AI Is Starting to Judge Your Employees
Two things happened in the same week, and together they mark a line being drawn. Mainstream HR platforms began shipping AI agents that draft performance reviews from evidence gathered across a company’s systems. And California moved a bill, nicknamed the No Robo Bosses Act, that would bar employers from letting AI alone decide who gets fired. AI has finished with HR paperwork and started on people. Whether that turns out well for your business depends almost entirely on where you decide the machine stops and the manager starts.
Why AI in reviews can be an improvement
It is worth resisting the reflex that AI near people decisions is automatically sinister. The status quo is not great. Most performance reviews are written from memory under time pressure, which reliably over-weights the last month and whoever happens to be most visible to the manager. An AI that assembles what actually happened across the year, projects delivered, feedback given, goals met and missed, can produce a fairer, better-evidenced starting point than recollection. The problem was never AI gathering evidence. It is AI forming the verdict.
Where the line goes
The distinction that keeps you both effective and defensible is between assembling information and making a judgment.
| Good use of AI | Keep human |
|---|---|
| Assembling the year’s evidence | Deciding the rating |
| Drafting a first version to edit | Owning what the review says |
| Surfacing patterns a manager missed | Promotion and pay decisions |
| Summarizing feedback fairly | Discipline and termination |
This is the same boundary we drew around AI screening résumés, now applied to the people already working for you. Hiring and firing sit at opposite ends of the employment relationship, and the rule holds at both ends.
The Canadian reality, regardless of California
A California bill does not bind a business in Calgary, but do not take much comfort from that. Canadian employment and human rights law already expects decisions about discipline and termination to be defensible, with reasoning you can articulate. If a decision is challenged, the system flagged them is not a defence, it is an admission that nobody can explain the call. Which means the practical standard is identical no matter which legislature moves first, and it connects to the broader question of who is accountable when AI is wrong: the answer is always you.
Tell people, and watch the patterns
Two last habits make the difference between AI improving your people processes and quietly poisoning them. First, tell employees where AI is used in evaluation. Discovering it later feels like concealment, and it converts a reasonable tool into a trust problem, the same logic behind disclosing AI to customers. Second, look periodically at whether AI-assisted assessments are landing differently across groups, because the thing scale does best is turn a small bias into a systematic one. Get those right and AI makes your reviews fairer and faster. Get them wrong and it industrializes whatever was already unfair.
Frequently Asked Questions
What is actually changing in HR tools?
AI is moving from helping with HR paperwork to weighing in on people themselves. In 2026, mainstream HR platforms began shipping AI agents that draft performance reviews from evidence gathered across a company’s systems, and consultancies started publishing frameworks for running whole HR functions around agentic AI. At the same time, lawmakers noticed. California’s proposed No Robo Bosses Act would bar employers from letting AI alone decide who gets fired. The technology and the guardrails are arriving in the same season, which is a useful signal about where the line is being drawn.
Is it wrong to use AI in performance reviews?
Not inherently, and used well it can make reviews better rather than worse. Managers routinely write reviews from memory, which favours recent events and whoever is most visible. AI that assembles the actual evidence, projects completed, feedback given, goals met, can produce a fairer starting point than recollection. The problem is not AI gathering evidence. It is AI forming the judgment. A draft assembled from real work that a manager then reads, corrects, and owns is a genuine improvement. A rating the manager rubber-stamps is not.
What is the "No Robo Bosses" idea about?
It is a proposed California law that would prevent employers from relying on AI as the sole decision-maker in consequential employment actions, most notably termination. The principle behind it is simple and likely to spread: decisions that seriously affect someone’s livelihood should have a human being accountable for them, able to explain the reasoning. Whether or not that specific bill passes, the direction is clear across jurisdictions. If your process lets software effectively decide who goes, you are on the wrong side of where the rules are heading.
Does this apply to Canadian employers?
The California bill would not bind a Canadian business directly, but the underlying obligation already does in substance. Canadian employment and human rights law expects that decisions about hiring, discipline, and termination are defensible, and that you can explain the reasoning behind them. Saying the system flagged this person is not an explanation, and it will not protect you if a decision is challenged. So the practical standard is the same regardless of which legislature acts first: a human decides, and that human can articulate why on grounds related to the job.
What is a sensible policy for a small business?
Draw the line at judgment. Let AI do the gathering and drafting: pull together the evidence, summarize feedback, draft the first version of a review, surface patterns a busy manager might miss. Keep humans making every actual determination about rating, promotion, discipline, and termination, and make sure that person can explain the decision in their own words. Tell employees where AI is used in the process, because finding out later feels like something was hidden. And check periodically whether AI-assisted assessments are treating groups differently, since scale turns a small bias into a systematic one.
Use AI in HR without losing accountability
We help Canadian businesses put AI to work on evidence and drafting in people processes, while keeping judgment, explanation, and responsibility with your managers.
Related Articles
AI Just Got Much Better at Sign Language
Your Human Reviewer Is Probably Rubber-Stamping
Using AI to Screen Résumés? Read This First
AI consultants with 100+ custom GPT builds and automation projects for 50+ Canadian businesses across 20+ industries. Based in Markham, Ontario. PIPEDA-compliant solutions.