No critical function of the organization is still managed with the tools and methods it used twenty-five years ago. None, except one: the design of training plans. In the very same steering committee where geographic expansion decisions are made using predictive artificial intelligence, where sales forecasting is optimized with machine learning models, and where internal fraud is analyzed with advanced analytics, next year’s training plan is still being built today with spreadsheets, in-person interviews, PDF forms, and decisions based on the HR team’s intuition. The paradox can’t hold much longer.
At CAE we have spent more than forty-five years accompanying large companies, consultancies, and universities in building their training programs. We have seen every technological wave the sector has produced: mass in-person training, early e-learning, the corporate LMS, the cross-functional catalog, microlearning, gamification. None of them has transformed the very nature of training design as deeply as artificial intelligence applied to needs analysis and automatic generation of personalized learning paths is doing right now. This article is a structured explanation of why the previous paradigm is exhausted, what ultra-personalization in corporate training really means, and how the transition materializes.
The structural problem: corporate training is still designed as it was in 1995
When an HR leader at a large company faces the design of the annual training plan, the typical workflow doesn’t differ substantially from the one applied twenty-five years ago. Needs identified by managers are gathered, cross-referenced with performance evaluations, checked against the organization’s strategic objectives, a skills matrix is built in a spreadsheet, courses from the available catalog are assigned based on expert approximation, and the plan is launched. The process can take weeks or even months in a mid-sized organization, and months in a large corporation with thousands of employees.
This workflow has four structural problems that the organization accepts as an unavoidable cost, without ever questioning them. The first is that it produces generic plans: when personalization carries such a high cost in HR team work-hours, the response is to segment by job families rather than by individual. The second is that course assignment is approximate: the person deciding what training suits each employee cannot possibly know the full training catalog or every course available in detail. The third is that the plan is static: it is defined once a year and isn’t adjusted until the next cycle. The fourth is that traceability back to business objectives gets lost along the way: the skills matrix turns into a list of courses that is rarely connected to subsequent operational KPIs.
These four problems aren’t questionable in themselves. They are the logical consequence of the method used. With spreadsheets, interviews, and expert decisions, there is no humanly possible way to produce a personalized plan with optimal content assignment, updated in real time, and traceable all the way to business results. Artificial intelligence changes this equation radically for the first time.
The three hidden costs of manual analysis that no organization measures
The usual perception of the cost of manual analysis is limited to the time spent by the HR team. This is a significant miscalculation. The real cost includes three dimensions that rarely show up in internal reports.
First hidden cost: senior talent’s time
Identifying training needs and designing the annual plan are tasks that absorb a significant share of the time of senior HR profiles — precisely the people who would contribute the most to the organization through other, higher-value strategic work. When an L&D lead spends 40% of their time on administrative and coordination tasks, the organization is paying a senior salary for work that can be automated.
Second hidden cost: training misallocation
Analyses published by leading sector consultancies such as Josh Bersin Research indicate that a very high proportion of corporate training is spent on skills the employee already has or doesn’t need for their actual role. In other words, the loss doesn’t only come from the forgetting curve: it happens at the source, through incorrect assignment. In organizations with more than a thousand employees, this cost is substantial and systematically invisible.
Third hidden cost: misalignment with strategy
The Future of Jobs Report from the World Economic Forum documents the speed at which critical skills are changing across virtually every sector. A training plan designed with manual, annual methods is already outdated by the time it launches. The misalignment between the training delivered and the skills that are actually critical for the business at any given moment is probably the most expensive hidden cost of all, and also the hardest to quantify directly.
What ultra-personalization in corporate training really means
The term personalization has been so worn down by training marketing that it’s worth reclaiming with precision. Personalizing training doesn’t mean putting the employee’s name on the welcome screen. It doesn’t even mean offering a menu of courses to choose from. It means the training system determines, for each employee, which skills they need to develop based on their current role, their specific gaps, their career goals, and the organization’s strategic priorities, and translates that analysis into a concrete learning path.
When that process is done for one or two employees, it’s called coaching or mentoring, and it’s expensive. When it’s done for thousands of employees simultaneously, with individual precision and continuous updating, it’s called ultra-personalization. And until very recently, it simply wasn’t technically feasible.
Ultra-personalization has three conditions that set it apart from earlier versions. The first is genuine individualization: each employee receives a distinct plan, not a minor variant of a common one. The second is scalability: individualization holds up whether there are ten employees or ten thousand. The third is continuous updating: the plan adjusts automatically as the employee’s skills, the role’s requirements, or strategic priorities change.
These three conditions are what artificial intelligence makes possible to fulfill simultaneously. No manual method, however sophisticated, can do it. Analyses from strategic consultancies such as McKinsey highlight that large-scale ultra-personalization is one of the AI applications with the strongest demonstrated return in the talent function, and no organization that has experienced applying it wants to go back to the previous method.
The transformation happens in three leaps, not one
It’s a mistake to assume that moving from manual analysis to AI-driven ultra-personalization is simply a change of tool. It is a complete transformation of the process, unfolding across three consecutive leaps. Understanding this sequence helps in planning the transition and avoiding the frustration of organizations expecting immediate results where, in reality, a structured progression is required.
First leap: automating data collection
The first step is replacing interviews, forms, and spreadsheets with the automated capture of data on each employee’s actual skills. This is achieved by combining specific assessments, analysis of the employee’s digital footprint across corporate tools, and historical performance and prior training data. In this leap, the organization stops depending on a manager’s memory and availability to have accurate information about its workforce’s skills.
Second leap: automated decision-making on the learning path
With accurate data available, artificial intelligence can decide what training each employee needs with far greater precision than a manual expert analysis. This isn’t about replacing human judgment in strategic decisions, but about freeing senior profiles from repetitive operational decisions so they can focus on the decisions that genuinely require judgment.
Third leap: automated execution and continuous updating
The third leap is the one that truly makes the difference. Once the training decision is automated, the organization can afford continuous updating of each employee’s training plan, adjusting it in real time as their skills or the role’s requirements change. This turns the training plan from an annual document into a living capability of the organization.
The five requirements an AI system must meet to deliver genuine ultra-personalization
As with artificial intelligence in any application, it isn’t enough to simply announce that it’s being used. These are the five technical and methodological requirements that separate a system that delivers real ultra-personalization from one that only claims to in its marketing copy.
First requirement: access to a broad, up-to-date catalog
A ultra-personalization system can only recommend training that already exists. If the available catalog is limited, personalization degrades quickly. That’s why systems that work rely on training catalogs that are broad, continuously updated, and capable of rapidly incorporating new content. The LearningHub CAE catalog, with more than eight thousand five hundred courses and constantly growing, is an example of the scale of content that ultra-personalization demands.
Second requirement: a semantic model of skills
The AI needs to understand the relationship between the employee’s skills and the training content available. This requires a rigorously built semantic model of skills, not a simple list of topics. Without this model, recommendations are superficial and often incorrect.
Third requirement: quality data on the employee and the role
Input data determines the quality of the recommendations. A system fed with incomplete, outdated, or low-quality data will produce mediocre recommendations. An organization committing to ultra-personalization must also commit to building and maintaining a solid layer of talent data.
Fourth requirement: the ability to learn from real usage
A good ultra-personalization system learns from actual usage: which recommendations are accepted, which are rejected, and what results they produce. Without this continuous improvement loop, the system stagnates. With it, recommendations get better every month.
Fifth requirement: clear human governance
Although AI automates most of the process, strategic decisions — which skills are priorities, which budget goes to which areas, which groups require special treatment — must remain human. A system without clear governance produces recommendations misaligned with strategy and eventually loses the trust of the leadership team.
The Imagine case within the CAE ecosystem
The five requirements above aren’t theoretical: they describe the conditions we applied when building Imagine, our AI-powered solution for automatically creating training plans. Imagine analyzes individual skills, detects the training gaps between what the employee can do and what the role requires, and generates personalized learning paths in seconds, automatically incorporating courses from the catalog, as well as any proprietary or official content the organization wants to include.
The transition from manual analysis to ultra-personalization is a change of method, not just of tool. At CAE we accompany that change with the same methodological rigor we apply to traditional training design: understanding the client’s context, precisely defining the personalization criteria to be applied, building the necessary layer of talent data, and activating Imagine with human support until the organization internalizes the new paradigm.
Frequently Asked Questions
Is it really unviable to keep using spreadsheets for training analysis?
In small organizations, it can still be enough. In medium and large organizations, the limitation shows up quickly: the level of personalization achievable is insufficient for the current pace of skills change. The cost of maintaining the manual method grows exponentially with workforce size.
Does AI-driven ultra-personalization replace the HR team?
No. It frees HR from mechanical, repetitive tasks so it can focus on higher-value strategic work: supporting managers, designing leadership pathways, managing critical talent, shaping corporate culture. The team’s role is elevated, not eliminated.
How long does it take an organization to move from manual analysis to ultra-personalization?
Technical activation can be completed in a matter of weeks, but the cultural change and internalization of the new method take between three and six months. Organizations that approach the transition as a change of method, not just a change of tool, achieve sustainable results.
Is ultra-personalization compatible with subsidized training such as FUNDAE?
Fully. Automatically generated learning paths can qualify as subsidizable training actions if they meet the requirements for duration, SCORM or xAPI traceability, and assessment. Subsidy management is actually simpler, since the system automatically maintains full traceability.
Can ultra-personalization also be applied to universities and educational institutions?
Yes, and it’s probably one of the areas with the greatest potential. Generating personalized learning paths per student makes it possible to complement formal education with tailored, real-time plans.
Conclusion: the end of manual analysis isn’t a threat, it’s a liberation
The manual-needs-analysis paradigm is exhausted. Not because it was ever bad — for decades it was the best method available — but because the conditions that made it the best method have disappeared. The speed at which critical skills change, the size of workforces, the availability of talent data, and the maturity of artificial intelligence applied to training decisions have made possible something that was technically unviable until very recently: genuine, large-scale ultra-personalization of corporate training.
Organizations approaching this transition thoughtfully are discovering two things at once. The first is that their HR teams feel freed from mechanical work and find time for tasks that bring real value to the organization. The second is that training impact rises significantly, because each employee receives exactly what they need, at the moment they need it. Both benefits happen together, not as a trade-off.
At CAE, after forty-five years, we remain convinced that pedagogy is the heart of any serious training solution. Artificial intelligence applied to ultra-personalization doesn’t change that conviction — it amplifies it. Because when each employee receives the right training at the right time, the pedagogical potential of that training is fully unleashed. And that, ultimately, is what we’ve been pursuing for decades.
Want to know how Imagine could transform the training-design process in your organization? Get in touch with our team for an initial conversation about your specific context.