# Boris Kehr — Senior Product Designer Stockholm, Sweden · boris@boriskehr.se · https://www.linkedin.com/in/boriskehr > I think in systems, design for outcomes I’m a senior product designer with a passion for education and pedagogy. I connect technology, business and human needs to create sustainable products. ## Selected work ### From zero to 300,000+ learners in AI education (EdAider) Building Sweden’s largest AI learning platform for educators. ### Better experience in job-search (Cassil) Platform helping designers through the application process. ### Researching a broken hiring system (Ghosted) Understanding why job-hunting is so hard and where a product can genuinely help. ## Case study: From zero to 300,000+ — Building the largest AI education platform for teachers was not enough. Role: Head of design, Product strategy, Research, Product design Generative AI reached schools before many teachers had the time, confidence or support to understand it. The conversation moved quickly between excitement and fear, while schools lacked a shared foundation for making informed decisions. EdAider set out to make AI understandable, practical and responsible for educators. Not to turn every teacher into an AI enthusiast, but to give them enough knowledge to decide where the technology was useful and where it was not. ### What I did As Head of Design, I helped build EdAider’s commercial AI education offer from its earliest stage through delivery and continued development. I worked across product direction and hands-on design: - Product strategy and concept development - Research with teachers and school leaders - Learning experience and course structure - Interaction and interface design - Learning content and practical exercises - Design-system development - Alignment across education, technology and business We turned a broad, fast-moving and often intimidating subject into a flexible online course. Short modules used plain language to explain how AI works, its possibilities and risks, and what responsible use could look like in education. Teacher and student perspectives connected the subject to everyday practice. Practical examples gave participants a safe way to experiment. Teachers could study independently, while schools could use the course as a shared professional-development programme and follow participation. ### The outcome: 300 000 in the first year The course grew from zero into a significant commercial product. In the first year, more than 300,000 seats were sold across teachers, school leaders, students and guardians. That created a major new source of revenue for EdAider. But the headline concealed an important problem. Many schools purchased access without achieving the activation and completion we wanted. Strong sales did not guarantee that teachers started the course, finished it or felt confident applying what they learned. The product was working commercially. It was not creating consistent change. That shifted the question from “How do we improve the course?” to “What needs to happen around the course for learning to become practice?” ### What the research changed We compared schools that had begun using AI confidently with those that struggled to get started. The biggest differences were not course content or access to better technology. They were organisational. Successful schools had visible leadership support, protected time for learning, a safe environment for experimentation and examples grounded in real teaching situations. Teachers shared a common language and understood why the organisation wanted them to engage with AI. Where those conditions were missing, flexible self-paced learning easily became learning for “when there is time.” In schools, that time rarely appeared by itself. The course could provide knowledge. Only the organisation could create the conditions for practice. ### The design response We stopped treating course access as the complete intervention and began designing support around it. I helped create a field guide that translated the research into practical guidance for school leaders and teaching teams: establish a purpose, involve leadership, schedule time, create psychological safety, begin with small classroom examples and follow up on confidence rather than completion alone. We also created a kickstart course for new customers. Instead of beginning with “here is your login,” onboarding began with the conditions required for the programme to work: why the school was investing, who needed to be involved, when teachers would learn and how experimentation would be supported. I do not have a reliable metric for the independent effect of these additions, and I do not want to invent one. Their value was a better product model grounded in how change actually happened inside schools. ### What I learned This work changed how I think about educational products. Strong content and a clear interface are necessary, but they cannot compensate for missing time, leadership, trust or psychological safety. The senior design problem was not simply making the course better. It was seeing the wider system around it, identifying what the digital product could support and being honest about what only the organisation could provide. ### Reflection A course can deliver knowledge. Turning knowledge into practice requires designing the conditions around it. ## Case study: Nobody was the villain — Investigating why hiring fails candidates and recruiters at the same time. Role: Research, Interviewing, Synthesis, Writing, Design, Build I expected finding a new role to be straightforward. I had twenty years of experience, a strong portfolio and an established network. Instead, I spent almost a year applying, being ghosted and hitting walls I could not see. Eventually, the experience stopped feeling like a difficult job search and started feeling like a verdict on my relevance. The question became: is this personal, or is it structural? ### What I did Over two months, I interviewed 24 designers, recruiters and hiring professionals across Sweden. Every interview was anonymised. Studying both sides was deliberate. Candidate-only research could become a complaint. Recruiter-only research could become a defence. I wanted to examine the overlap: where people on opposite sides of the same process described the same failure without knowing the other side experienced it too. I synthesised the interviews into twelve chapters, then compared the qualitative findings with ten published sources, including Swedish labour-market data, large-scale job-search studies and research into AI hiring and personality testing. The result was Ghosted: a public research report with an eighteen-minute audio version for people who did not have the time or energy to read the full study. ### What I found The problem was not one broken side. It was two groups of capable people doing careful work inside an infrastructure that throws much of that work away. **Volume had replaced signal.** One recruiter described receiving 300 applications within a day. One designer had tracked 700 applications since September. Recruiters responded to volume with faster filtering. Candidates responded to rejection and silence by applying more widely. Each side’s rational response made the other’s problem worse. **The visible market was not the real market.** Formal applications created the appearance of an open, merit-based system, while senior hiring often happened through relationships and private networks. Candidates were optimising for the process they could see, even when that process was not how many roles were filled. **AI accelerated the loop.** Candidates used AI to pass automated filters. Recruiters used AI to process the resulting applications. Job descriptions, CVs and cover letters began to converge, making genuine motivation and working style harder to see. The clearest shared principle was: AI for logistics, humans for judgment. In practice, hiring often ran in reverse. Machines screened and scored while people scheduled, tracked and processed volume. ### Decisions that shaped the report I preserved contradictions instead of editing them away. When large surveys reported positive experiences with hiring AI and my recruiter interviews described it as a noise amplifier, I published the disagreement as a finding. I removed names, employers and identifying details, even where recognisable participants would have added authority. The openness of the interviews mattered more than borrowed credibility. I also refused to end with a proposed solution. The research described what people wanted, but it did not pretend that one product or framework could repair a system of incentives, budgets, habits and unequal power. ### The outcome Ghosted turned a private experience into a shared view of the system around it. It showed that the exhaustion on both sides was connected rather than competitive: candidate desperation created recruiter overload, and recruiter overload created more candidate desperation. The report did not prove that every hiring process works this way, nor was the sample statistically representative. Its value was qualitative: revealing recurring behaviours, tensions and workarounds that broad market statistics could not explain on their own. The research later became the foundation for Cassil, a candidate-controlled career product, a separate case study. Ghosted’s outcome was the understanding required to decide what was, and was not, worth building. ### What I learned I began the research expecting to find a villain: careless recruiters, entitled candidates or irresponsible automation. I did not find one. Recruiters wanted to treat people better but could not do it at application volume. Candidates wanted to be honest but learned that performance was safer. Both sides adopted tools that helped them survive individually while making the shared system worse. The experience changed how I think about research. Personal frustration can identify a real problem, but it cannot be the conclusion. The work is to test the story you arrived with hard enough that it can be replaced by a better one. ### Reflection I thought hiring was broken because people had stopped caring. Ghosted showed me that people were still trying; the infrastructure between them was failing. ## Case study: A candidate-owned memory — Becoming more useful with every career decision. Role: Solo product design, Strategy, Brand, Build My research into the Swedish design hiring market found a system failing both candidates and recruiters. The complete findings are documented separately, in a case study called Ghosted. Cassil focuses on one smaller part of that system: the repeated work carried by candidates. Designers describe themselves again and again across CVs, applications, interviews, spreadsheets and isolated AI conversations. Each tool holds one piece of the search, but none of them build a useful memory of the person: how they work, what they want, why one opportunity feels more relevant than another. I wanted to reduce that repetition without automating the decisions that should stay human. ### What I did Three tools sharing one candidate-controlled memory. A profile captures how someone works. An application board records the roles they pursue and what they think about them. A coach helps them reflect on what they want next. Each surface learns from the same user-approved context, so the person never has to start again. I designed the product architecture, interaction model, visual identity, data structure and boundaries myself, then used Claude Code to implement it, testing and refining the working behaviour as I went. The most important decisions were about what Cassil should refuse to do. There is no candidate score, no auto-apply and no employer view. AI can organise information and suggest interpretations; the candidate approves, corrects or removes them. ### The outcome so far Cassil moved from an idea to a working product. In August 2026, a hand-picked group of 20 designers began testing it over four to six weeks. The largest risk is the first twenty-to-forty-minute interview. If people do not finish it, or if the resulting profile does not feel useful quickly, the rest of the product does not matter. The pilot is testing whether people: - Complete the interview and recognise themselves in the result - Return to save and annotate jobs - Correct or reject the system’s interpretations - Find the coach more useful as context accumulates - Experience less repeated work across the search I do not yet have retention, revenue or evidence that Cassil works at scale. The honest outcome is a functioning product with real testers and clear assumptions still waiting to be proven. ### The most important thing I learned Creating a product is much harder than vibe-coding a convincing prototype. Current tools make it remarkably fast to build interfaces and add features. That can create the feeling of progress while avoiding the questions that determine whether a product actually exists: - Will people use it more than once? - Is the problem important enough for them to pay? - How will they discover it at the moment they need it? - Can I reach those people without spending more than they are worth as customers? - Does each new feature improve the core value, or only make the prototype look more complete? The difficult work shifted from “Can I build this?” to “Should this exist, for whom, and how will it survive?” A polished experience is not enough. Without users, repeated value and a credible way to reach the market, it is still a prototype, however complete it looks. That changed how I evaluate product work. Building is no longer the finish line. It is the point where the real questions begin. ### How it works **Profile.** A written interview explores how the user shapes work, collaborates, handles feedback, approaches risk and leads. Cassil turns the conversation into a structured profile. Every interpretation traces back to evidence, and the user approves every line. **Application board.** The board keeps live and inactive applications together. Saving a role prompts a short reflection: what appealed, what created doubt, and where the role would stretch the user. Those answers add context for future job comparisons. ### The coach The coach draws on the profile, saved jobs and previous reflections. Instead of beginning with an empty conversation, it can identify patterns across the search and help the user clarify what to pursue next. ### What sits underneath The data model stores interpretations together with the evidence behind them. Profile statements can be flagged and corrected, giving the user control while showing where the system reads people badly. Saved-job reflections create a second layer of information based on choices rather than self-description. Over time, the product can become more specific without asking the user to maintain another profile. This is the part of Cassil I would continue testing: not how many features it can contain, but whether its shared memory becomes valuable enough for people to return, trust it and eventually pay for it. ### Reflection A polished experience is not enough. Without users, repeated value and a credible way to reach the market, it is still a prototype, however complete it looks. ## Writing I write about the experience of being a designer, seeing things from the design perspective, and the systems that shape human behaviour. ### Are we stuck in an AI loop, AI loop, AI loop? Candidates using AI to write cover letters, picking out the right keywords, hoping that just the pure amount of applications sent will help them land the job they want. Nobody really knows what it takes to pass the first sorting. They try out everything just to see what sticks. Even if all criteria are met, there is no guarantee that your application will make it to the next step. The bar is invisible, and the feedback signal is almost always absent. What’s left is just to try out everything and hope for the best. It feels more like a lottery than a system that promotes merit. #### Performing for each other Many candidates I’ve spoken to don’t even want to use AI because they feel their voice is getting lost. At the same time, they don’t want to be left behind when they suspect everyone else is using AI and getting an upper hand. They suspect recruiters are using AI for screening too, making it impossible to know who they’re actually talking to. Is it a human or a machine? In the worst cases, both sides show up performing — the candidate rehearsing answers, the recruiter reading AI-generated questions from a screen. What we get is candidates creating an abundance of applications just to hope to get through, making recruiters use AI to sort out the vast majority on unclear premises. Everyone is worse off. Neither side can stop. #### The vanishing signal When applying for jobs was cumbersome and hard, and writing a good personal letter took real effort, it was a sign of dedication. A personal letter could actually tell quite a lot about who you are and whether you are a good fit for the role. Now this signal has disappeared. Those who use AI carelessly lose their personality in the process. Those who still write by hand can’t be trusted either — everything is suspected to be AI slop and has therefore lost its value. Meanwhile, the recruiters who are actually making hires have mostly abandoned the formal system altogether. They’re working their personal networks instead. The system gets overwhelmed by noise. The good is mixed with the bad, sorting it out takes enormous effort, and the personal skills that actually matter most are the hardest to communicate. Ghosting has become so normal that one candidate I interviewed stopped categorizing it — no tag, no color code, just silence filed away as expected. #### Stuck in the loop Everyone feels stuck in this hamster wheel. Candidates are busy trying to survive, find the job, and just keep afloat. Hiring managers are swamped. They don’t have time or the mandate to recreate the whole system. They don’t like it, but they don’t see any other way. That’s the most common thing I heard — on both sides. Both sides are using AI to survive a system they wish didn’t exist. It’s like watching people build better weapons to fight in a war nobody wants. I’m not anti-AI. I use it constantly. But I’m watching two communities optimize their way into mutual distrust, and it’s hard to see the endgame. #### A design problem, not a technical one The AI arms race in hiring isn’t a technical problem. It’s a design problem. The system incentivizes the wrong behavior. Until that changes, more AI will just mean better gaming, not better matches. This series comes from six weeks of interviews — 24 conversations with designers looking for work, recruiters trying to hire them, and hiring managers caught in the middle. I’m sharing what I learned because I believe both sides deserve a better process. One built on transparency, dignity, and actual human contact. ### Should hiring be more like dating? One question kept popping up while I was doing the interviews for the Ghosted Report. This can’t be a unique problem. There need to be others struggling with exactly the same thing. “What other area spends this much energy trying to connect two humans to see if they are a perfect match?” So I called Olga Engwall to find out. She is a former UX colleague, now building MinglMe, a fresh take on dating. I needed to know: have they actually solved the problem of translating data into feelings? Is there a way to use AI for more than just efficiency and automation? Or are we all just stuck in the same trap? I needed a new perspective. I got more than one. Olga has spent years thinking about why people connect, and more specifically, why the systems we built to help them connect keep getting in the way. Her diagnosis of dating apps is sharp: they turned human connection into a volume problem. Swipe left, swipe right, five hundred people processed before breakfast. The apps make it feel like efficiency. What it actually produces is numbness. Every major dating app is owned by the same company, Match Group, and they all have the same basic design underneath the different branding. Some sell themselves as casual, some as serious. All of them keep you in the app as long as possible, because that is how they make money. A match is actually bad for the business model. Like a medicine that does not want to get you healthy, just keep you taking more of it. Sounds familiar? The Ghosted report found the same mechanics running through hiring. Candidates send applications into the void. Recruiters drown in noise they cannot convert into anything useful. Both sides perform for a system that serves neither of them. Olga’s summary was insightful. “You hire A person without THE person. The personality gets removed from the process.” It is a good line. I heard similar versions of it from designers and recruiters who had never met her. Dating apps let you filter on height, age, interests, distance. Hiring does the same with years of experience, job titles, keyword matching. Both create the feeling of control while completely bypassing the thing that actually decides whether it works: what happens when two people are in a room together. Olga cited research on this. Even a perfect match on Big Five personality scores, the most scientifically validated personality model we have, tells you almost nothing about chemistry. Two people can score identically across all five dimensions and still sit across from each other and feel absolutely nothing. You cannot quantify the moment it clicks. MinglMe is built on exactly this idea. Instead of matching people on profiles and hoping for the best, the app puts people in shared activities. Cooking class, escape room, pub quiz, whatever. You buy a ticket. The chat opens. You show up. The activity is the thing, not the date. You are busy for an hour doing something together, and in that hour you see things no profile could ever show: how someone handles losing a game, how they treat a stranger, whether they are generous or stingy with praise. You cannot keep the front window up for ninety minutes while also trying to crack a code with four people you just met. Olga’s phrase for what this creates: “You see who they actually are.” A designer I interviewed for the Ghosted report, without knowing anything about MinglMe, said almost the same thing about his first in-person visit at a company he later joined. He saw the office. He watched how people greeted each other in the corridor. He got a coffee and felt whether this place could be part of his life. The formal interview mattered less than those first twenty minutes of just being there. One designer put it: “You taste the coffee. That tells you more than anything they ask you.” Two people from completely different worlds landing on the same metaphor. That is not a coincidence. That is a signal. #### There are no soulmates Both in dating and hiring there is this illusion of the perfect candidate. One person who solves everything. Your job is just to find them. In reality, people with totally different skills, personality, and ways of thinking can all solve the same problem, but have completely different effects on the people around them and the organization they join. Not better or worse. Just different. One might strengthen what is already there. One might challenge it. One might bring something the job description never thought to ask for. Most hiring processes are designed to fill a role as written. Olga is arguing for something harder: look past the current problem to where you are actually heading. The difference is between filling a position and making a bet on a direction, where the direction matters more than the starting point. That thinking is not in any ATS system I have ever seen. Hard skills are the entry ticket, not the destination. Your diploma, your experience, your portfolio: they prove you can do the job. But they are backwards-looking. Like stocks, you cannot judge future performance only by looking at history. At some point you need to get into the room, see if you connect, find out whether you are actually heading in the same direction. We do not marry someone who looks good on paper. We do not become close colleagues with someone we do not like or cannot collaborate with. When there is a genuine will to connect, the other problems become solvable. I started this conversation looking for confirmation. I already believed the hiring system was broken and I wanted another perspective on why. What I found was someone who had stopped asking that question. Olga is not trying to improve the filtering machinery. She is not building a better swipe. She looked at the problem and asked something different: how do we get people into a real situation together, as fast as possible? That question does not exist in hiring yet. It probably should. Olga Engwall is co-founder of MinglMe. ### Bad UX made me do it Last year I spent almost a full year trying to find a new job. I thought it would be straightforward. Twenty years of experience, a decent network, solid portfolio. I was completely wrong. I got ghosted more times than I could count. Hit walls I could not see. And somewhere along the way I stopped feeling like a senior professional looking for a new opportunity, and started feeling like an irrelevant has-been that nobody needed. #### Rage against the machine The anger came next. I needed someone to blame. The recruiter. The algorithm. The job market. The AI. None of it stuck, because none of it was quite right. Eventually I got tired of being angry and decided to switch focus. I wrote about it and it became one of my most recognized things I have ever published on LinkedIn. Apparently a lot of people were sitting with the same feeling and just needed someone to say it out loud. #### Stop chasing waterfalls I stopped chasing design jobs and enrolled in AI and machine learning instead. Suddenly I had purpose and agency. Instead of writing meaningless applications I wrote code, learned the math and principles behind what we lazily call AI. I learned more in a few months than for the past years. #### Don’t look back in anger A year later, with less desperation and more curiosity, I went back in. Two months of interviews with 20+ people across all sides of the hiring process. Trying to find the shape of the thing and understand how all parts fit together. #### Wicked games Hiring is a wicked problem. And what is happening to designers in this market is not one thing going wrong. It is many things colliding: the economics of design work shifting, AI rewriting what the job actually is, companies pausing on juniors, then seniors, then everyone while they figure out what a design team even looks like now. Wars, recessions, a post-pandemic hangover, organizations rethinking everything from the ground up. None of those things in isolation explains what happened to me or to the people I interviewed. All of them together do. What I did not expect to find was hope. Designers getting creative about how they show their value. Staying visible, building things, refusing to disappear quietly. Hiring managers pushing back against full automation, keeping humans in the decision loop, delegating the administration to the machine instead of the judgment. Small moves, but the right ones. #### Here comes the sun I do not know if this report fixes anything. That was never the goal. The goal was to understand it well enough to stop feeling like it was just me, and maybe to make the conversation a little more honest on both sides. That felt worth two months of unpaid work on a Sunday afternoon.