![\"\"](\"https:\/\/solutionsreview.com\/edtech\/files\/2026\/08\/AdobeStock_368278340-768x384.jpeg\")
<\/p>\n<\/p>\n
MyPassion.ai’s’s Marco Kohns offers this commentary on the direction gap and how AI made learning content cheap while guidance remains the scarce good. <\/strong><\/em>This article originally appeared in Insight Jam<\/a>, an enterprise IT community that enables human conversation on AI.<\/span><\/strong><\/em><\/p>\n In June 2026, The Conference Board reported that 68.9 percent of US workers are satisfied with their jobs. That is the highest reading in the survey’s 39-year history, up from 68.3 percent a year earlier and a long climb out of the 42.6 percent low of 2010. It is good news, and for anyone evaluating learning technology it is close to useless. A satisfaction rate counts heads. It says nothing about what the other three in 10 need, whether more training would help them, or whether the platform you are about to sign for addresses any of it.<\/p>\n That gap matters more this year than last, because artificial intelligence (AI) has moved the scarce part of the learning stack.<\/p>\n Generation Solved the Wrong Bottleneck<\/strong><\/p>\n Nearly every education technology (EdTech) product announcement I have read this year points at the same layer: generation. Course outlines from a prompt, adaptive explanations, instant translation, a tutor that answers at 11pm, assessments written in seconds. The work is good and it has pushed the marginal cost of producing a competent lesson toward zero.<\/p>\n When producing content costs almost nothing, content stops being the thing you are buying. The bottleneck moves one step earlier, to the decision sitting in front of the catalog: which of these thousand available things should this specific person do next, and on what evidence.<\/p>\n Our own data says that decision, rather than access, is where adult learners are stuck.<\/p>\n The Only Content That Survives Abundance Is Content Nobody Else Can Produce<\/strong><\/p>\n I have run this experiment on my own site. We had published more than 60 blog posts built the way everyone builds them: keyword research, competent structure, a competitor’s page open in the next tab. Not one of them carried a number that did not already exist somewhere else on the internet. They performed accordingly.<\/p>\n What changed the response was not better writing. It was publishing our own response data, with the base stated for every percentage, the methodology written out, the caveats included, and an open license inviting anyone to reuse it. That material is the only content we have produced that other people ask to reuse, and the reason is not craft. Nobody else can regenerate it. A model can rewrite a listicle in seconds. It cannot produce the answers 6,837 people gave to a question we asked.<\/p>\n This is the pattern behind everything else in this piece, and it cuts both ways across an EdTech purchase. If your learning content can be reproduced by anyone holding the same model subscription, your learners will value it at roughly what it cost to make, which is close to nothing. If the vendor you are evaluating is a wrapper over a general model and a licensed content library, the same is true of theirs.<\/p>\n What People Questioning Their Careers Say Is Blocking Them<\/strong><\/p>\n Between January 29 and July 28, 2026, 6,837 people completed the\u00a0free career quiz<\/a>\u00a0we run at\u00a0MyPassion.ai<\/a>, from 156 countries. This is a self-selected sample of people actively questioning their direction, not a representative population panel. Every figure below states the base of respondents who answered that specific question, and none of it supports an economy-wide claim. It describes the people inside the question rather than the average of everyone.<\/p>\n Asked to name their core career struggle, 52.9 percent of 3,656 respondents chose a self-knowledge problem. Of those respondents, 26.5 percent said they have no idea which direction to pick, and 26.4 percent said they cannot tell what they are good at. Only 20.5 percent reported the opposite problem of having too many interests to choose between.<\/p>\n That split has a direct consequence for buyers. A catalog, a skills taxonomy, and a recommendation engine trained on job postings all address supply. For the majority here, supply is not the constraint. What is missing is information about themselves, and no volume of available courses resolves it.<\/p>\n The pattern holds where you would least expect it. Of 6,515 respondents describing their situation, 19.1 percent said they are in a well-paying career and want to change anyway. Among that group, 59.2 percent have eight or more years of experience (n = 1,035). Dissatisfaction concentrates at seniority, among people who already have money, access, and a track record.<\/p>\n The Finding That Should Concern Anyone Selling Learning<\/strong><\/p>\n Respondents pick one core struggle from a fixed set. One option reads “I’m stuck or bored in what I do now.” The share choosing it rises at every step of the education ladder: 6.4 percent of high-school respondents, 14.9 percent of bachelor’s graduates, and 20.1 percent of those holding a master’s degree or higher. The bases are 1,553, 1,078, and 568 respectively.<\/p>\n This is observational data from one point in time and it cannot establish that education causes boredom. Several readings are available. More education tends to buy work with more autonomy, which should reduce boredom, and it also raises what a person expects work to give them. What the gradient does establish is that the group with the most formal learning behind them is the group most likely to describe their current work as empty. Credentials buy access reliably. On this evidence they do not buy engagement.<\/p>\n Set that next to what the same respondents name as their strongest source of absorption. Asked what most reliably puts them in flow, 29.9 percent of 2,330 chose learning, going deep on a topic that grips them, ahead of creating at 22.4 percent and being with people at 21.8 percent. The people most likely to be bored are the ones who spent the longest in environments built around learning, and learning is the most commonly named source of absorption. Roles that stop teaching people anything appear to be the roles that lose them.<\/p>\n For an EdTech buyer, that is the opportunity and the warning in the same sentence. Appetite for learning is not the problem. Direction is.<\/p>\n Completion Is a Relevance Problem<\/strong><\/p>\n One operator number is worth more to a buyer than any of the above. Our questionnaire takes a median of 6.9 minutes and 75.4 percent of people who start it finish every question. There is no video, no gamification, and no reward. It holds attention because the output is about the person answering.<\/p>\n Anyone who has watched a corporate learning module get abandoned at 30 percent knows the standard diagnosis: the content was too long, too dry, or badly produced. That diagnosis sends buyers shopping for better production. The likelier explanation is that the learner could not connect the module to anything about their own situation. Engagement is downstream of relevance, and relevance is downstream of knowing who the person is.<\/p>\n What a Direction Layer Does<\/strong><\/p>\n Four capabilities separate a direction layer from a catalog with better manners.<\/p>\n It infers rather than asks.<\/strong>\u00a0Self-report has a ceiling, and half our respondents told us plainly that they cannot describe what they are good at. A system that only reorganizes what a learner typed into an intake form inherits that ceiling. Ask a vendor which inputs the system uses that the learner never entered, and what it does with them.<\/p>\n It shows its reasoning and invites disagreement.<\/strong>\u00a0A recommendation a learner cannot interrogate is a recommendation they will abandon quietly. The output should carry the evidence behind it, in language the learner can push back on.<\/p>\n It terminates in an action, not a course.<\/strong>\u00a0A recommendation whose final step is “enroll here” is a sales funnel. A direction layer ends in a role to explore, an internal move to apply for, a project to volunteer for, or a conversation to book, with the learning attached to that move rather than standing in for it.<\/p>\n It stands on data a competitor cannot regenerate.<\/strong>\u00a0Ask what the system knows that is not available to every buyer with the same model subscription: outcome records from prior cohorts, longitudinal data on who moved where and whether the move held, assessment responses at scale. This is the question that separates a product from a prompt, and it is the one most likely to go unasked in a demo.<\/p>\n How to Pilot This in 90 Days<\/strong><\/p>\n Define the outcome metric before you take the first demo. Completion, seat time, and satisfaction scores will all improve under a competent vendor and none of them tell you whether anyone found direction. Better candidates: the share of participants who make a documented next-step decision within 30 days, internal mobility applications filed, and persistence in the chosen path at 90 days.<\/p>\n Pick the cohort most likely to be stuck rather than the easiest one to recruit. On our data that is not new hires. It is people with eight or more years of experience who are performing well and quietly questioning the whole thing.<\/p>\n Run a control arm. Give a comparable group the existing catalog plus one structured conversation with a manager or a counselor. If the software cannot beat a good conversation, you have learned something worth more than the pilot cost.<\/p>\n The Failure Mode<\/strong><\/p>\n The risk in this category is not that AI gives learners no direction. It is that it gives them confident direction with nothing underneath. A model will produce a plausible career recommendation for any input you hand it, at any level of evidence, and a person will act on it. That is a heavier consequence than a badly recommended course.<\/p>\n So make this the last question in every evaluation: what does the system do when it does not know? A product that can say “there is not enough here to advise you yet, answer these three things” is telling you it has a confidence threshold. A product that always has an answer is telling you it has none.<\/p>\n Content was the constraint for 20 years, and AI has removed it. Whoever buys as though it were still the constraint will end up with a faster way to deliver material that nobody needed in the first place.<\/p>\n","protected":false},"excerpt":{"rendered":" MyPassion.ai’s’s Marco Kohns offers this commentary on the direction gap and how AI made learning content cheap while guidance remains the scarce good. This article originally appeared in Insight Jam, an enterprise IT community that enables human conversation on AI. In June 2026, The Conference Board reported that 68.9 percent of US workers are satisfied […]<\/p>\n","protected":false},"author":1484,"featured_media":536,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"inline_featured_image":false,"footnotes":""},"categories":[14],"tags":[75,74],"acf":[],"yoast_head":"\n![\"\"](\"https:\/\/solutionsreview.com\/security-information-event-management\/files\/2025\/03\/Insight-Jam-Logo-2025-Square.png\"){kind=logo}
<\/a><\/p>\n