that discussion.
A finance manager reviewing an AI-assisted forecast can examine how the analyst checked the source data and dealt with assumptions that affected the result. The employee’s explanation of an unusual movement helps establish whether they understood the figures well enough to advise the business. Where the analyst changed a recommendation after investigating a discrepancy, that decision belongs in the record of their performance.
In software development, a review can examine whether an employee understands the code they submit and can diagnose problems when it fails. Recruitment staff face a different responsibility when AI helps organise candidate information: they must check that the resulting assessment is supported by evidence relevant to the role. Each appraisal should connect the employee’s decisions to the consequences of the work they were trusted to deliver.
HR teams using AI to summarise employee surveys face another kind of judgement. A recurring concern in the responses requires interpretation before it becomes a recommendation to management. The reviewer should examine whether the employee considered the limits of the sample and understood the organisational circumstances surrounding the findings.
These discussions can draw on existing project records and a small sample of significant assignments. Employees should be able to explain where AI helped and what they did before accepting its output, with the depth of the review matching the importance of the work. A routine email calls for a proportionate check; a report influencing a major investment requires a more substantial account of the evidence.
The same approach helps employers make sound promotion decisions. Someone moving into greater responsibility needs opportunities to demonstrate that they can reason through unfamiliar problems and explain consequential choices. Managers should build those opportunities into talent development programmes, with assignments and coaching that develop the expertise expected at the next level.
Why Access to AI tools Matters for Fair Assessments
Fair comparisons require an understanding of the conditions under which people work. Approved tools and relevant training affect what an employee can achieve, while access to suitable devices determines whether those tools fit into daily tasks. Employers should establish what support was available before attributing differences in output to individual capability.
Microsoft and LinkedIn’s 2024 Work Trend Index provides useful international context. Among respondents who used AI at work, 78% reported bringing their own AI tools into the workplace. The wider survey covered 31,000 knowledge workers across 31 markets, showing how much employee experimentation was taking place through tools individuals selected themselves.
For a Nigerian organisation assessing employees across several locations, that raises practical questions about access. HR should establish whether staff had approved licences and reliable connectivity during the review period, then examine whether training was accessible and relevant to their roles. Expectations about AI proficiency should be communicated with enough time for employees to practise and receive feedback.
The appraisal also needs room for work that helps other people perform. An experienced employee who develops a dependable way to check AI-generated reports contributes to the quality of the whole team’s output. Managers can recognise that contribution through evidence that colleagues adopted the method and that it improved the work, supported by feedback from those who used it.
This gives employers a way to recognise learning that spreads beyond one person’s results. It also encourages employees to share useful practices, because the performance system makes their contribution to colleagues visible.
Managers’ Responsibilities When AI Supports Appraisals
AI also changes how managers prepare appraisals and the support they need to lead their teams. A tool can organise project notes or produce a draft review, giving the manager a starting point for the conversation. The quality of that review depends on the evidence supplied and the care taken to examine what the draft says about the employee.
A confident narrative can give incomplete records an appearance of authority. If the source material omits time spent resolving a difficult client issue, the appraisal can carry that omission into a rating. Managers should check significant claims against the underlying records and give employees a practical opportunity to supply relevant context.
Data protection becomes part of this responsibility whenever a system processes personal information about employees. Section 24 of the Nigeria Data Protection Act 2023 requires fair and lawful processing, with transparency about how data is used. Employers must also address accuracy and collect information appropriate to the stated purpose, alongside the Act’s requirements concerning retention and security.
Section 37 addresses decisions based solely on automated processing that produce legal or similarly significant effects, subject to specified exceptions and safeguards. Section 28 requires a prior impact assessment where processing is likely to create a high risk to people’s rights and freedoms. Employers introducing systems that score or profile staff should assess these obligations with their data protection advisers before using the results in consequential employment decisions.
Within the appraisal process, employees should understand which records inform a rating and how they can challenge an error. The manager conducting the review needs enough information and authority to investigate a recommendation, with time to consider the employee’s explanation. HR should be able to establish how the final decision was reached when a rating affects someone’s career.
Testing Revised Criteria Before They Affect Pay and Promotion
A useful starting point is a department where employees already use approved AI tools regularly. HR and the department head can examine completed assignments to identify how the work has changed, then review whether their performance management framework captures the responsibilities employees now carry. Involving staff at this stage helps uncover checking and follow-up work that managers would otherwise miss.
From there, the team can revise the criteria for a small set of responsibilities. In customer service, that could mean examining resolution quality alongside handling time and accounting for the complexity of a case. A marketing team would need to connect faster delivery to the campaign’s purpose, with commercial outcomes assessed in light of the decisions and resources the employee controlled.
Managers should apply the proposed criteria to completed work and discuss the evidence behind their ratings. Where they disagree, HR can examine whether the criteria are unclear or whether the record needs more context. This exercise helps establish what “meeting expectations” means before the revised approach influences pay or promotion.
During the next cycle, regular conversations can capture significant decisions while the details remain fresh. An employee who catches an error in an AI-assisted analysis should have the chance to explain what happened and how they resolved it. Those discussions also allow managers to adjust support as
