You know the situation. You\u2019ve designed a great training programme. The participants were enthusiastic, the evaluation forms looked promising, and the trainer had clearly put in a lot of effort. But three months later, you\u2019re sitting opposite the director, who asks: \u201cWhat has this actually achieved?\u201d And you realise your answer isn\u2019t really convincing. <\/p>\n
This is the central problem facing L&D in 2026. Not the lack of impressive programmes and annual plans. Not the lack of engaged employees. But the lack of an answer to the question that is relevant today: has anything changed in how people behave in the workplace? And if so, how do you demonstrate that? <\/p>\n
Proving the ROI of soft skills training is a complex part of the job for many L&D professionals. And that is precisely the problem PractAIce is trying to solve; not by building better training programmes, but by making behavioural change visible and measurable.<\/p>\n
Most training evaluations stop at what Kirkpatrick calls the first or second level: how satisfied were participants, and what did they learn? These are useful indicators, but they say little about what actually changes in the workplace. According to data from the American Society for Training and Development, Kirkpatrick Level 3 \u2014 behavioural change \u2014 is measured in only 25% of cases, and Level 4 \u2014 organisational results \u2014 in just 15%<\/a>. <\/p>\n This is not because L&D professionals do not want to measure it. It is because it is difficult. Behavioural change in soft skills \u2014 for example, how someone gives feedback, how someone deals with resistance, or whether someone remains calm during a difficult conversation \u2014 is by definition elusive if you do not have a structured way of observing it. <\/p>\n And yet it is precisely this data that senior management is asking for. According to research by the Brandon Hall Group, 57% of L&D leaders are experiencing increasing pressure to demonstrate ROI, and that pressure has only grown in recent years<\/a>. At the same time, 75% of organisations say that better aligning L&D with business strategy is their top priority for 2025. The gap between what L&D delivers and what the business needs lies not in the quality of the training but in the measurability of the results.<\/p>\n There is a reason why management training and soft skills training are so often dismissed as \u2018nice to have\u2019: the effects are visible in behaviour, not in figures. And behaviour is difficult to quantify. Especially if the measurement takes place only afterwards, separate from the training itself. <\/p>\n The Kirkpatrick model provides a clear framework for this. Level 3 of the model measures whether participants actually apply what they have learnt in their work \u2014 and that is, as Kirkpatrick himself states, the first level at which you can truly say anything about the effectiveness of a training programme<\/a>. Everything prior to that, such as satisfaction scores and knowledge tests, is internal evidence for the training function itself. What the business needs starts at Level 3. <\/p>\n But how do you obtain Level 3 data for soft skills? Traditionally through manager observations, 360-degree feedback or performance reviews. These methods have value, but they are time-consuming, subjective and come too late, often weeks or months after the training has taken place. By that time, the link between what was learnt and the measured behaviour has often faded. <\/p>\n That is the gap. And it is a structural gap, not one you can close with a better satisfaction survey.<\/p>\n PractAIce works differently from traditional training, and that difference is precisely what matters when it comes to ROI. Every time an employee completes an AI role-play, the platform generates detailed behavioural data: how did someone score on active listening? How was feedback structured? How did someone respond to an escalation in the conversation? Or how effectively does someone handle conflict resolution?<\/p>\n That data isn\u2019t based on self-reporting or a manager\u2019s impression. It\u2019s based on what someone actually did in the simulation, how often, and whether that improved over multiple sessions. That\u2019s Kirkpatrick Level 3 data: not added retrospectively, but incorporated into the platform\u2019s design from the outset. <\/p>\n For L&D professionals, this makes a difference to how you approach boardroom discussions. Instead of \u201c95% of participants rated the training as good or very good\u201d, you can say: \u201cEmployees scored an average of 54% on constructive feedback skills at the start; after four practice sessions, that was 71%, with a clear curve in individual progress.\u201d That is a conversation you can have with the director.<\/p>\n One of the most consistent findings in training research is that a single learning session is not enough to bring about behavioural change. People learn something, return to the workplace, and the pressures of everyday life pull them back into old patterns. This is not a lack of motivation; it is neurology. <\/p>\n Deliberate practice, the concept developed by psychologist K. Anders Ericsson, shows that expertise does not arise from experience or natural talent, but from focused, repeated practice on specific behavioural aspects, with immediate feedback after each attempt<\/a>. For soft skills, this means: not just one role-play in a group session, but running through the same scenario ten times until the new behaviour becomes automatic.<\/p>\n That is what PractAIce makes possible. Employees can repeat a scenario without relying on a trainer or a colleague. They practise at their own pace, at a time that suits them, and thus build up, step by step, the behavioural repertoire that the training aimed to achieve. This is particularly relevant for management training and leadership development: those skills fade most quickly if there is no structured opportunity for practice after the training.<\/p>\n It starts with a shift in how you design training programmes. Kirkpatrick recommends designing backwards: start at Level 4: what business outcome do you want to achieve? Then work backwards to determine what behaviour is required for that (Level 3), what people need to learn to achieve it (Level 2), and what the training should look like (Level 1). <\/p>\n In practice, for soft skills training this means: determine in advance which competencies you want to develop, make them observable, and build in a way to collect those observations. So not afterwards, but as part of the training design itself. <\/p>\n PractAIce directly supports this approach. L&D professionals can create scenarios for each competency that align with their own organisational context \u2013 in other words, the kind of conversations employees have in their specific roles, with dynamics that are recognisable to their team. These customised role-plays not only deliver a better learning experience, they also generate more relevant behavioural data.<\/p>\n And that is what changes the conversation with the director. Not \u201cwe have trained 120 employees\u201d, but: \u201cwe can demonstrate that the feedback culture in Team X has improved in concrete terms, and here is the data to back that up.\u201d<\/p>\n How do you measure the ROI of soft skills training? What is the Kirkpatrick model and why is Level 3 so important for L&D? How do I convince senior management of the value of soft skills training? Is AI-based management training more effective than classroom-based training? L&D is not an activity centre. It is a strategic function that contributes to how people collaborate, communicate and perform. But that contribution is only visible if you make it measurable. <\/p>\n Proving the ROI of soft skills training is not simply a matter of fancier evaluation forms. It starts with the design: what behaviour do you want to see, how do you make that observable, and how do you build in structured practice so that what has been learnt really sticks? The fact<\/a> that 80% of professionals say that soft skills such as communication are only becoming more important due to AI<\/a> does not make the question any smaller, but it does make it more urgent. <\/p>\nWhy soft skills are so difficult to measure and what that means for your approach<\/h2>\n
How PractAIce generates behavioural data that really makes a difference<\/h2>\n
Soft skills training that sticks: why repetition is key<\/h2>\n
What L&D professionals can do differently<\/h2>\n
Frequently asked questions about measuring the effectiveness of soft skills training<\/h2>\n
<\/strong>The best way to measure the ROI of soft skills training is to define behavioural indicators in advance that you want to see improve. Consider: how often does someone give feedback, how does someone react in conflict situations, how does someone score on active listening in simulated conversations. PractAIce automatically generates this behavioural data for each practice session, enabling pre- and post-assessment without placing an additional observation burden on managers. <\/p>\n
<\/strong>The Kirkpatrick model divides training evaluation into four levels: reaction, learning, behaviour and results. Level 3 (behaviour) is the first level that truly indicates whether training has worked in practice. Satisfaction scores (Level 1) and knowledge tests (Level 2) measure what happened during the training; Level 3 measures what changes on the work floor afterwards. For soft skills in particular, this is the crucial evidence. <\/p>\n
<\/strong>The strongest arguments are concrete and behavioural. Not \u201cthe training was well-received\u201d, but \u201cafter the programme, employees demonstrably scored higher on specific communication competencies, and here is the progression across multiple practice sessions.\u201d PractAIce enables this type of reporting without the need to manually collect observational data. <\/p>\n
<\/strong>Not necessarily more effective in terms of content, but certainly in terms of transfer. The fundamental problem with classroom-based management training is that there is no structured practice environment once the training day is over. AI-driven practice via PractAIce fills that gap exactly: employees can repeat scenarios whenever they need to, with immediate feedback, without relying on a trainer.<\/p>\nFinally: the conversation you want to be able to have<\/h2>\n