AI in learning and development: what most L&D teams are getting wrong

Here’s a quick summary

• 93% of Australian organisations use AI in some capacity, but AHRI found that adoption remains concentrated in operational areas rather than strategic capability building.
• 75% of organisations say HR is driving AI training, but 63.6% of HR departments have provided no advice to the business on what AI actually changes about how people work.
• Jobs and Skills Australia found generative AI is more likely to augment roles than replace them, but automation potential is concentrated in routine tasks, exactly the ones that dominate many L&D functions today.
• 70% of Australian employees feel their employer isn’t adequately preparing them for an AI-driven future, and only 17% have received any AI training.
• The AITD’s research is direct: ‘The core challenge is not a lack of skills. It is a design gap.’
• L&D’s role is shifting from content supply to capability design, building the judgement, adaptability, and team-level performance that AI makes more valuable.

Every enterprise L&D leader eventually notices the same thing. The programmes get delivered, and the completion rates look fine. Six months later, the same capability gaps are back, and nobody in the business seems to have noticed they never really left.

AI has entered that conversation as the fix. The pitch is familiar: generate content faster, personalise learning experiences at scale, and close skills gaps with predictive analytics. AHRI’s December 2025 data shows 93% of Australian organisations are already using AI in some capacity.

But the adoption numbers don’t tell you that almost all of that AI effort is landing on the production side of L&D, content drafting, training administration, policy writing, completion tracking, not the impact side. AHRI’s own analysis is blunt about it. AI use in Australian HR ‘remains concentrated in operational areas such as HR analytics, training delivery, recruitment and policy drafting.’ At the same time, workforce planning, talent management, and performance management lag.

Where most L&D teams are pointing their AI attention right now

When you look at where spend and effort are concentrated in AI within L&D functions across Australia, a pattern emerges. AHRI’s breakdown across HR functions concludes that AI in L&D is most used throughout reporting and analytics, training and development, recruitment and onboarding, and HR policy drafting. It’s used least in performance management and workforce planning.

This forms a gap in an L&D AI strategy. The areas where AI adoption is highest are all operational, including producing, administering, and tracking content. The areas where it’s lowest, like AI for employee development, are the ones that determine if any of that content changes how people work.

Across L&D teams, the same five use cases come up repeatedly:

Content drafting

Generative AI is now a core tool for L&D teams drafting modules, rewriting policy content, and building scenario scripts. It’s faster than starting from scratch, and most teams have adopted it in some form. But just because production is faster, this doesn’t fix a design problem. If the original module was a passive slide deck that nobody applied after completing it, producing five variations of it in half the time doesn’t improve anything; it just creates a larger library of training that doesn’t transfer or provide genuine learning outcomes.

Although AI for workplace training in Australia can help in some ways, there are instances where a human approach is needed. AI promises real-time feedback at faster speeds, yet teams will spend hours reviewing and editing outputs. The issue isn’t with what AI generates; it’s the lack of time and effort that humans invest in defining what the training needs to do before AI is used.

Quiz generation

AI tools can generate knowledge-check questions from source material almost instantly. Useful for scaling compliance assessments, but it reinforces a model where training success is measured by the ability for someone to recall information, not their ability to apply it under real working conditions. Lam et al. (2024) also found that AI-generated multiple-choice questions were comparable in quality to human-written ones, but only when the AI was given explicit prompt instructions including a discipline-relevant framework and example questions. Generic prompting didn’t get there.

LMS administration

Automated enrolment, progress nudges, overdue reminders, and reporting dashboards. This is where AI can save hours of admin time, and it’s one of the least controversial use cases. Employment Hero estimates the average employee spends close to two hours a day on administrative work that could be automated. However, the risk lies in mistaking administrative efficiency for learning effectiveness.

Translation

AI-powered translation has made multilingual rollouts faster. For organisations operating across states or with diverse workforces, this is a practical win. But translated content is still content, and content alone is rarely what was missing. There is also still room for error depending on what languages are being translated, which can lead to inconsistencies in translation quality.

Chatbots and virtual learning assistants

Always-on support tools that answer learner questions, surface relevant resources, and guide people through modules. They reduce the load on facilitators and help learners who get stuck outside business hours. They don’t replace the facilitated conversation where someone works through a difficult scenario with a colleague.

When the vast majority of your AI investment is accelerating production, and almost none is improving whether training changes behaviour, you’re implementing AI on the wrong side of the equation.

Danger 1 – Using AI to make bad training faster

Australian employers are investing billions of dollars into workplace training every year. Yet, the Productivity Commission’s 2025 workforce enquiry found that training participation has stagnated. The spend is high, but the participation is low and persistent gaps continue to appear.

The ABS’s 2024–25 Work-Related Training and Adult Learning survey puts a number on the disconnect. Participation in work-related training fell to 19%, down from 23% in 2020–21. A significant share of work-related training in Australia is compliance-driven, with modules that are knowledge-intensive, delivered online, repetitive, and not necessarily designed to change practice.

This is where AI content creation can become a danger, because the AI tools in L&D aren’t solving the pressing issues, and it’s happening at scale. Content is being produced, and modules are being completed, but the gap is between completion and real-world application.

The organisations we work with through instructional design projects tell us the hard part isn’t building the content, it’s identifying the skill gaps, scoping the right learning objectives and mapping them to real behaviours before any content is written. AI can draft faster when the thinking has been done prior; it can’t replace the human ability to do the actual thinking.

Danger 2 – Treating AI as a subject to teach, when it’s a shift in how work happens

The second danger can be harder to argue against, as it seems like a natural progression for L&D professionals adopting an AI learning platform. Teach people what generative AI is, how to write prompts, and which tools to use. This sounds responsible; however, it treats AI as a subject or another skill to add to your repertoire, when it’s really shaping how the work gets done.

The data shows the gap. AHRI’s December 2025 Work Outlook found 75% of Australian organisations say HR is driving AI training. Yet the AHRI/QUT study found 63.6% had provided no advice to the business on what AI actually changes about how people work. HR is training people about AI without helping them adapt their roles to it.

There’s a practical difference between teaching someone to write a prompt and helping a team redesign how they prepare a client brief now that AI drafts the first version. One is a training session, while the other is a facilitated conversation about what judgement, quality control, and decision-making look like when the routine parts of a role are handled by AI technology. The organisations pulling ahead are the ones investing in teaching their people how to use the AI software they’re implementing.

Danger 3 – Ignoring what AI replaces in L&D

Jobs and Skills Australia’s Gen AI Capacity Study is direct about the pattern: augmentation outweighs automation across most roles, but the higher need for automation is concentrated in routine administrative tasks. Entry-level positions built on standardised, repetitive work could also be affected.

For L&D, that finding reads uncomfortably close to home. Content drafting, LMS administration, compliance module production, and assessment generation are routine production tasks, and they’re exactly where AI is most capable of operating without human involvement. JSA also found that a significant portion of employees are already using generative AI independently and without employer oversight.

But AI isn’t replacing the parts of L&D that drive results. The capabilities growing in demand are human-centric: designing programmes that change behaviour, building judgement and decision-making, and working through complex scenarios together. These are the skills that get more valuable as AI handles the routine. AI capability isn’t a threat to the function, it’s something every L&D professional needs to learn how to incorporate into the way they design, deliver, and measure employee learning.

AI is collapsing the distance between knowing and doing

For decades, juniors learned context by doing, the associate who formatted the board pack absorbed how governance worked, the analyst who manually pulled data learned what the numbers meant, and executing low-level tasks was the training ground for higher-order judgement.

AI collapses that pathway. A junior employee can now produce a polished strategy brief in minutes, the output looks senior, but the understanding behind it may not be there. As Chirag Joshi told the AICD: ‘AI has helped us lift the floor and lift the ceiling too. However, it also risks removing the stairs.’

Entry-level roles in AI-exposed fields are now seven times more likely to demand judgement, stakeholder management, and strategic thinking. Training can no longer assume task execution will naturally build these capabilities, because AI is removing the tasks.

This changes what L&D is for. Employees need to be trained as ‘human-at-the-helm’ gatekeepers, accountable for controls, output accuracy, and alignment with organisational values. That’s a capability programme built through facilitated training and monitoring employee performance.

What the strategic L&D function looks like from here

The gap here isn’t a lack of skills, it’s not knowing how to design effectively. L&D’s job is to design the conditions where capability develops. That means shifting from static skills maps to adaptability, from individual proficiency to team-level performance, and from front-loaded training to continuous, supported development in the flow of work.

AITD identifies four shifts that define what strategic L&D looks like now:

• From skills to adaptability. Static skills maps can’t keep pace. Capability depends on how people interpret context, apply judgement, and adapt in unfamiliar situations.
• Human strengths as the differentiator. As AI handles the procedural and analytical work, value increasingly sits in judgement, ethical reasoning, contextual intelligence, and influence; capabilities that can’t be automated.
• From individuals to teams. Performance now depends on how well human capability and AI contributions integrate at the team level, not on individual proficiency alone.
• Readiness as a continuous state. Past qualifications and experience no longer predict performance in fast-changing environments. Readiness must be built through ongoing supported exposure.

The Institute of Managers and Leaders reinforces this from the management side. AI isn’t eliminating managers, Jobs and Skills Australia projects management roles will grow 14.9% to 2035, but it is reshaping what management demands. Seven out of ten Australian employees felt their employer wasn’t adequately preparing them for an AI-driven future, and only 17% had received any AI training at all. When AI handles the routine, the remaining work is harder: higher-stakes decisions, ambiguous situations, and ethical grey areas.

This means strategic L&D functions need to design programmes and learning experiences where people can build the judgement, adaptability, and collaborative capability that AI makes more valuable.

What MCI is seeing in the market

The organisations asking MCI for help are asking questions like what does our workforce need to be able to do differently, and how do we build that at speed? The shift in the conversation, from platform selection to capability design, is the clearest signal that the L&D function is starting to move. If you’re rethinking what your L&D function needs to look like in this shift or working out which parts of your current approach AI should and shouldn’t touch, we welcome the conversation.

Frequently asked questions

Does AI in workplace training raise privacy concerns for Australian organisations?

Yes. The OAIC’s October 2024 guidance clarifies that feeding personal data into AI tools, including employee learning records and performance data, is considered a high privacy risk activity. Some Australian organisations still lack AI-use policies for HR, so this is a governance gap worth closing early.

Where should a small L&D team start with AI?

Start with auditing where your team’s time goes. If most of it is content administration and compliance reporting, that’s where integrating AI saves time. Redirect the freed capacity toward work AI isn’t able to do, such as scoping learning objectives, designing for application, and having conversations that build judgement.

How do you measure where AI is improving L&D outcomes?

This shouldn’t be measured by completion rates. The measures that have impact are behavioural, such as people applying what they learned over a long term period. This could be monitored over 30, 60, and 90 days to see if the capability gaps are closing or persistent. If AI is making your functions faster but the numbers haven’t moved, you’re measuring production over impact.

Will AI replace L&D professionals?

AI is already having an impact on roles involving content drafting, LMS administration, and assessment generation. Strategic roles are growing, capability diagnosis, behaviour-change design, instructional design that connects learning to business outcomes, and designing how humans and AI work together. L&D professionals who move from content production to capability architecture will be more valuable.