Skip to content
JiraVision
All articles

Ethical AI

Ethical considerations in AI-powered project management

The ethical question is rarely whether the model is correct. It is who is accountable when a person acts on it — and whether they could have said no.

3 min read24 October 2026JiraVision team

A person weighing a decision, hands resting on a desk

AI in project management is rarely a robot making decisions. It is a recommendation, a risk score, a suggested owner, a summary of a thread nobody read. That framing matters: the ethical question is almost never "is the model correct" but "who is accountable when a person acts on it".

Explainability is a feature, not a nicety

If a tool flags a task as at-risk, the person receiving that flag needs to know which inputs produced it. Without that, they either ignore the system entirely or defer to it uncritically — both of which are worse than not having it. Prefer models and prompts whose reasoning can be surfaced in plain language alongside the recommendation, and store the reasoning with the record so it can be audited later.

This is why every metric in JiraVision prints its own formula. A number you cannot reproduce is not a measurement, it is an assertion.

Watch for proxy discrimination

Historical project data encodes past decisions, including unfair ones. A model trained on who was assigned stretch work, whose estimates were trusted, or who got flagged as a bottleneck can reproduce those patterns while looking purely statistical. Two guards help:

  • Review any per-person score for patterns across teams, locations, tenure or role before it influences a decision.
  • Keep human judgement in the loop for anything that affects a person's work allocation, appraisal or reputation. Automate reminders, not consequences.

Be honest about the data

Tell people what is collected, what is used to train or tune models, and what is retained. Wellbeing and sentiment signals deserve the strictest treatment: they are the most useful inputs an AI system can have and the easiest to misuse. Aggregate by default, restrict access to individuals, and never let a check-in score become a performance input without the team knowing.

Design the override path first

Every automated recommendation needs a visible way to disagree, a record that the disagreement happened, and no penalty for using it. If dismissing a suggestion is harder than accepting it, the system is not assisting — it is deciding.

In JiraVision a dismissed recommendation is archived with the reason you gave, so the disagreement is legible later. That archive is not a nice-to-have; it is the audit trail for the claim that a human stayed in the loop.

The metrics we refuse to build

Some measurements are technically easy and organisationally corrosive. These are not on the roadmap:

  • Lines of code, commits per person, or any raw activity count presented as productivity.
  • Keyboard time, idle detection, or screen monitoring of any kind.
  • Per-person velocity rankings shown to the group.
  • Sentiment analysis of chat content used as a performance signal.

Publishing a refusal list is itself part of the ethics work. A team cannot consent to a metric they cannot predict.

A practical checklist

  1. Can a person see why this recommendation was made?
  2. Can they override it, and is the override recorded?
  3. Who is accountable for the outcome — the tool, the operator, or the manager who approved it?
  4. What personal data went in, and who can still read it?
  5. When was the last time anyone checked the outputs for unfair patterns?

Answering those five questions before rollout turns most ethical risk into an ordinary product decision. Leaving them unanswered is what turns a useful assistant into a liability.

Written by the JiraVision team. Corrections and disagreements welcome via the contact page.

Send feedback

Keep reading

AI Scrum Master

The complete guide to AI Scrum Masters

What an AI Scrum Master actually does: where the draft comes from, what the rationale contains, and why nothing reaches your board until a person accepts it.

Read article

AI Scrum Master

How AI changes the agile routine

The ceremonies an AI assistant genuinely improves, the ones it leaves alone, and the five numbers worth measuring in your own workspace instead of trusting ours.

Read article

Read the next one in your inbox

Double opt-in, no tracking pixels, and an unsubscribe link in every message.

Create an account