Host Country Open Stage

Summary

This discussion featured presentations from four Norwegian companies demonstrating how AI and technology can address critical challenges in healthcare and public services. Ivar Hukkelberg from Deep Insight opened by explaining how their company uses AI and mathematics to optimize hospital resource management, particularly in surgical departments. He highlighted that operating rooms remain empty 40% of the time despite long waiting lists, largely due to inadequate planning tools like pen and paper or Excel spreadsheets. Deep Insight developed an AI-powered solution called “Deep Inside Hero” that uses predictive models to forecast surgical demand and generate optimized scheduling plans. Working closely with hospitals like Luisenberg in Oslo, they achieved a 12% increase in surgical activity and treated over 1,000 additional patients within a year.

Vegard Samestad-Hansen from Vilmer addressed the growing problem of loneliness among elderly populations, noting that over one million elderly people in the Nordic countries experience loneliness. His company created an activity-as-a-service platform that matches volunteers and paid contributors with elderly residents in nursing homes, delivering over 100,000 positive moments while increasing activity levels by two to four times at a quarter of the usual cost. The presentation from CIMA, delivered by Manav Rihel Kumar and Jakob Sverre Løvstad, focused on diversity management in organizations. They explained that diversity alone doesn’t automatically create value but requires proper management, and their approach involves analyzing organizational data to create tailored leadership training programs.

Finally, Malcolm Langford from Trust, one of Norway’s new AI research centers, discussed the challenge of building trustworthy AI systems. He emphasized that while people prefer human decision-makers over AI alone, they favor the combination of humans working with AI. Trust focuses on developing AI that is accurate, interpretable, inclusive, safe, and scalable, working with partners across public and private sectors. These presentations collectively demonstrated how Norwegian organizations are leveraging AI and technology to create more efficient, inclusive, and trustworthy solutions for societal challenges.

Keypoints

## Major Discussion Points:

– **AI-driven healthcare resource optimization**: Deep Insight’s presentation focused on using AI and predictive modeling to solve hospital scheduling inefficiencies, particularly in surgical departments where operating rooms sit empty 40% of the time despite long waiting lists. Their solution resulted in 12% increased activity and over 1,000 additional patients treated.

– **Technology-enabled elderly care and combating loneliness**: Vilmer’s platform addresses the growing care gap for elderly populations by creating an “activity-as-a-service” model that matches volunteers and paid contributors with nursing home residents, delivering 2-4 times more activities at a quarter of the cost while addressing widespread loneliness among over one million Nordic elderly.

– **Diversity management through data-driven leadership training**: CIMA presented their approach to transforming diversity from a compliance issue into value creation through statistical analysis of organizational cultures, using multi-mediation analysis across 7,000 data points to create tailored leadership training programs that address specific diversity challenges within individual organizations.

– **Trustworthy AI development and governance**: Trust research center outlined the critical challenge of building AI systems that are not only accurate but also interpretable, safe, inclusive, and scalable, particularly in high-stakes decision-making contexts like immigration and court systems, emphasizing the need for proper governance frameworks and calibrated trust levels.

## Overall Purpose:

The discussion appears to be a conference or symposium showcasing how various Norwegian organizations are leveraging AI and technology to address significant societal challenges across healthcare, elderly care, workplace diversity, and AI governance. Each presenter demonstrated practical applications of their solutions with measurable results and real-world impact.

## Overall Tone:

The tone throughout the discussion was consistently optimistic and solution-oriented. All presenters maintained a professional, confident demeanor while discussing serious societal challenges. The tone remained forward-looking and innovation-focused, with speakers emphasizing positive outcomes, measurable results, and the potential for scaling their solutions. There was no notable shift in tone across the presentations – all maintained an entrepreneurial spirit combined with social responsibility messaging.

Speakers

– **Ivar Hukkelberg**: Chief Revenue Officer at Deep Insight, expertise in AI and mathematics for healthcare resource management and surgical planning

– **Vegar Samestad Hansen**: CPO (Chief Product Officer) at Vilmer, expertise in technology solutions for elderly care and activity-as-a-service platforms

– **Malcolm Langford**: Representative of Trust (AI research center), expertise in trustworthy AI, AI governance, and AI applications in legal/administrative decision-making

– **Manav Rihel Kumar**: From CIMA, expertise in insight-based leadership training and diversity management in organizations

– **Jakob Sverre Lovstad**: From CIMA, expertise in statistical analysis and multi-mediation analysis for diversity and organizational performance

**Additional speakers:**

None identified beyond the provided speakers names list.

# Discussion Report: AI and Technology Solutions for Societal Challenges

## Executive Summary

This discussion featured presentations from four Norwegian organisations demonstrating applications of artificial intelligence and technology to address challenges in healthcare, elderly care, diversity management, and AI governance. The speakers presented technology-enabled solutions that focus on optimising existing resources and human-AI collaboration.

## Detailed Presentation Analysis

### Healthcare Resource Optimisation Through AI

**Ivar Huckelberg** from Deep Insight addressed inefficiencies in healthcare delivery, specifically that operating rooms remain empty 40% of the time while hospitals maintain extensive waiting lists for surgical procedures. He explained that many hospitals still rely on pen-and-paper systems or basic Excel spreadsheets for resource allocation decisions.

Huckelberg founded Deep Insight with mathematicians “who believe in using our knowledge to create a positive impact for society.” He described how he almost “infiltrated” a hospital by “postponing as a doctor” to understand their problems firsthand. His key argument was that traditional approaches to healthcare challenges are unsustainable: “We cannot just hire ourselves out of the problem to put more people into the healthcare sector to meet the increasing demand. Instead, we have to think differently around how we utilise the current resources we have.”

Deep Insight’s AI-powered platform employs predictive modelling to forecast surgical demand and generate optimised scheduling plans. Working with Luisenberg Hospital in Oslo, the system achieved a 12% increase in surgical activity and treatment of over 1,000 additional patients within a year. Huckelberg emphasised that the solution augments rather than replaces human decision-making, allowing planners to leverage AI insights whilst maintaining control over scheduling decisions.

### Technology-Enabled Elderly Care Solutions

**Vegard Sammerstad-Hansen** from Vilmer addressed the elderly care crisis, noting that over one million elderly people in the Nordic countries experience loneliness. This challenge is compounded by projected healthcare staff shortages that will increase from 7,000 to more than 50,000 within 2040.

Hansen shared a personal anecdote about his grandfather, who said “okay then I will lie down here, stare at the blank wall” when faced with limited activity options. This experience highlighted an important insight: “For every elderly living out there, there are actually five people that want to contribute with some good moments.”

Vilmer’s activity-as-a-service platform matches volunteers and paid contributors with elderly residents in nursing homes. The platform has delivered over 100,000 positive moments whilst increasing activity levels by two to four times at a quarter of the usual cost. Hansen’s approach transforms elderly care from a resource scarcity problem into a coordination challenge that technology can address.

### Data-Driven Diversity Management

**Manav Rihel Kumar** and **Jakob Svære Løsta** from CIMA challenged conventional approaches to diversity management. Kumar questioned fundamental assumptions: “Diversity doesn’t automatically yield better results. Diversity brings a lot of complexity, but also a potential to do better.”

Kumar argued that diversity competence is becoming a basic leadership skill due to globalisation, but emphasised that realising diversity’s value depends on how organisations manage the associated complexity.

Løsta presented their methodological approach, which involves statistical analysis of 7,000 data points using multi-mediation analysis. CIMA connects diversity initiatives to concrete business metrics such as billable hours and sick leave, creating business cases for diversity programmes. Their approach recognises that diversity’s impact varies significantly by organisation, requiring tailored solutions rather than universal approaches.

### Trustworthy AI Development and Governance

**Malcolm Langford** from Trust, one of Norway’s new AI research centres announced two weeks ago by the Prime Minister, addressed building trustworthy AI systems for high-stakes decision-making. Trust employs 150 researchers and 35 PhD candidates working on AI governance frameworks.

Langford shared research findings about public attitudes towards AI: “Research shows that if citizens are asked if they would like a court decision or an administrative decision to be made by a human or an AI system or a robot, they would prefer a human… But when we also ask citizens, would you prefer a human with AI to make a court decision or administrative decision, we prefer that combination more than just a human.”

Trust focuses on developing AI systems that are accurate, interpretable, inclusive, safe, and scalable, with applications in immigration and court systems. Langford mentioned their work with 4 million documents on 120,000 asylum decisions as an example of their immigration system analysis.

The research centre tests different AI models with immigration directorates and courts, including data analytics of decision-maker behaviour. Langford emphasised the importance of “veridical AI” and uncertainty quantification, ensuring that users neither over-trust nor under-trust AI systems.

## Key Themes and Insights

### Human-AI Collaboration

All speakers advocated for human-AI collaboration rather than AI replacement. Huckelberg’s hospital planning system augments human decision-making, while Langford’s research shows public preference for human-AI combinations over purely automated systems.

### Context-Specific Solutions

Speakers consistently emphasised tailored approaches. CIMA’s diversity management recognises organisational differences, while Trust’s AI governance requires careful attention to specific implementation contexts.

### Resource Optimisation vs. Scaling

Both Huckelberg and Hansen argued against traditional scaling approaches, instead advocating for technology-driven solutions that optimise existing resources. Huckelberg focused on better utilising hospital capacity, while Hansen identified untapped community willingness to contribute to elderly care.

### Stakeholder Engagement

Successful implementation requires careful attention to end-user needs. Huckelberg described building trust with hospital planners, while Hansen’s platform depends on volunteer engagement.

## Challenges and Future Considerations

The speakers identified several ongoing challenges. Scaling AI healthcare solutions across different hospital systems remains complex, as does maintaining long-term volunteer engagement in elderly care platforms.

Langford highlighted emerging challenges as AI systems scale, particularly managing interactions between multiple AI systems and preventing emergent behaviours. He emphasised the need for proper uncertainty communication and calibrated trust in AI systems.

## Conclusions

The discussion demonstrated how Norwegian organisations are leveraging AI and technology for societal benefit. The presentations showed technology’s potential to optimise existing resources, coordinate community engagement, and solve problems that traditional approaches struggle to address effectively.

The consistent emphasis on human-AI collaboration and context-specific solutions provides practical guidance for technology implementation across diverse domains. The speakers’ focus on optimisation rather than simple scaling suggests that many societal challenges may be more addressable through better coordination and resource utilisation than previously assumed.

Ivar Hukkelberg: Please welcome to the stage, from Deep Insight, Ivar Huckelberg. Thank you so much for the invitation to come here today to share the story about Deep Insight and how we use AI and mathematics to create a better and more sustainable healthcare service. My name is Ivar and I’m the Chief Revenue Officer in Deep Insight. And before we dive into the story about us, I want to quickly present the company. And we’re founded by mathematicians who believe in using our knowledge to create a positive impact for society. And what better way to do that than using our knowledge in the healthcare sector. So let’s take a step back and see the global picture right now that’s concerning the healthcare sector. And everyone here today, either themselves or someone you know, will at some point in your life be in need of healthcare services. And the time from you get ill to you get the treatment can be quite psychologically demanding, especially if you constantly get postponed the date you are to be treated or cancel the surgery you have been booked into. And the waiting time is growing longer and longer. How can we solve this when the population is getting older and older? And believe me, I want to get old as well. That’s a great thing. But we are also giving birth to fewer people. We cannot just hire ourselves out of the problem to put more people into the healthcare sector to meet the increasing demand. Instead, we have to think differently around how we utilize the current resources we have. And in that way, still provide sustainable healthcare in the future. That is why Deep Insight is on a mission where we want to help hospitals manage and plan their resources with AI to ensure that as many people as possible receive the help they need when they need it. And for us, where can we make the biggest impact? And for us, that was in the surgery department, one of the core services of the hospitals. And if you look at how the planning is done today, there is immense amount of data going in to making these plans and ensuring the operation runs smoothly. And all these puzzle pieces that are to fit together to create this surgical program is really hard for the planners at hospitals to do when the current tools they use are pen and paper and Excel spreadsheets. They lack good tools and support to actually utilize the resources the best way possible. So the consequence is that the operating room are empty 40% of the time, even though the waiting lists are mile long. So what we then think of is that using AI and backed up by research, we think of planning through demand-driven way, where we actually use AI and predictive models to understand what type of surgery, how many types of surgery for different types will come in the next two, three, four months, and then map that up to the hospital’s resources that they have available so they are better prepared. But to create this solution, we cannot do it ourselves, then it wouldn’t work. We need to include the hospitals when developing these kind of solutions. So we’ve been working with a couple of hospitals here in Norway, one of them Luisenburg here in Oslo. And what we have done is like, before we even wrote the first line of code, we went up there, talked with them, understanding their problems and what feature they really need. We even went as far as, well, me almost infiltrating them, postponing as a doctor to really feel what it is like to be in their shoes. And also, it’s important to be close to them, to make them trust what we create with the algorithms and AI, for them to trust the output our solution gives. And together with them, we created a solution we call Deep Inside Hero for the heroes that work in the healthcare sector. And what we do is, basically, we use AI to generate plans in the future that tells the hospital what are to happen on a given operating room on a given day. So it could be on operating room one, we should do three knee prostheses. And that makes it easy for the planners to slot in the right patient and to fill that OR. It also accounts for critical bottlenecks in the system, so we don’t push more patients through than there are beds available, for example. And it also gives the planners suggestions on what patient of the thousands of patients on the waiting list who does fit in the program and are ready for surgery. And does it work? We’ve had it running for over a year at Luisenberg, the hospital. And the result has been incredible, 12% increase in the activity, vast amounts of time saved on planning. And these are numbers that top management and politicians screams of joy with when they see. But in the end, it’s about the patients. And this accounts for more than a thousand more patients treated by planning differently and using AI. That’s a thousand more people that are able to go back to their normal life sooner rather than later. That is why we are here. That is how we can use AI and mathematics to still provide a sustainable healthcare service in the future. My time here is soon up, but our company is just in the beginning of the journey and we are going to make some huge waves going forward in the healthcare sector. So thank you so much for your time. Please welcome to the stage, from Vilma, Vegard Sammerstad-Hansen.

Vegar Samestad Hansen: Hello, thank you for the opportunity to present here. I’m Vegard Sammerstad-Hansen, the CPO of Vilmer. So what we have done is that we have created more than 100,000 good moments for the elderly using our own developed technology. We do this with an activity-as-a-service model and our platform is today engaging thousands of volunteers and paid contributors to contribute with good moments for the elderly. The interesting thing with our technology is that we can actually deliver a two to four times increase in amount of activities in nursing homes at only a quarter of the cost. So it’s a very sustainable solution and the market is also attractive to be in as a private company. And why are we doing this? We are doing it because loneliness and inactivity is one of the largest healthcare challenges we do have and also probably one of the largest problems we have as a society today. So today over one million elderly in all of the Nordics do experience loneliness. I remember my own grandfather at his final days. He usually said to me when I left him from my usual visits, okay then I will lie down here, stare at the blank wall. So this is exactly what we want to do something about. And it’s about time because the care gap is really growing at this time. Healthcare workers say that they don’t have time. There is a healthcare staff shortage that will increase from 7,000 to more than 50,000 within 2040. And families say that they don’t have time to contribute, come for visits and help out the elderly. However, the local citizens do say that they do want to help out. There are a lot of people that want to contribute with good moments for elderly. And this is what we are trying to do something about. I’m sure that the one of you that are within this sector is very familiar with these graphs. The timing is definitely now, we can’t wait. It’s something that has been around for a long time. the latest 20 years, but right now we really see that the gap between healthcare staff and LRA is hitting us and hitting also all institutions across the world. So let me tell a bit more about what we are doing. What we try to do is to make the elderly’s life look more like this. We bring in physical activities, everything from dog visits to someone taking a coffee cup with an elderly, or someone coming taking them for a walk. We also usually engage some celebrities to arrange bigger events such as concerts and other larger events. How we are doing it? We have developed our own matching platform that maps out all the different activity needs from the institutions, the elderly living there, and based on that we recruit and engage contributors from the local society. And for every elderly living out there, there are actually five people that want to contribute with some good moments. Some of them want to help out maybe once a month, others want to help out 12 hours a week, some do it for free, and some want to get a bit of a smaller payment that we also can arrange. Per today, there aren’t really any specialized solutions that take advantage of this unused workforce. So they haven’t been asked before, we come in usually. So our matching and coordination platform really connects these two needs together and then we can increase the activity level that you have today. The result is a digital activity plan, a pre-made week filled with good moments for every elderly living in institutions that contains both digital activities, physical activities, and everything that the elderly living there is interested in. It looks like this. There are some images of the activities we recently have arranged, all of these are from the previous week. Here we have a store contributing with some free food, here we have some dog visits, some digital lectures, a horse coming for a visit to a nursing home, and a lot of varying activities. In the end here, we actually are able to reduce coordination costs that we do have from before. When it comes to arranging activities, we increase the activity level by two to four times, and at the same time we increase the quality at least according to our own NPS service to the citizens on the activities itself. So we are just one out of many examples on how technology can really do something about the elderly way, and at the same time keep the quality of the life that I need in the end. Thank you.

Ivar Hukkelberg: Please welcome to the stage, from CIMA, Manav Rihel Kuma, and Jakob Svære Løsta.

Manav Rihel Kumar: Hi everyone, we’re going to speak a little bit about how we can do insight-based leadership training, and connect diversity to value. So around 2015, there were numerous reports coming out that linked diversity to value creation, but they didn’t really say how do you get the value out of diversity. And there was a saying in the private sector and public sector that diversity could be the new goldmine, so everyone were looking to how can we become more diverse. But the truth is that diversity doesn’t automatically yield better results. Diversity brings a lot of complexity, but also a potential to do better. So whether you deal with the complexity in a good way, and you actually get to the value that diversity can bring, that depends on how you manage diversity in the organization. And our job is to figure out what that depends on. So every organization is a village of its own, with its own culture, with its own way of doing things. So there is no one-size-fits-all when it comes to releasing diversity potential. So you need to figure out what is the truth, and what is the culture in your organization, and how is that hindering or promoting a diversity potential.

Jakob Sverre Lovstad: All right, so as Manuel said, this is kind of the old model, so to speak. But at this point, we have analyzed something like 7,000 data points in Norway, across a multitude of different industries. I mean it’s retail, arts, oil and gas, whatever you want to point a stick at, we’ve been there. And this simplified model has been proven not to work. As he said, it all depends. And it also is individual to the organization. The mistake is often to take US research in particular, and transfer it to individual organizations far from wherever they did the original research, and that has shown to be wrong. So there has to be a different approach. And when we say that something depends, statistically speaking, we have to do something called multi-mediation analysis. So basically, as you bring more diversity into the organization, you see, as I said, that there’s more complexity. But how that complexity is handled is mapped through the psychosocial variables individual to the organization. So when we work, we come into the organization and we see this kind of thing, all right? We see that the diversity we map out through statistical methods, they do affect well-being, but also other variables like sick leave, turnover, how much billable hours you have per person, things like that. But we can also say, well, how does that happen? In this particular picture, you see that the main variables of interest are authenticity and development opportunities, which is, no sorry, competency utilization, which is something that we could typically see. So if we’re going to work with this particular organization, we would say, well, for diversity, we need to do something about the ability to be authentic. Is there enough psychological safety here? That kind of thing. We also have to look at whether or not people, especially diverse people, get to utilize their competence in their jobs. And that’s where we have to look to see how we can create more value for the company. And this is another example of the same kind. Here you see, instead of well-being, we have billable hours. So it’s both the soft variables and the hard variables. This makes it possible for organizations to create business cases out of something that is often perceived as a very soft domain. We can transfer that into, you know, hard line, billable hours, for example, sick leave, turnover, things that do matter to the bottom line. Also, when we are saying diversity, we have an expanded model of what diversity is. At least in Norway, most of when you hear people talking about diversity, it boils down to either ethnicity or gender. We also map out things like age, neurodiversity, socioeconomic class, all the things that contribute to the diversity experience. In addition, we also look at perspectives or perspective diversity, which means that all of us here have, you know, our perspectives, our opinions, our feelings and so on under the hood. That is also shown to be very, very important in the workplace for performance. So, sir.

Manav Rihel Kumar: Yeah. So what we do when we work with organization is that we use the method that Jakob described to do an analysis of what is happening in your village, right? So that’s the organization. And based on that, we tailor make a leadership or a training program. And then we measure again after we’ve conducted it to see if you got the bottom line result you wanted. So, for example, if your aim was to improve subjective well-being in the organization, we do the analysis. We see how much subjective well-being is affected by diversity. We do the training and then we measure again to see if your diversity sensitivity for subjective well-being has gone down afterwards. So what we base all this on and why diversity is important is that leadership is usually easy when it comes to similarities. So if you apply your leadership skills, but you select people who are the same as you, leadership is quite easy. But leadership becomes challenging when you introduce diversity and it could be a little bit diversity that’s introduced or a lot. And the way the world is developing right now with more globalization, more diversity coming into workplaces, diversity competence is a basic leadership skill. That is something that every organization needs to focus on when developing leaders for the future. And this is our slogan. So what we do is to build future or diversity-ready organizations through statistical methods and tailor-made leadership training programs. And our primary goal is to take diversity from something that you do out of sympathy to something you do to create value. So looking at how we can create value based on the differences that exist in the organization. Thank you very much. Thank you for your time.

Malcolm Langford: Please welcome to the stage, from Trust, Malcolm Langford. Research shows that if citizens are asked if they would like a court decision or an administrative decision to be made by a human or an AI system or a robot, they would prefer a human. Even more, research shows that we will tolerate failures by a human decision-maker much more than an artificial decision-maker. So in many contexts, particularly hard cases like decision-making, AI has a lot of work to do to make itself trustworthy. But when we also ask citizens, would you prefer a human with AI to make a court decision or administrative decision, we prefer that combination more than just a human. And that shows the path or the potential for using AI in smart ways in which we don’t compromise our moral values. So how do we get there? I represent Trust, which is one of the six new AI centers in Norway, in research and innovation, which were announced two weeks ago by the Prime Minister. We are 150 researchers, 35 PhD candidates, who are focused on how do we get AI which is accurate, interpretable, for example, it’s explainable, it’s aligned with our values, it’s inclusive of the voices that we want to hear when designing AI, it’s safe, it’s not going to cause harms, it’s sustainable, it’s also well-governed, and particularly, and something that’s often missed in definitions of trustworthy AI, it works at scale. The more and more we start to roll out AI systems, the more and more we lose control of them in practice. And AI systems start to speak to AI systems and create emergent behaviors. So how do we have systems to govern this? We have 18 research partners in Norwegian institutions. We have 44 user partners from the public sector, the private sector, so from Equinor, a multinational, to the court administration. We also have civil society organizations like Amnesty, and we work with international partners like Cambridge, University of California, and so forth. We have 14 research areas, which includes everything from veridical AI, how we get truth and transparency in the way we pull together data. This keeps on moving for some reason. I can’t seem to stop it, but anyway, it’s got a mind of its own. It’s controlled by AI, perhaps. I’ll keep on trying to keep it back. We also are interested in responsible uncertainty quantification. So when you get an AI prediction, what sort of number are you going to get as to how certain the AI system is? Maybe it’s 85% certain with a 10% confidence interval. What does that mean in practice? And how can we communicate to you in a language that you as a non-data scientist will understand what are the real risks here? Then how do we attribute what the system is doing? Explain what are the core features in this? How do we look at what’s causing those predictions? How do we work with foundation models, generative AI, emergent behaviors when AI systems start to speak to each other, and so forth? We also have 15 action clusters with everything from healthcare, mobility, case processing, which I’ll come back to, infrastructure, dealing with criminality and the climate crisis, but also technical solutions. For example, anomaly detection. How can you use AI to detect that something might be wrong in your system? That ID fraud is happening, or some form of criminality is happening in an ocean surveillance system. So if we come back to the example that I gave you at the beginning, from the immigration directorate in Norway, we’ve got 4 million documents on 120,000 asylum decisions. A very sensitive area of decision making with massive consequences for people’s lives. A lot of personal data in those decisions, which makes it very difficult for the immigration directorate to work with AI on it or share it with researchers. But we developed a data infrastructure where we have all of these documents and decisions in a secure location that no one outside can access. And we can train machine learning models, we can even use large language models, like our own version of ChatGPT, to find patterns here to help decision makers avoid bias, to work more quickly, to perhaps improve the quality of their writing and their decision making. But we also have a system with automatic de-identification, so that researchers and case workers can read individual decisions with an automatic masking of all the personal data, and it’s calibrated according to their research needs. And what we’re now doing in a new project, working with the immigration directorate and the courts, is testing out different models of AI. From very simple help mechanisms, where they get better access to information, for example on the human rights context in the countries asylum seekers are coming from, or summaries of facts and law, to more auto-generated aspects of decisions. But then we’re testing out how do decision makers use that AI, and we’ll look at the data analytics of their clicks and their eye movements, we’ll have focus group interviews, we’ll have expert assessments of the types of decisions they are making. Because at the end of the day, I think these four core elements need to be there when we are talking about trustworthy AI in practice, particularly in high risk areas, but areas in which we have a lot of possibility to not just increase efficiency, but improve quality, and that we do. The first one is to understand AI in context. What are the risks, but what are the user needs? What do decision makers and asylum seekers and their lawyers need from AI systems? Two, what sort of models for trustworthiness can we develop? In the system we’re developing, we’re including explainable dimensions, uncertainty quantification dimensions. Thirdly, we need system thinking and governance. We don’t know really the risks of AI, so we need a lot of scaffolding around it. We need feedback systems to help us at early stages learn what is going wrong and what is going right. And finally, we need to calibrate trust. Sometimes AI systems look like they’re working too well, and we trust them too much, and we need to test them properly and communicate back to citizens, users, judges and decision makers how accurate, how effective the AI predictions are, and what warnings need to be in place. Thank you very much. Thank you. Thank you. Thank you.

Agreement points

Technology can address resource constraints and staffing shortages in healthcare and social services

Speakers

– Ivar Hukkelberg
– Vegar Samestad Hansen

Arguments

Healthcare systems face growing demand with aging populations and resource constraints, requiring AI-driven solutions for sustainable service delivery

Healthcare staff shortages will increase from 7,000 to over 50,000 by 2040, but local citizens want to contribute to elderly care if properly coordinated

Summary

Both speakers recognize that traditional approaches of hiring more staff cannot solve the growing demand in healthcare and elderly care. They advocate for technology-driven solutions that optimize existing resources and tap into underutilized community capacity.

Topics

Development | Economic

Context-specific solutions are essential rather than one-size-fits-all approaches

Speakers

– Manav Rihel Kumar
– Jakob Sverre Lovstad
– Malcolm Langford

Arguments

Diversity doesn’t automatically create value but requires proper management to handle complexity and unlock potential benefits

Statistical analysis of 7,000 data points shows diversity impact varies by organization, requiring tailored approaches rather than one-size-fits-all solutions

AI systems need proper scaffolding including feedback systems, uncertainty quantification, and calibrated trust to work effectively in high-risk areas

Summary

All three speakers emphasize that effective solutions must be tailored to specific organizational contexts and circumstances rather than applying generic approaches. This applies to diversity management, AI implementation, and trustworthy technology deployment.

Topics

Economic | Sociocultural | Legal and regulatory

Similar viewpoints

Both speakers demonstrate that technology platforms can dramatically improve efficiency and outcomes in healthcare-related services while reducing costs. They both provide concrete evidence of successful implementations with measurable results.

Speakers

– Ivar Hukkelberg
– Vegar Samestad Hansen

Arguments

AI-generated surgical scheduling increased hospital activity by 12% and treated over 1,000 additional patients through better resource utilization

A matching platform can connect elderly care needs with community volunteers, delivering 2-4 times more activities at quarter of the cost

Topics

Economic | Development

Both speakers advocate for treating diversity as a strategic business capability rather than just a social good, emphasizing the need for data-driven approaches to demonstrate and optimize diversity’s business value.

Speakers

– Manav Rihel Kumar
– Jakob Sverre Lovstad

Arguments

Diversity competence is becoming a basic leadership skill due to globalization and increasing workplace diversity

Multi-mediation analysis can connect diversity to hard business metrics like billable hours and sick leave, creating business cases for diversity initiatives

Topics

Economic | Sociocultural

Unexpected consensus

Human-technology collaboration as the optimal approach

Speakers

– Ivar Hukkelberg
– Malcolm Langford

Arguments

AI-generated surgical scheduling increased hospital activity by 12% and treated over 1,000 additional patients through better resource utilization

Citizens prefer human-AI collaboration over purely human or AI decision-making, showing the path for acceptable AI integration

Explanation

While coming from different domains (healthcare operations vs. trustworthy AI research), both speakers converge on the principle that the most effective approach involves humans and AI working together rather than AI replacing humans entirely. This consensus is unexpected because it bridges practical implementation and theoretical research perspectives.

Topics

Development | Human rights | Legal and regulatory

Community engagement and stakeholder involvement as critical success factors

Speakers

– Ivar Hukkelberg
– Vegar Samestad Hansen
– Malcolm Langford

Arguments

Operating rooms remain empty 40% of the time despite long waiting lists due to poor planning tools, which AI can optimize through demand-driven planning

Healthcare staff shortages will increase from 7,000 to over 50,000 by 2040, but local citizens want to contribute to elderly care if properly coordinated

Testing AI in sensitive areas like immigration decisions requires secure data infrastructure and comprehensive evaluation of decision-maker behavior and outcomes

Explanation

All three speakers, despite working in different domains, emphasize the importance of involving end-users and stakeholders in solution development. This consensus spans healthcare operations, elderly care, and AI governance, suggesting a fundamental principle about successful technology implementation.

Topics

Development | Economic | Human rights | Legal and regulatory

Overall assessment

Summary

The speakers demonstrate strong consensus around several key principles: technology as a solution to resource constraints, the need for context-specific approaches, human-AI collaboration over replacement, and stakeholder engagement in solution development. They share a pragmatic approach that combines social impact with business viability.

Consensus level

High level of consensus on fundamental principles despite working in different domains. This suggests emerging best practices for responsible technology implementation that prioritize both effectiveness and stakeholder needs. The implications are significant for policy and practice, indicating that successful technology solutions require careful attention to context, collaboration, and community engagement rather than purely technical considerations.

Different viewpoints

Approach to AI implementation in healthcare vs. general decision-making

Speakers

– Ivar Hukkelberg
– Malcolm Langford

Arguments

AI-generated surgical scheduling increased hospital activity by 12% and treated over 1,000 additional patients through better resource utilization

AI systems need proper scaffolding including feedback systems, uncertainty quantification, and calibrated trust to work effectively in high-risk areas

Summary

Hukkelberg presents AI implementation as having achieved clear success with concrete metrics in healthcare operations, while Langford emphasizes the need for extensive safeguards and testing before AI can be trusted in high-risk decision-making contexts. Their approaches differ in the level of caution and scaffolding deemed necessary.

Topics

Development | Human rights | Legal and regulatory

Unexpected differences

Level of AI implementation readiness

Speakers

– Ivar Hukkelberg
– Malcolm Langford

Arguments

AI-generated surgical scheduling increased hospital activity by 12% and treated over 1,000 additional patients through better resource utilization

Testing AI in sensitive areas like immigration decisions requires secure data infrastructure and comprehensive evaluation of decision-maker behavior and outcomes

Explanation

It’s unexpected that two AI advocates would have such different perspectives on implementation readiness. Hukkelberg presents AI as already successfully deployed and delivering results, while Langford emphasizes that AI in sensitive decision-making is still in careful testing phases requiring extensive safeguards. This suggests different risk tolerances and implementation philosophies within the AI community.

Topics

Development | Human rights | Legal and regulatory

Overall assessment

Summary

The speakers showed minimal direct disagreement as they presented complementary rather than competing solutions. The main areas of implicit disagreement centered on AI implementation approaches and the balance between innovation speed and caution in sensitive applications.

Disagreement level

Low to moderate disagreement level. The speakers were largely aligned on identifying problems (aging populations, healthcare shortages, need for better technology solutions) but differed in their approaches to implementation and risk management. This suggests a healthy diversity of approaches within the technology and AI community rather than fundamental philosophical divisions. The implications are positive, showing multiple viable paths forward for addressing societal challenges through technology.

Partial agreements

Similar viewpoints

Both speakers demonstrate that technology platforms can dramatically improve efficiency and outcomes in healthcare-related services while reducing costs. They both provide concrete evidence of successful implementations with measurable results.

Speakers

– Ivar Hukkelberg
– Vegar Samestad Hansen

Arguments

AI-generated surgical scheduling increased hospital activity by 12% and treated over 1,000 additional patients through better resource utilization

A matching platform can connect elderly care needs with community volunteers, delivering 2-4 times more activities at quarter of the cost

Topics

Economic | Development

Both speakers advocate for treating diversity as a strategic business capability rather than just a social good, emphasizing the need for data-driven approaches to demonstrate and optimize diversity’s business value.

Speakers

– Manav Rihel Kumar
– Jakob Sverre Lovstad

Arguments

Diversity competence is becoming a basic leadership skill due to globalization and increasing workplace diversity

Multi-mediation analysis can connect diversity to hard business metrics like billable hours and sick leave, creating business cases for diversity initiatives

Topics

Economic | Sociocultural

Key takeaways

AI can significantly improve healthcare efficiency by optimizing resource utilization – demonstrated by 12% increase in hospital activity and treatment of 1,000+ additional patients through better surgical scheduling

Technology platforms can address elderly care challenges by matching community volunteers with care needs, delivering 2-4 times more activities at 25% of traditional costs

Diversity management requires organization-specific approaches backed by statistical analysis rather than generic solutions, with proper management converting diversity from complexity into measurable business value

Trustworthy AI implementation requires human-AI collaboration rather than full automation, with comprehensive governance frameworks including explainability, uncertainty quantification, and calibrated trust

Healthcare workforce shortages (projected to grow from 7,000 to 50,000+ by 2040) necessitate innovative technology solutions and community engagement models

AI applications in sensitive decision-making areas like immigration require secure data infrastructure, comprehensive testing, and continuous monitoring of decision-maker behavior and outcomes

Resolutions and action items

Deep Insight will continue expanding their AI-driven surgical planning solution beyond the successful Luisenberg hospital implementation

Vilmer will scale their volunteer-matching platform to address the growing elderly care gap across the Nordics

CIMA will continue developing tailored diversity management programs based on statistical analysis of organizational data

Trust research center will test different AI models with immigration directorate and courts, including data analytics of decision-maker behavior and expert assessments

Unresolved issues

How to scale AI healthcare solutions across different hospital systems and healthcare contexts

Long-term sustainability and volunteer retention in elderly care matching platforms

Standardization challenges for diversity management approaches across different industries and organizational cultures

Regulatory frameworks and ethical guidelines for AI implementation in high-stakes decision-making areas

Public acceptance and trust-building for AI systems in sensitive government and healthcare applications

Technical challenges of AI systems communicating with each other and managing emergent behaviors at scale

Suggested compromises

Human-AI collaboration model for decision-making rather than full AI automation, addressing public preference for human involvement while leveraging AI capabilities

Gradual AI implementation with comprehensive testing and feedback systems to build trust while managing risks

Secure data infrastructure allowing AI development while protecting personal information in sensitive applications

Tailored rather than universal approaches to diversity management, acknowledging organizational differences while maintaining core principles

We cannot just hire ourselves out of the problem to put more people into the healthcare sector to meet the increasing demand. Instead, we have to think differently around how we utilize the current resources we have.

Speaker

Ivar Hukkelberg

Reason

This comment reframes the healthcare crisis from a resource scarcity problem to a resource optimization challenge. It challenges the conventional thinking that more staff equals better healthcare outcomes and introduces the concept that intelligent resource management could be more effective than simply scaling human resources.

Impact

This insight establishes the foundational premise for AI intervention in healthcare throughout the discussion. It shifts the conversation from traditional solutions to innovative technological approaches, setting up the rationale for why AI-driven planning and optimization are necessary rather than just convenient.

The operating rooms are empty 40% of the time, even though the waiting lists are mile long.

Speaker

Ivar Hukkelberg

Reason

This paradoxical statement reveals a counterintuitive reality that challenges assumptions about healthcare efficiency. It demonstrates that the problem isn’t necessarily lack of capacity, but rather poor utilization of existing resources, which fundamentally changes how we should approach healthcare optimization.

Impact

This statistic provides concrete evidence for the resource optimization argument and validates the need for AI-driven solutions. It transforms the discussion from theoretical to practical, showing that technology can address real, measurable inefficiencies in the system.

Diversity doesn’t automatically yield better results. Diversity brings a lot of complexity, but also a potential to do better. So whether you deal with the complexity in a good way, and you actually get to the value that diversity can bring, that depends on how you manage diversity in the organization.

Speaker

Manav Rihel Kumar

Reason

This comment challenges the popular assumption that diversity is inherently beneficial, introducing nuance to a topic often treated simplistically. It reframes diversity from a moral imperative to a management challenge that requires specific skills and approaches to unlock its value.

Impact

This insight shifts the diversity conversation from idealistic to pragmatic, introducing the concept that diversity requires active management to be effective. It establishes the foundation for their data-driven approach to diversity management and challenges organizations to move beyond surface-level diversity initiatives.

Research shows that if citizens are asked if they would like a court decision or an administrative decision to be made by a human or an AI system or a robot, they would prefer a human… But when we also ask citizens, would you prefer a human with AI to make a court decision or administrative decision, we prefer that combination more than just a human.

Speaker

Malcolm Langford

Reason

This observation reveals a sophisticated understanding of human psychology regarding AI acceptance. It identifies that the path to AI adoption isn’t replacement but augmentation, and that public acceptance depends heavily on how AI is positioned and integrated with human decision-making.

Impact

This insight reframes the entire AI implementation discussion from replacement to collaboration. It suggests that successful AI deployment requires careful consideration of human psychology and trust-building, influencing how all the speakers’ solutions should be positioned and implemented.

For every elderly living out there, there are actually five people that want to contribute with some good moments. Some of them want to help out maybe once a month, others want to help out 12 hours a week, some do it for free, and some want to get a bit of a smaller payment… Per today, there aren’t really any specialized solutions that take advantage of this unused workforce.

Speaker

Vegard Samestad Hansen

Reason

This comment reveals an untapped resource that challenges the assumption that elderly care is limited by available caregivers. It identifies a market failure where willing contributors exist but lack proper coordination mechanisms, suggesting that technology can solve coordination problems rather than resource scarcity.

Impact

This insight shifts the elderly care discussion from a resource shortage problem to a coordination and matching problem. It demonstrates how technology can unlock existing social capital and creates a model that other speakers could potentially apply to their domains.

Overall assessment

These key comments fundamentally shaped the discussion by challenging conventional assumptions and reframing problems in innovative ways. Rather than accepting traditional limitations (staff shortages, resource constraints, diversity challenges), the speakers consistently identified how technology could optimize existing resources, unlock untapped potential, and solve coordination problems. The comments created a cohesive narrative around intelligent resource utilization, human-AI collaboration, and data-driven problem-solving. Most importantly, they shifted the conversation from ‘what we lack’ to ‘how we can better use what we have,’ establishing technology not as a replacement for human judgment but as a tool for amplifying human capability and optimizing complex systems. This reframing approach became the common thread connecting healthcare planning, elderly care, diversity management, and trustworthy AI implementation.

How can we solve the growing waiting times in healthcare when the population is getting older and we’re giving birth to fewer people?

Speaker

Ivar Hukkelberg

Explanation

This represents a fundamental demographic challenge that requires innovative solutions beyond just hiring more healthcare workers, making it crucial for sustainable healthcare planning.

How can we better utilize current healthcare resources instead of just hiring more people?

Speaker

Ivar Hukkelberg

Explanation

This question addresses the core problem of resource optimization in healthcare systems facing staffing shortages and increasing demand.

How can we make hospital planners trust AI algorithms and their outputs?

Speaker

Ivar Hukkelberg

Explanation

Trust in AI systems is critical for successful implementation in healthcare settings where human lives are at stake.

How can we address the healthcare staff shortage that will increase from 7,000 to more than 50,000 within 2040?

Speaker

Vegar Samestad Hansen

Explanation

This represents a massive projected shortage that requires immediate attention and innovative solutions to prevent a healthcare crisis.

How can we better engage the unused workforce of people who want to help elderly but haven’t been asked before?

Speaker

Vegar Samestad Hansen

Explanation

This identifies an untapped resource that could help address the care gap for elderly populations.

How do you get the value out of diversity in organizations?

Speaker

Manav Rihel Kumar

Explanation

While research links diversity to value creation, the practical mechanisms for realizing this value remain unclear and organization-specific.

How can we create business cases out of diversity initiatives that are often perceived as soft domains?

Speaker

Jakob Sverre Lovstad

Explanation

This addresses the need to translate diversity benefits into measurable business outcomes to gain organizational support.

How do we get AI systems that work at scale without losing control as they start to interact with each other?

Speaker

Malcolm Langford

Explanation

This addresses the emerging challenge of managing AI systems as they become more interconnected and develop emergent behaviors.

How can we communicate AI uncertainty and confidence intervals to non-data scientists in understandable terms?

Speaker

Malcolm Langford

Explanation

This is crucial for proper decision-making when using AI systems, especially in high-stakes environments like legal or healthcare decisions.

How do we calibrate trust in AI systems to avoid both over-trust and under-trust?

Speaker

Malcolm Langford

Explanation

Proper trust calibration is essential for effective human-AI collaboration and preventing both misuse and disuse of AI systems.