AI Governance.
Using AI in project controls under clear, human-centred governance.
Overview
AI governance is what keeps AI use trustworthy: data quality, confidentiality, explainability and — above all — human accountability for every consequential output. It is the difference between useful AI and dangerous AI.
What this area covers
Why it matters
This is a core part of the PCL-AI body of knowledge — assessed as part of the credential and applied on real projects. It connects to the wider discipline and, increasingly, to responsible AI use, so professionals can demonstrate the integrated judgement PCI certifications assess.
Taught, then tested.
Every knowledge area maps to the examination blueprint and is assessed through realistic, scenario-based questions — not rote recall. Explore the full body of knowledge or the certification roadmap.
Common questions
Is this part of the PCI examinations?
Yes — this is one of the knowledge areas assessed in the PCI examinations. The exam is built around the twelve-competency model, so each competency, including the governed use of AI, is tested as part of an integrated whole rather than in isolation. The emphasis is on applying it with judgement in realistic project scenarios, not on reciting definitions.
Do I need prior expertise?
No prior expertise is needed to get involved or to begin preparing. For certification specifically, the entry requirement is around three years of relevant professional experience in any field rather than a particular qualification — the aim is to keep the credential open to capable people from many backgrounds. What matters is your ability to meet the standard the assessment sets, which you can work towards at your own pace.
How does AI fit in?
AI runs through everything PCI certifies, but always under the principle at the heart of the standard: AI proposes, the professional disposes. AI governance is treated as a competency in its own right, and the responsible use of AI is woven through the other competencies too. The point is not to use AI for its own sake, but to use it well — validating, explaining and owning AI-assisted outputs so that accountability stays with a competent human.
Why this matters
This matters because a credential earns its value from substance, not marketing — clear standards, fair process, transparent governance and honesty about status. Everything in the institute's resources is written to that test: genuinely useful to professionals and employers, and never claiming more than is true today.
PCI builds in the open. That means being candid about what is in place and what is still developing, refusing to publish invented data or figures it cannot stand behind, and letting the community shape what gets prioritised. Trust, earned this way, is harder to lose.
- Substance over marketing
- Fair, transparent process
- Honesty about our status
- Responsible, governed use of AI
Governing one output at a time
Governance is easiest to grasp at the scale of a single deliverable. Before an AI-assisted figure enters a forecast, report or recommendation, a disciplined practitioner works through five checkpoints.
- Define the decision. Know what the output will feed before you generate it. The consequence of being wrong sets the depth of review the output needs.
- Check the inputs. Confirm the provenance and sensitivity of the data first. Commercially confidential information never goes into a tool your organisation has not approved.
- Validate independently. Test the output with a method that does not share the tool's assumptions — a manual estimate, a historical comparison, a rule-of-thumb cross-check.
- Record the reasoning. Note which tool was used, on what inputs, and why the result was accepted, adjusted or rejected. If it cannot be explained, it cannot be defended.
- Own it by name. A named, competent person signs the figure. Accountability never transfers to the tool — that is the heart of human oversight.
Three tests of well-governed AI use
Well-governed AI use has a recognisable texture. When reviewing an AI-assisted deliverable, look for three properties — and be wary when any is missing.
Reproducible
Another practitioner, given the same inputs and the stated method, could regenerate something materially similar. If the output changes on every attempt and nobody can say why, it is not yet evidence.
Explainable
The path from input data to final figure can be narrated without resorting to “the model said so”. The explanation must satisfy someone accountable for the decision, not merely someone curious about the tool.
Challengeable
There is a real route for a reviewer to reject the output — and evidence that the route gets used. A process under which no AI output has ever been rejected is not governance; it is decoration.
PCI applies the same tests to its own operations through its Responsible AI Framework and AI decision policy.