Change Signals · AI & work
The tool is in the room
Canada’s AI-transparency consultation closed. The work questions it raises are already here.
This week’s public signal · Canada
The consultation closed on 23 September 2026.
This was a policy-development process, not a new legal requirement. ISED says a “What We Heard” report will follow.
Read the ISED consultation page- Signal type
- Technology reshapes work
- Jurisdiction
- Canada
- Public status
- Consultation closed
AI & work
A team can adopt an AI tool without calling it a transformation.
It starts as a shortcut. A faster first draft. A way to summarise calls, route requests, prepare a customer response, screen a queue, or find the thing nobody can find in the shared drive.
Then the work changes shape.
Someone sees a recommendation before someone else sees the underlying information. A manager approves an output they did not create. A customer receives a response no one can fully explain. The person who used to make the first call now receives a finished-looking answer and is expected to make the last one.
Nothing may be wrong. The people involved may be doing sensible things with a useful tool.
The work can still stall.
This week’s signal
On September 23, the Government of Canada closed its consultation on advancing AI transparency. It was not a new legal requirement for employers or businesses. It was a policy-development process about how transparency and accountability may need to work as AI becomes more present in Canadian products, services, and organisations.
The consultation named five areas where transparency questions are becoming harder to avoid: identifying AI-generated content; knowing when someone is interacting with an AI system; understandable information about a system’s capabilities and limits; tracking serious incidents; and tracking the activity and interactions of AI agents.1
That is a public-policy signal. It is also an organisational one.
When an AI system starts shaping a workflow, the question is not only whether the tool works. It is whether the people around it still know how the work moves.
The line that becomes hard to see
Most teams do not need a seminar on transparency before they can use a new tool. They need to know where judgment still lives.
If an AI assistant drafts the response, who owns the promise inside it? If an agent routes a request, who notices the exception it was not trained to recognise? If a manager relies on an AI-generated summary, where does the original account go when the summary is wrong, incomplete, or too confident?
These are not only technology questions. They are questions about authority, visibility, trust, and handoffs.
They are also easy to miss because each person can still be doing their job well. The operations lead sees a faster queue. The IT team sees approved software. The manager sees a polished response. The person closest to the customer sees the part of the work that no longer quite fits, but may not have a clean place to put that concern.
The friction sits between them.
What a diagnostic would look for
An Unflocked organisational diagnostic would not begin by auditing the AI tool or scoring the organisation’s “AI readiness.”
It would begin with confidential conversations across the people carrying the changed work. Not to manufacture consensus. To find where the practical account of the work has split.
The listening would stay close to a few ordinary questions:
- What now arrives in your hands already decided, summarised, or routed?
- Where do you override the tool, work around it, or add an unofficial check before the work can move?
- Which decision still needs a person, but no longer has a clearly named owner?
- What information can one part of the team see now that another cannot?
The point is not to decide in advance that the tool is a problem. It is to make the changed conditions visible before reasonable people build separate workarounds around them.
Then the people with authority to act can decide what needs attention: a handoff, a decision right, a customer commitment, a disclosure, a review step, or a conversation that has been postponed because the new workflow feels temporary.
When the next step is Praxis
Sometimes the organisation can already name the problem. The difficult part is that the people involved need to practise a different conversation while the stakes are still manageable.
That is where Praxis fits. Not as training on the software. As a structured room where a team can rehearse the decision conditions the changed workflow has created: raising a concern about a confident-looking output, disagreeing with an automated recommendation, clarifying who can override it, or telling a customer what the organisation can and cannot stand behind.
The room makes hesitation, recovery, and decision patterns visible before they have to play out in front of a client, a regulator, or a frustrated colleague.
The signal is not the diagnosis
Canada’s consultation does not tell us which organisations are prepared for AI-enabled work, which teams are struggling, or what rules may eventually be adopted. It is not evidence that a specific company has a problem.
It is a public sign that the questions have moved beyond the tool itself.
The useful question for a leader is smaller and closer to the work:
“What has this tool changed about who knows, who decides, and what happens when the answer is not enough?”
If that question keeps returning, bring it to the people living the work. The change may have started in the software. The next decision still belongs to the team.
Source
- Have your say on advancing AI transparency in Canada. Innovation, Science and Economic Development Canada. The consultation closed on 23 September 2026. ISED says a “What We Heard” report will follow.