How do you rebuild the systems a business runs on without stopping the business? What does technology actually owe the people using it? And why is the least exciting part of AI the part that decides whether it works at all?
To find the answers, it helps to speak to someone who has spent their career sitting exactly where technology, process, and people meet.
Kjersti Olafsbye Unelsrød is Head of Technology Projects and Portfolio at KID / HEMTEX, one of Scandinavia’s leading retail chains. Before retail, she spent around eight years in consulting at PwC, working with international companies on business technology - and before that, she wrote a master's thesis on using technology to make planning processes faster and more precise.
That was back in 2012, and it set the direction for everything since.
KID / HEMTEX are unusual among retailers in that the group controls its own supply chain end to end. That brings enormous advantages, and it also brings the same challenge almost every established retailer is now facing: a system landscape built up over decades, holding together a business that customers now expect to behave like an app.
Because the hard part isn't buying the technology. Everyone can buy the technology.
The hard part is sequencing the change, carrying the people with you, and doing it all while the stores stay open.
We talked to Kjersti about legacy systems, fragmented stacks, what good technology actually looks like, why AI is boring before it's powerful, and what she'd tell a retailer who is too frightened to start.
Kjersti, let's start at the beginning. Where does your passion for people and technology come from?
It's a long story.
It really started with my master's thesis back in 2012. It was about helping planning processes become more efficient and more precise by introducing the right technology to support them.
And even then, I understood something that’s stayed with me since: when processes are supported by the right technology, it creates superpowers. That's where the business value comes from. Being able to understand both sides - the processes and the technology - is really valuable.
So that's where I found my area, and ever since it’s always been about the same question: how can we use technology to add business value?
That's still what my passion is about. Redirecting human energy towards the work only humans can do.
When you arrived at KID / HEMTEX, what was the challenge in front of you?
Honestly, it's a common challenge in retail as a whole.
There's a lot of technical debt. There's a fragmented system landscape. A lot of retailers have grown through a patchwork of systems that were never designed to talk to each other.
But the surroundings have changed, and the customers have changed. They expect us to know everything about them. They don't expect any gap between the physical stores and the digital side. They expect to see stock levels, buy through the app, then go to the store and pick it up.
That's the challenge. The expectations come from the digitalisation of everything - we live in a world where apps meet you wherever you need them to and give you exactly what you expect. But most retailers' technology stacks were never built to do that.
So when the stack isn't ready, what do you look for? What does good technology actually look like?
I think we need to start with the customers.
Customers expect us to be one company. They expect us to know our inventory levels in real time. They expect a fresh, well-designed experience. And they expect the people in the stores to give them good service.
So the way we approach it is: where do we want to be? What's the customer journey? Where are we now? What's the assessment? And how do we get from here to there?
Then you build the roadmap from that. We need to get real control of our inventory first. Then we go to the legacy systems and change them. Then we can give the people working in the stores much better tools, so they can do what humans do - the relational part, the judgement, handling the situation in front of them.
That's the whole logic. Let the systems do what systems can do, so people are free to give good service, remember their customers, and give proper advice about the products.
You've talked about technology creating superpowers. What should the best technology be doing for people?
Machines are brilliant at the repetitive, rule-based things. Humans are irreplaceable at the relational.
In retail, so much of the legacy problem shows up as people doing manual, repetitive, non-value-adding tasks just to hold things together. What excites me is replacing those things - and then giving people the space to do what people are actually made for.
Then there's the relational side. You have to read the emotional texture in the store and meet the customer where they are. When everything else is comparable, that's what really differentiates a brand.
Creativity is incredibly important too. AI is something you need to use, absolutely. But it creates on the basis of what it has already seen. When you want to make something extraordinary for your brand, you need people to combine different concepts into something new.
Finally, there's judgement. You can use technology for rule-based decisions, but decisions under uncertainty are different. If the data is incomplete, or there are a lot of other factors to weigh, technology can show you the data - but a person has to make the call.
So it all comes back to enabling people to do what they're best at.
I don't get excited about technology replacing people. I get excited about technology supporting people.
What are the real benefits of moving from a monolithic legacy system to modern, cloud-based architecture?
The last few decades have been about building large suites. And that's easy at the start, because everything is in one system.
But as soon as you need to add other applications, you have a huge problem. You lose speed. And you can't develop at the pace retail actually demands, because everything moves so fast. It's all about getting things done.
So modernising the tech stack is really hard, because you have to do it while operations are running. We're changing the engine while the car is still moving.
But once you're through it, and you've moved more towards best-of-breed, the applications run much faster. You can connect them to each other in a completely different way. Everything is ready to adapt to whatever you want to do next.
You can set a goal three years ahead, but by the time you finish the first year, it will already have changed, because new technology has arrived. That flexibility is the point.
How did you approach it in practice? Did you change everything at once?
We're still in that transition.
But we centralised our warehouse about a year ago, with a lot of automation. And we changed the ERP, the warehouse management system, and the automation at the same time.
So there was a lot of work on the sequencing. In what order should this happen? How do we reduce the risk? And for me, it always comes back to the people - there are a lot of people working in these programmes, and it ultimately ends with them.
The other constraint in retail is that you don't have many windows. Q4 is about stability. Don't touch it, because that's where the earnings are. Then you have Q1, where you have more space to try something new.
A lot of retailers are held back by fear. They want to modernise, but they're scared the car will simply stop. What would you say to them?
I understand it, because these legacy systems have often been there longer than the people have. You open the doors and skeletons fall out. There's always a new surprise.
But there are a lot of smart people in every company. Put them together and work out which part should be done first.
You have to divide the elephant - as we say in Norway - break it into smaller pieces. I would never advise anyone to do one huge change all at once. If you can put things into production in stages and then test through the whole value chain, you become much more confident.
And it's really a constant series of trade-offs. You're never doing this in a vacuum – you're always weighing which window gives you the lowest risk against which window is the most convenient for the business. For instance - our stores in Norway are closed on Sundays, but not in Sweden, so Sunday becomes a natural window for us to push changes in Norway. Find those windows – where you can make a change and contain the risk to a small, controlled slice of the business. That's a much better place to be than pretending you can eliminate risk entirely, because you never can. And ensure that you have a plan to reverse the changes if it doesn't work.
Everyone wants to talk about AI. How important is the data layer underneath it?
AI is really boring if you break down what you have to do before you can put a model into your stack.
You need to start with: which data do I have? Which data do I not want AI to reach? You need that governance in place before you can take the full value out of it.
And that's the shift we're seeing in the trends too. AI has moved from pilots into production. You need to be able to use it in your operations - demand forecasting, dynamic pricing, personalisation.
But it works in that order. Once you have the data right, AI is genuinely powerful for precision and efficiency.
What excites you about where retail and technology are going?
I think every big challenge in retail has an opportunity inside it.
Modernising the tech stack is really hard, but it's also the chance to free people up for higher-value work. It's the chance to rethink processes in a way competitors can't copy - because everyone can buy the same technology. The difference is how your people adopt it. That's where the advantage actually comes from.
And for retailers who control their own supply chain, there's a real opportunity to turn that into a genuine advantage rather than treating it as a cost centre, which is how it's been viewed for a long time.
So there are a lot of upsides inside these challenges. If you can get through them, that's when the real thing starts.
Final question. If I ran a hundred stores on a twenty-year-old legacy system and I came to you and said, "please help" - what would your top three tips be?
Firstly, it’s too large to handle all at the same time, so break the scope into smaller pieces and make a transparent plan handling the dependencies and the necessary order. Ensure that you have the target picture clearly defined and communicated
Secondly, include the right people on every level from board to key employees from the start. Ensure involvement and a sense of responsibility.
Finally, since you need to fix the engine while the car is driving, don’t strive for perfection and a to large scope. Ensure that the absolute must-have requirements are stable and released first to production. Find ways to do small-scale tests in production and verify the results before releasing the mass. Then build, test, and release the rest of the functionality.
