Why Most Corporate AI Training Programmes Are a Waste of Money
Quick answer: Why do most corporate AI training programmes fail? Because they teach people what AI is rather than how to use it in the specific context of their work. Generic digital literacy training produces the appearance of AI capability without reality. Effective AI upskilling is function-specific, use-case specific, and tied to actual workflow change not a module people complete and forget.
The corporate AI training market is booming. Every major consulting firm, every HR technology platform, and every corporate university is selling something. Completion rates are being reported. Certifications are being issued. Leadership teams are ticking boxes and feeling like they've addressed the upskilling problem.
Most of it is not working.
That's a blunt assessment, and I want to be precise about what I mean. The problem isn't that the content is wrong. Most AI literacy training is technically accurate. The problem is that it's answering questions nobody on the frontline is actually asking and it's not answering the questions they are asking.
After working with organisations across financial services, healthcare, retail, and government on their AI adoption challenges and after delivering the Upskilling Our People with Agentic AI keynote to leadership teams navigating exactly this the pattern is consistent. The organisations making real progress on AI capability are doing something fundamentally different from what most corporate training programmes are doing.
What most corporate AI training actually delivers
The typical corporate AI training programme looks something like this: a series of modules covering what generative AI is, how large language models work, the history of AI, a brief overview of tools the company has licensed, some guidance on responsible AI use, and a certificate at the end.
The people who complete it walk away understanding AI better as a concept. What they don't walk away with is any meaningful ability to use AI tools differently in the specific context of their actual job. The gap between 'understanding AI' and 'using AI to do my job better' is enormous and most training programmes never cross it.
The gap between 'understanding AI' and 'using AI to do my job better' is enormous. Most training programmes teach the first and assume the second follows. It doesn't.
The specific reasons training programmes fail
They're designed for HR, not for learners
Corporate training programmes exist partly because organisations genuinely want to build capability and partly because leadership teams need to demonstrate to boards and investors that they're addressing the AI readiness question. The second motivation produces training that looks good in reports and in board presentations, high completion rates, nice dashboards, certifications with the company logo but doesn't produce behaviour change.
Training designed for HR is optimised for completion. Training designed for learners is optimised for application. Those are very different design problems, and most corporate training programmes are solving the wrong one.
They're generic when they need to be specific
An AI literacy module for a procurement team and an AI literacy module for a marketing team are often the same module. But the AI tools, the use cases, the specific workflows being changed, and the specific judgments being automated are completely different. Generic training that doesn't speak to the specific context of a specific role in a specific function doesn't produce the 'how does this change what I do tomorrow morning?' insight that actually drives behaviour change.
The organisations that are seeing real capability gains are running function-specific training built around the actual workflows, the actual tools deployed in that function, and the actual questions that function's professionals are facing. That training is harder to build and can't be sold at scale to every company in every industry. Which is why most vendors don't offer it.
They treat AI literacy as a knowledge problem, not a behaviour problem
Knowing that AI can summarise documents doesn't mean you'll start using it to summarise documents. Knowing that AI can help structure analysis doesn't mean your team will restructure their analytical workflows. Knowledge is necessary but insufficient for behaviour change. Behaviour change requires practice, feedback, and most importantly permission.
The permission piece is underappreciated. Many employees who have completed AI training are still not using AI tools in their work because the signals from their organisation about whether AI-assisted outputs are acceptable, whether using AI is a sign of laziness, and whether their manager will value or discount AI-augmented work are ambiguous or negative. Training that ignores the cultural and managerial context produces learned information that doesn't translate into changed behaviour.
They focus on current tools rather than durable principles
AI tools are changing faster than training can track. A programme built around a specific tool deployed six months ago may already be teaching workflows that the current version of the tool makes obsolete. Worse, it may be teaching workarounds for limitations that no longer exist, or missing capabilities that have been added since the curriculum was written.
Effective AI upskilling builds capability at a level of abstraction that doesn't become obsolete with the next model release: how to evaluate AI output critically, how to identify when AI is confidently wrong, how to structure prompts that produce genuinely useful results, how to know which tasks are appropriate to delegate to AI and which require human judgment. These principles are durable in ways that specific tool tutorials are not.
What actually works
Use-case-first design
Start with a specific task that a specific role performs regularly, and build the training around how AI changes that task. Not 'here's what AI can do' 'here's how your quarterly business review preparation changes when you use AI to pull and synthesise the data, and here's how you validate that output before putting your name on it.'
This requires interviews with the actual people doing the work. It requires access to the real workflows, the real tools, and the real constraints. It's slower and more expensive to build than generic training. It also produces results.
Manager capability as a prerequisite
The single most reliable predictor of whether team members will change their behaviour after AI training is whether their manager has changed theirs. If a manager has not meaningfully integrated AI into their own work, they will consciously or not signal to their team that AI integration is optional, experimental, or not really the priority. Training programmes that don't include a specific component for managers are building on a foundation that will undermine their own results.
Measurement tied to workflow change, not completion
If the metric for your AI training programme is completion rate and certification count, that's what you'll get. If the metric is 'what percentage of the finance team is using AI to draft first-pass variance analysis' or 'what is the average time from brief to first draft in the content team before and after training,' those are the outcomes you can actually learn from. They're also harder to measure which is why most organisations don't measure them.
Iteration, not events
A two-day AI bootcamp and a certification are training events. They produce a bump in awareness and, sometimes, a temporary change in behaviour. Genuine capability development is a continuous process of new tools, new use cases, new judgments as AI improves and as the organisation's deployment of AI matures. The organisations making real progress treat AI upskilling as an ongoing operating expense, not a one-time project.
The leadership communication problem underneath the training problem
Something I've observed consistently across organisations: the reason AI training programmes don't produce behaviour change is often not the training itself. It's the absence of a clear, credible, function-specific answer to the question employees are actually asking: 'What specifically should I be doing differently, and what happens to my role if I do?'
Generic training can't answer that question. Only a manager who understands both the AI tools and the specific work of their team can answer it. Which is why investing in manager capability and investing in giving managers the specific, honest answers to those questions almost always outperforms investing in more training modules.
This connects to a theme I cover extensively in the Culture in an AI World keynote why culture is the variable that determines whether AI training produces capability or just completion rates.
The question employees are actually asking is not 'what is AI?' It's 'what specifically should I be doing differently, and what happens to my role if I do?' Generic training cannot answer that.
What organisations should do instead
Audit actual workflows before designing any training understand what work AI could change before building curriculum.
Identify the three to five use cases per function where AI would produce the most value, and build training specifically around those.
Include managers explicitly train them first, and make their adoption visible.
Create explicit permission structures to make clear that AI-assisted outputs are acceptable and explain how quality is maintained.
Measure behaviour change, not completion track actual workflow change in the months after training.
Treat upskilling as ongoing, not as an event budget for continuous learning, not a one-time programme.
For leadership teams navigating this challenge: the Upskilling Our People with Agentic AI keynote is built around exactly this, what actually works for workforce AI readiness, grounded in what organisations across sectors are doing. You can also explore the full range of keynote topics or book a conversation to discuss what your audience specifically needs.
Frequently asked questions
Why do corporate AI training programmes have low impact?
Most corporate AI training programmes have low impact because they're generic, knowledge-focused, and disconnected from the specific workflows and real decisions of the people going through them. Knowing what AI is doesn't change how you work. Only training that addresses specific use cases in the specific context of a specific role combined with clear managerial permission to use AI and explicit expectations about quality produces genuine behaviour change.
What makes an AI upskilling programme effective?
Effective AI upskilling programmes share four characteristics: they're use-case specific rather than generic, they include managers explicitly rather than treating management as separate from the training problem, they measure workflow change rather than completion rates, and they treat upskilling as a continuous process rather than a one-time event. The organisations seeing real capability gains have all four of these elements. Most programmes have none of them.
How should organisations measure the ROI of AI training?
Stop measuring completion rates and certification counts. Those measure training activity, not capability development. Measure workflow change: time-to-first-draft before and after, percentage of team members using AI tools in specific workflows, quality of AI-augmented outputs compared to unaugmented outputs, and most directly business outcomes in the functions where AI has been most deliberately deployed. These are harder to measure. They're also what actually matters.
Should AI training be mandatory in organisations?
Mandatory training is appropriate for foundational AI literacy and for specific compliance requirements around AI data governance and responsible AI use. For capability development, actually learning to use AI tools effectively mandatory training tends to produce compliance rather than engagement. The organisations seeing the best results are making foundational training mandatory and advanced capability development compelling: people want to participate because they can see the value to their own work.
What role should managers play in AI upskilling?
Managers are the single most important variable in whether AI training produces behaviour change. If managers haven't meaningfully integrated AI into their own work, they signal consciously or not that adoption is optional. If managers are actively using AI tools and discussing how they're changing their own work, that signal propagates through the team. Invest in manager capability first, and make that adoption visible and specific, not just aspirational.