We all know that Artificial Intelligence is rapidly changing the way we work. So far, much of the discussion has focused on productivity gains. AI can summarise documents, search regulations, draft reports, analyse data and generate learning content in seconds. Tasks that once took hours can now be completed almost instantly.
For Training & Competency professionals, the conversation has largely focused on how AI can help us streamline the time it takes to create learning content or complete a task. However, I think the more interesting question is not what AI can do for us, but given all that is now available at the touch of a button, what do humans still need to learn?
Historically, competence has been built around three core components:
Knowledge
Do they know the rules, regulations, products and processes?
Understanding
Do they understand what those rules mean and why they exist?
Application
Can they apply that knowledge appropriately in real-world situations?
Most training and assessment frameworks are built around these principles, and they have generally worked well for decades. We teach knowledge, test understanding and assess application through scenarios, observations, quality monitoring and case studies, but AI is beginning to challenge one of the underlying assumptions.
If AI can increasingly support knowledge, understanding and application, perhaps Training & Competency frameworks need to evolve beyond those three components. Perhaps a fourth component is emerging
Traditionally, if someone could answer questions correctly, complete assessments and work through case studies, we had reasonable confidence that they understood the subject matter. Today, even with the current limitations of AI, it can generate answers to knowledge-based questions and, with careful prompting, assist someone in applying that knowledge to a prescribed set of circumstances. This does not make knowledge less important, in fact, it may make it more important. The real challenge is now proving that the knowledge genuinely exists.
Consider an experienced adviser, compliance professional or T&C specialist using AI. They do not use it to do their job. They use it to accelerate parts of their job:
- AI can find information.
- AI can organise information.
- AI can even suggest conclusions.
However, what it cannot reliably do (currently) is exercise professional judgement. The experienced professional knows when something feels incomplete. They recognise when context is missing. They spot assumptions, inconsistencies and potential customer harm. Perhaps most importantly, they know when to challenge the answer.
That is where competence becomes interesting.
If AI can increasingly support knowledge, understanding and application, perhaps Training & Competency frameworks need to evolve beyond those three components. Perhaps a fourth component is emerging:
Challenge
- Can an individual identify when an answer is incomplete?
- Can they recognise when a recommendation appears reasonable but may create a poor customer outcome?
- Can they identify bias, missing information or vulnerability indicators?
- Can they challenge the output rather than simply accept it?
In a regulated environment, these questions matter. Imagine an AI-assisted lending decision undertaken through an online application. The affordability assessment passes, the policy requirements are met, and the recommendation is technically correct. Yet during the implementation process, which includes a scheduled interaction with the customer, they appear confused, distressed, or unable to fully understand the implications of the decision being made. Does the employee recognise the risk and act on it? Or do they accept that the system has followed the rules and allow the loan to go ahead? Historically, many assessments have focused on identifying the correct answer. Increasingly, we may need to assess something different.
Instead of asking, “What would you recommend?” we may need to ask, “What concerns you about this recommendation?”
Instead of asking, “Which answer is correct?” we may need to ask, “What is missing from this answer?”
Instead of asking, “Can you apply the process?” we may need to ask, “Should the process be challenged in this situation?”
These are fundamentally different assessments. They require critical thinking rather than recall, they require professional scepticism rather than procedural compliance, they require judgement rather than repetition and they require a colleague to ‘Act with Integrity’ and ‘Act with Due Skill, Care and Diligence’.
This has significant implications for Training & Competency professionals. If AI can generate training content, model answers and realistic scenarios, our role may increasingly shift towards creating opportunities for challenge rather than simply testing knowledge. Future assessments may include:
- Deliberately flawed recommendations.
- Missing customer information.
- Hidden vulnerability indicators.
- Technically compliant but ethically questionable outcomes.
- Conflicting pieces of evidence requiring professional judgement.
The objective would no longer be to identify who can find the answer fastest, but it would be to identify who can recognise when the answer should be questioned. This also raises a practical question: how should we assess this in an AI-enabled world?
A multiple-choice test may still have a place for checking baseline knowledge, but it is unlikely to be enough to evidence professional judgement. Remote assessment software may still be useful, particularly where it presents adaptive scenarios, asks learners to critique AI-generated outputs, or requires written reasoning rather than simple right-or-wrong answers. Some aspects of competence may be better assessed through old-fashioned discussion. Classroom or virtual workshops, professional conversations, role plays, case study reviews and peer challenge sessions have always allowed assessors to hear how someone thinks. They reveal whether a colleague can explain their reasoning, challenge assumptions, identify missing information and recognise where customer harm could arise.
In practice, the future may require an increasingly blended approach:
- Remote tools to test knowledge and present realistic AI-generated scenarios.
- Written responses to assess reasoning and professional judgement.
- Facilitated discussions to explore challenge, ethics and customer outcomes.
- Observations and quality assurance to confirm whether those behaviours appear in real work.
Importantly, I do not see this as an argument for reducing standards or removing qualifications. In fact, it is quite the opposite, as without sufficient knowledge, understanding and experience, how can anyone effectively challenge an AI-generated recommendation?
For years, Training & Competency has focused on proving that people know what to do. The next challenge may be proving that they know when not to do it. I think the foundation we know remains essentially the same, but what changes is simply how we evidence competence. Knowledge remains important, understanding remains important, application remains important but, in an AI-enabled world, they may no longer be enough. The future competent employee will need to demonstrate the fourth capability of challenge, the ability to recognise when an apparently correct answer is incomplete, inappropriate or potentially harmful. As learning professionals, we need to consider how we will test and evidence this appropriately in an AI-driven world, while ensuring we do not unnecessarily overburden either the individual or the journey to reach competence.
