Day 5: Draft a Performance Review

The Concept

The annual performance review sits in an awkward position in most organisations. Everyone agrees it matters. Surveys consistently show that both managers and employees find the process stressful and frequently unhelpful. And yet the form keeps appearing on the HR calendar, the ratings keep being assigned, and the calibration conversations keep happening — often with less rigour than the process assumes. The problem is not that performance management is unimportant. The problem is that the most common format for it — a document written by a manager, once a year, under time pressure — produces outputs that are too vague to be useful and too inconsistent to be fair.

Understanding why this happens is the first step to fixing it. Managers are not bad at giving feedback. Most of them are good at it in real time, in conversation, in the context of a specific piece of work. They are bad at constructing a written document from memory that covers twelve months of performance, uses the organisation's approved competency framework, avoids legally problematic language, and still sounds like something a human being wrote. That is a significant writing and analytical task, and few managers have been trained to do it.

The Real Purpose of a Performance Review

It is worth being clear about what a performance review is actually for, because the answer shapes how you use it. It is not primarily a historical record of what happened — though it serves that function. It is not primarily a rating that determines a pay increase — though it may feed into that. At its best, a performance review is a structured moment of clarity between a manager and an employee about where the employee is relative to expectations, what has driven their results, what is getting in the way of higher performance, and what the plan is for the period ahead.

Understood that way, the value of the document is not in the document itself. It is in the quality of the conversation the document enables. A well-written review gives the employee specific, actionable information they can engage with. A vague one gives them something they cannot argue with or act on — which is why the most common employee response to a performance review is a resigned shrug rather than genuine reflection.

Why Vague Reviews Harm Employees and Managers

Vagueness in performance reviews creates two distinct problems. For the employee, it denies them the information they need to improve. "Good communicator" tells a person nothing useful. "Effectively translated complex technical analysis into language the executive team could act on in the Q3 board presentation" tells them exactly what they did well and implies what they should keep doing. The difference between those two descriptions is not just clarity — it is the difference between feedback that builds someone's career and feedback that fills a form.

For the manager and organisation, vagueness creates legal and fairness risk. A performance review that describes a poor performer as "not a great cultural fit" or "lacks gravitas" — without specific behavioural evidence — is both useless as a development tool and potentially discriminatory in a way that becomes visible only when challenged. AI is particularly useful here because it can flag language that sounds evaluative without being evidential, and suggest how to reframe it in terms of observable behaviours. That is not about softening difficult feedback. It is about making difficult feedback defensible and useful.

Keeping the Manager's Voice in the Output

The most important instruction in today's prompt is the last one: preserve the manager's genuine assessment. AI has a tendency to produce reviews that are more diplomatically phrased than the manager's notes, which can inadvertently soften feedback that needs to land clearly. If your notes say a person missed three deadlines and consistently failed to flag risks early, the review needs to say that — specifically, with examples, and with the same weight it deserves. A review that smooths over performance problems is not kind. It is a disservice to the employee, who needs accurate information to make decisions about their own development.

Read the AI-generated draft as a structure, not as a final document. Where the language has become more hedged than your actual view, rewrite it. Where it has introduced competency-speak that dilutes the specificity of your examples, replace it with your own words. The AI handles the architecture; you bring the honest assessment that makes the architecture worth building.