AI Training: Impressive Credentials. Zero Production Scars.

The AI training market is full of certified experts who have never taken a project from pilot to production. Boards are paying premium rates to find out the hard way.

Video summary: https://youtu.be/xDTjM5C8FoU

Sign up for my Substack daily AI newsletter here.

See my AI Training course portfolio for corporate Business Leaders here.

Article link: https://open.substack.com/pub/johanosteyn/p/ai-training-impressive-credentials?r=73gqa&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/

The pattern is familiar to every executive who has commissioned AI training for their organisation. A provider is engaged. A programme is delivered. The trainers are credentialled, articulate, and confident. The workshops are well-structured. For a period of weeks, sometimes months, there is genuine energy — people experimenting with tools, sharing prompts, discovering what the technology can do. And then, quietly, the excitement subsides. The tools that seemed transformative in the workshop become optional in the workflow. The use cases that looked compelling in the slides remain stuck in the ideation phase. Nothing has changed operationally. The organisation has spent significant money, and the return on that investment is an interesting couple of weeks and a cohort of employees who are more comfortable with AI than they were before, but no more capable of implementing it than they ever were.

This is not a failure of enthusiasm. It is a failure of the training itself — and the failure has a precise cause. Almost every corporate AI training programme in existence today teaches people how to use AI. Almost none of them teach people what to use it for. And the reason that the second question goes unanswered is not that it is unanswerable. It is that answering it requires the kind of practical, implementation-level judgment that can only be built by having been inside a large organisation, identified a genuine business problem, and navigated the full journey from that problem to a working solution in a production environment. That experience is not something a trainer acquires from a curriculum. It is something they earn from the work — and most corporate AI trainers have never done the work.

CONTEXT AND BACKGROUND

The training market’s failure is not hypothetical. It is documented, quantified, and consistent. A DataCamp survey of more than 500 enterprise leaders in the US and UK, conducted with YouGov in March 2026, found that 82% of organisations provide some form of AI training, yet 59% report a persistent AI skills gap. The training exists. The capability does not materialise. Only 35% of those organisations describe their AI upskilling as mature and organisation-wide. The rest are training people without producing the outcomes that the training was commissioned to deliver.

The EY Survey for 2026 sharpens the picture further: 88% of organisations use AI in the workplace, but only 28% have managed to empower their people to actually transform operations. That gap — between using AI and transforming with AI — is precisely the distance between knowing how and knowing what for. Employees across these organisations are prompting, experimenting, and generating content. They are not identifying the high-value use cases in their specific organisational context, designing the workflow integrations that would make those use cases sustainable, or building the internal capability to maintain what they have built once the training provider has left.

The Josh Bersin Company has described a $400 billion corporate training market being disrupted at a quickening pace — but the disruption is not solving the underlying problem. The market is producing awareness at an unprecedented scale. The gap between that awareness and operational capability is not closing. What is being sold, at premium rates, is familiarity with the technology. What organisations need, and what the market is largely failing to provide, is the judgment to know where that technology belongs in their specific operational context.

MIT researchers reviewed more than 300 publicly disclosed AI implementations in 2025 and found that just 5 per cent of integrated AI pilots generated millions of dollars in measurable value. The most consistently cited cause of failure is not the technology. It is execution — specifically, the absence of the organisational judgment required to identify the right problems, design the right solutions, and manage the transition from a promising pilot to a production system that delivers returns.

INSIGHT AND ANALYSIS

The distinction between how and what for is not subtle, but it is almost universally missed in the design of corporate AI training programmes. Teaching an executive how to use an AI tool means teaching them to prompt, to navigate an interface, to understand what the model can and cannot do. It is useful knowledge and it is the right starting point. But it is only a starting point, and most programmes treat it as a destination.

Teaching an executive what to use AI for means something entirely different. It means helping them map their organisation’s specific workflows, identify where AI would produce the highest return, understand what data is required and whether it is available, anticipate the change management challenges that will arise when the tool meets the reality of the organisation, and build the governance structure that will make the deployment sustainable beyond the initial enthusiasm. That is not a conversation about technology. It is a conversation about the organisation — and it can only be conducted credibly by a trainer who has had it before, in a real organisation, with real consequences for getting it wrong.

The CEO of General Assembly, Daniele Grassi, named the market’s central problem precisely in 2026: comfort does not equal capability. Organisations offering one-size-fits-all AI training end up with people at the top using AI the same way as interns — for email drafting, basic summaries, and prompt experimentation — while the genuine transformation opportunities remain unidentified and unpursued. The training produced familiarity. It did not produce the map from that familiarity to operational value.

That map is what experienced implementers carry into a training room that credentialled theorists do not. An experienced implementer has been in a large organisation when the initial AI excitement subsided, and has understood why it happened. They have seen the pattern: a tool is introduced, enthusiasm builds, and then the first time an employee encounters a genuine operational challenge — a workflow the tool does not fit, a data quality problem the prompt cannot solve, an integration requirement that nobody planned for — the enthusiasm becomes frustration and then abandonment. They know that the antidote to that pattern is not more enthusiasm. It is use case identification that begins with the organisation’s actual problems, not with the technology’s capabilities.

The three most consequential things that experienced implementers know, and that standard training programmes almost never address, flow directly from this distinction. The first is that data readiness determines more about an AI initiative’s success than any other factor, and that the organisations most likely to fail are the ones whose training begins with the tool rather than with an honest assessment of whether the data required to use it is clean, accessible, and governed. RAND research cited in multiple 2026 industry analyses found that 80% of AI projects deliver no measurable business value, and that data problems account for a significant proportion of that failure. Gartner has predicted that 60% of AI projects unsupported by AI-ready data will be abandoned entirely. A trainer who has never conducted a data readiness assessment inside a large corporation cannot teach one.

The second is that the absence of a clear use case definition produces shadow AI — the condition in which employees, finding that the official AI tools do not solve their actual problems, quietly resort to consumer-grade models and type sensitive corporate data into systems the organisation has no governance relationship with and no visibility into. Shadow AI is not a technology problem. It is the direct consequence of training that teaches how, without teaching what for, leaving employees to find their own applications for a capability they have been taught, but not shown how to direct. The regulatory and reputational exposure this creates is significant and growing.

The third is that pilot purgatory — the condition in which an organisation manages dozens of experimental AI use cases that never reach production — is also a direct consequence of the same training failure. Organisations that cannot answer what they should use AI for cannot make the decisions required to take a pilot to production. They run more pilots instead. McKinsey’s 2026 data on AI maturity found that only 1% of corporate leaders describe their AI deployments as mature. The rest are permanently exploring. The trainers who left them there taught them to experiment. Nobody taught them to commit.

IMPLICATIONS

The implication for boards and executives is direct and practical. Before any AI training contract is signed, the evaluation of the provider must go beyond institutional credentials and curriculum comprehensiveness. The questions that determine whether a training provider can answer the what for question are different from the questions that evaluate their grasp of the how.

Ask them to describe a specific organisation whose AI initiative they supported, what operational problem that initiative was designed to solve, and how they helped the organisation identify that problem before the training began. A provider who has done this work will answer with precision. A provider who has not will describe the curriculum.

Ask them how they approach use case identification, and at what point in the engagement that conversation happens. If the answer positions use case identification as something that happens after the participants understand the technology, the provider is teaching in the wrong order. Knowing what to use AI for is the prerequisite for knowing how to use it well. The sequence matters, and experienced implementers know that.

Ask them what they teach about the transition from pilot to production, and specifically about the change management, data governance, and organisational alignment required to make that transition happen. If the answer focuses primarily on the technology, the provider is equipped to produce the initial excitement. They are not equipped to prevent the dwindling.

BCG has found that 70% of AI success is people, process, and change management — not algorithms or infrastructure. The organisations capturing AI’s value are not the ones with the most sophisticated training programmes. They are the ones whose training was designed by people who have navigated the full journey from problem identification through to production deployment, and who teach from the scars that journey leaves rather than from the slides that describe it.

CLOSING TAKEAWAY

The excitement that follows a well-delivered AI training programme is real. So is the dwindling that follows the excitement when the organisation discovers that knowing how to use AI is not the same as knowing where to apply it, or how to make that application stick. The pattern is so consistent across organisations of every size and sector that it has become the expected shape of a corporate AI journey, when it should be the exception.

The providers who break that pattern are not the ones with the most impressive credentials. They are the ones who enter a training engagement by asking what the organisation’s most important unsolved operational problems are, and who design everything that follows around answering that question. They teach the technology in service of the problem. They do not teach the problem in the service of the technology.

That is the difference between a trainer who has learned about AI and one who has implemented it. Before you commission the workshop, ask the question that separates them. Not which university trained you, but what have you built, where did it fail, and what did you learn from it that you now teach?

Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net