The interface trap: Why your AI adoption is failing (and how to fix it) | Kaisa Martiskainen

Is your team actually using AI, or are they just playing with it?

In this episode of Human X Intelligent, host Madalena Costa sits down with Kaisa Martiskainen, AI Operations lead, to uncover the hidden gap in corporate AI adoption. While usage metrics might be up, true understanding is often lagging. Kaisa explains why providing access to chatbots isn’t the same as building capability and how the 'Interface Trap' prevents organizations from seeing the real value of AI.

In this episode, we explore:

• The missing conceptual layer: Why mental models are more important than tool proficiency.
• The interface trap: How limiting AI to a chatbot window narrows your strategic vision.
• Human learning vs. Machine speed: Why humans need friction and failure to truly 'get' AI.
• Predictors vs. knowers: Understanding the three foundational concepts every employee needs before their first prompt.
• Beyond surface level: How to transition from "interacting" with AI to 'integrating' it into your organizational DNA.

If you’re a leader, manager or individual contributor feeling overwhelmed by the AI hype, this conversation will help you shift from reactive usage to intentional system thinking.

Connect with Kaisa Martiskainen:
→  LinkedIn: www.linkedin.com/in/kaisamartiskainen
→  Substack: https://mamaknowsai.substack.com

Human × Intelligent is a podcast at the intersection of design, AI and human agency. Hosted by Madalena Costa.
→   humanxintelligent.com
→   https://www.linkedin.com/company/human-x-intelligent/
→   https://www.instagram.com/humanxintelligent/
→   https://www.instagram.com/designwithmaddie/
→   https://www.linkedin.com/in/madalenafigueirasdacosta/

Guest bio
Kaisa works at the intersection of technology and human understanding. She helps organizations and individuals understand how to work with artificial intelligence in practical, thoughtful ways, focusing not just on tools, but on how technology changes the way people think, learn and make decisions

Chapter timestamps

• 00:00 – Use vs. Understand
• 01:09 – Real AI Adoption
• 02:49 – AI Mental Models
• 04:29 – The Metrics Myth
• 05:35 – How Humans Learn AI
• 07:21 – 3 Rules of Prompting
• 10:06 – The Interface Trap
• 12:38 – Access ≠ Capability
• 15:47 – AI as a Collaborator
• 17:28 – The Teaching Problem
• 19:54 – A Learning Challenge
• 21:43 – The "Ideal" AI Org
• 24:32 – The Best Investment
• 25:49 – Where to Learn More
• 26:57 – Final Takeaways

Concepts to explore further:
→ AI vs Automation
→ AI as a multiplier (not a fixer)
→ Problem-first vs technology-first thinking
→ User trust in AI systems
→ AI readiness (data, workflows, goals)

👉 Subscribe for more conversations on AI, product design and human-centered technology.

Transcript

SPEAKER_00 0:08

So most organizations today feel like they're adopting AI. People are using tools, prompts are being like shared, and usage metrics are going very up. But there's actually a gap that's much harder to see. Because using artificial intelligence is not the same as actually understanding artificial intelligence. In this episode of Human X Intelligent, I'm joining by Kaiser Martiskinen, a technology professional with over 15 years of experience across software development, products, and large-scale organizations. Currently leading AI operations and cross-team initiatives at SOP customer experience, Kaiser works at the intersection of systems thinking, organizational design, and human behavior, helping teams move beyond surface level adoption and build a more thoughtful relationship with artificial intelligence. Together, we will explore the missing conceptual layer in AI adoption, why access to tools don't create like real understanding, and why the future of artificial intelligence in organizations is not just a technology challenge, but a learning one. Kaisa, welcome to Human X Intelligence.

SPEAKER_01 1:15

Thanks for having me, Madalena.

SPEAKER_00 1:17

Do you think most organizations are actually adopting artificial intelligence, or are they just interacting with it?

SPEAKER_01 1:24

That's a really great question. I think most organizations are still just interacting with AI rather than fully adopting it. I also think many of them are reacting to AI without fully understanding the larger system that they are a part of when they're using AI. So, like you said, companies are experimenting with tools, they're running AI pilots and encouraging people to start applying AI in their work. And that's a good thing, but it doesn't automatically change how work is done. It doesn't quite go far enough. And real adoption really starts when people understand how AI affects their daily decisions and responsibilities. And in many organizations, right now, AI is still treated as something external. It's kind of like this layer that sits on top of the regular work instead of becoming part of how that work is done and how it's designed.

SPEAKER_00 2:22

That's interesting, especially what you said about like externally using these tools like a layer for the work and not actually adding to the process. Just it's there. They're just adding. And it it really sounds like that we are calling like adoption might be more like surface level or like the layered level, like you said, than we think. So if that's the case, then something deeper is missing, right? Might be like underneath that interaction. What do you think is this missing conceptual layer in how organizations approach artificial intelligence adoption today?

SPEAKER_01 2:57

So the missing layer, in my opinion, is understanding AI as a system that you operate within, not just as a tool that you use, but how can we understand this system? Because systems are often complex and it takes time to understand them. So before anyone can learn anything new, before we can learn something new, something as complex as AI, we need to be able to build some kind of a mental model for this new thing that we need to learn. And a mental model is a concept from education, and what it is is basically a simplified representation of how something works in real life. So in the case of AI, people need to understand how AI produces answers, what are the advantages and at the same time the limits of the system? People need to understand what responsible use looks like. So not just what use looks like, but how can we use AI responsibly? And also another factor that is really beneficial for humans to understand is that where does the human responsibility go when we're within the system? So when AI becomes part of our work and we're starting to use AI output in our work, who is responsible for that, especially in these situations where something goes wrong? So many organizations teach employees how to use specific tools, but they spend less time explaining how the system itself behaves. And without this conceptual understanding, it is very difficult for people to use AI intentionally.

SPEAKER_00 4:36

I liked how you framed that. Like the being able to build this mental model that comes from education. It's like the simple representation of what we're doing, like how do we produce the timing, what is responsibility? And I liked how you went right away to who is responsible for these outcomes. Because that would explain also why, even with high usage, something feels very inconsistent, right? It almost feels like activity is being mistaken for understanding. And why do you think that there is such a huge gap between artificial intelligence usage and the real understanding inside organizations?

SPEAKER_01 5:15

I'm sure there are lots of different reasons, but I think one of them is certainly that using tools is quick, but developing understanding takes much longer. So when people are using tools, AI tools, they can produce summaries and drafts and they can analyze things in minutes or in seconds. And that speed, the experience of the speed, creates this imprint that learning has already happened. But understanding happens through a lot of other things, such as repetition, comparison, feedback loops, and ultimately also failure. So many employees know how to generate an answer, but they're maybe they're not sure or they're still learning on how to evaluate that answer. They're learning how to improve their prompts or their instructions to the AI. And they're also learning to recognize situations where the system might actually be struggling. So that learning process is much slower than just using a new tool. And the reason for that is that humans learn slowly. Like we learn through friction. So the organization might be seeing people trying these tools, which it interprets as usage and adoption. And this might happen. Like the user statistics, they look good, but the story that they don't tell is that maybe the deeper understanding is not there yet, although people are playing around with the tools.

SPEAKER_00 6:52

Those gaps and that learning curve and everything that you share doesn't feel very accidental, right? It feels like uh it might come from artificial intelligence as introduced in the first place, like how we did it. Almost like we are um starting from this interface, and instead of thinking behind it, like having a logic, having the right questions, or at least some questions sometimes. What do people actually, what do you think that people actually need to understand in the first place before even using these tools and prompts even that will help them to start to make sense of what is happening and of what they're doing?

SPEAKER_01 7:29

It all comes down to the system and understanding the kind of system that they are working with. So um I I guess there could be multiple different points where someone could start, but I think there are three simple ideas that give a good foundation for understanding the system. So, first, AI predicts likely responses based on patterns in the training data. And it's easy for people to think, and it's easy for us to think that AI knows, but it's important for us to know that it is actually just predicting. Like it does not know, it just predicts. And second, the quality of the result depends really heavily on the information that we as the users provide. So context matters a lot here. So two different people asking the same or a similar kind of question from the same system might get a completely different answer because of the way they are prompting, based on what kind of wording they use, what kind of context they give, or even based on what the system already knows about them and is trying to kind of please them in a way with that answer. And then, third, the system can produce a confident answer, even when the answer could be completely wrong, or that it definitely needs some verification. So when people understand those behaviors, they have a little bit more clarity on what AI actually is and how they should think about it. And now they're able to give clearer instructions, provide examples, and maybe they're able to also review the output more carefully. Understanding AI definitely doesn't stop there. There's a lot more that I could talk about. But I think that's a good starting point. And it's something that is not, it's not always there when we when we're getting AI enablement. But once people get that basic foundational understanding of the system, that's usually the point when the tools start to feel reliable to most people.

SPEAKER_00 9:34

And what you're describing, it's basically like a mindset shift or a shift in perspectives, I would say. But most organizations simplify that complexity into something much more approachable. And we've discussed about this, like the chatbots, for example. And I wonder if that simplification will come as a cost or with a cost. And the question I have for you is based on this because how does this limit or how does limiting artificial intelligence to a chatbot shape or a shedbot feature or functionality, or whatever you want to call it? The way people understand or even or misunderstand what artificial intelligence actually is.

SPEAKER_01 10:14

Yeah, that's a really interesting question. And and this is something that I often see, especially at large organizations. So maybe some small startups, they might have a different tool stack, but this is definitely true for most big organizations. So there's a lot of talk about how people need to start applying AI in their work, but then in the end, they're only given a couple of chatbots to interact with. And um, I feel like there is something that I like to call the interface trap. So when the chat or the chatbot is the only AI that you can interact with, it at the same time, it makes AI feel a certain way. So it might make AI feel like a person, it might make it feel like a search engine, and it might make it feel like this helper that you have. And um something that um when you think about AI that way, if that's the way you experience it, a lot of things stay hidden. The larger system and it's all the possibilities stays hidden, the integrations that are possible, the capabilities to automate workflows, and and that specific part is like the real value that comes from AI. So all of that stays hidden if the only thing that we have is chatbots. And they can be, yeah, and and and I saw you had uh you had a different episode about this as well. It's such an interesting topic. And uh they uh chatbots are a really useful starting point, but just focusing on them, it really narrows the conversation. So many of the most valuable uses of AI happen outside of that chat window. And if the focus only stays on conversations, then I feel like organizations might be overlooking opportunities to actually improve their everyday processes. So the interface makes the technology approachable, and in some cases that is exactly what's needed, but it doesn't show the full picture.

SPEAKER_00 12:12

That that is kind of like a paradox, right? AI is becoming very easy to access, but it's not necessarily easier to understand. Because even with the same tools, like you were saying, or the same prompting, or the same usage but different ways, people seem to be very this very different outcomes. Exactly, like you were saying. But why do some people adopt artificial intelligence much much faster and more effectively than others? And even when everyone has access to the same tools.

SPEAKER_01 12:46

Yeah, no, that's um that's another really interesting and really complicated thing in AI adoption. So I think there is this myth that is that access equals capability, or access equals easy adoption. So they say, okay, now that vibe coding is basically free, everybody will just vibe code. And um, we've seen that uh many people vibecode, but it is certainly not everyone. And uh and the same applies for all the other AI tools as well. So access doesn't equal capability. So some of the some of the different kinds of layers that I think are underneath are again the mental models that we were discussing earlier. So certain people who are, let's say, programmers or they are otherwise technical, they might have already this existing mental model, which uh is pretty similar to what you would need for AI. So for these people, it's easier to just go ahead and try it out. And then also, while people talk about vibe coding as this solution to everything, the the reality is that, for example, programmers they see problems that probably look like software, and product managers, they see problems or they come up with solutions that probably look like products and it works like that for everyone. So then for people whose problems have a different shape, then for them, the tool is not immediately useful, even though they have access. So that just shows that kind of your position, your background, your existing mental models, they do matter in this case. But then at the same time, it it certainly doesn't mean that only technical people are the ones, the early adopters. But I think um curiosity also plays a major role. So many early adopters of AI are people who really enjoy exploring new tools and testing ideas, regardless of their job title or education. So I think this is um and this is actually a great thing. But for organizations, that means that people will start from many different places. So someone will move quickly into building solutions, while others will begin by learning how to use the system safely and confidently. And um that that is a challenge for company companies because they need to design these learning journeys that support the different starting positions, and they also need to be realistic with their expectations because not everyone will be able to advance at the same pace.

SPEAKER_00 15:31

So it's basically not about or not just about access, it's about like how t people think about or how they approach these systems, almost like some people already have a mental framework that fits artificial intelligence better, versus there's people who still need to do this work, this job. But what kind of mental models do people need to actually work well with these AI systems?

SPEAKER_01 15:55

So I think people need simple ways to interpret what they're seeing and experiencing when they're using these tools. So an example, kind of mental model would be that AI is a collaborator with limits, AI is a pattern generator, and that AI is actually a system component instead of being a solution. AI shouldn't be this hammer, and we shouldn't look at everything as nails when we when we use it. So that would be one example of a mental model that people can use to shape how they think about AI and how they think about their usage. So people need this kind of models to help them so they can decide when to trust something, when to question something, and when they should just ignore something. So these I these types of ideas actually guide people's everyday behavior. And then over time, that kind of habits will build confidence and consistency.

SPEAKER_00 17:01

Yeah. And basically, if that's true, then this becomes less of a tooling problem and more of a teaching problem because we are asking people to think in new ways, but we are not necessarily teaching them how. And I think that's very important because we can say use this, use this, use this, but if you don't teach certain people to do it, it's gonna be hard because, like you're saying, there's different types of mental models and how we can use them. So, why doesn't technical expertise automatically translate into effective AI education inside organizations?

SPEAKER_01 17:36

Yeah, so in in many organizations, I can see that the technical AI experts are sometimes put in charge of giving some of the AI enablement to other employees or or to mentor the ones that are maybe a bit hesitant about AI, but it's very I've seen this thing that's very experts-centered. And I think there are some benefits to this kind of approach, but I think in some other with some other aspects, I don't think it's necessarily such a good idea when it uh when it comes down to actually helping people with that system view that they need to get. So I think technical experts understand the technology very well, but teaching something like that to other people, it requires different kinds of thinking. So teaching AI effectively requires zooming out and looking at the big picture, not necessarily focusing on the technical details, which is what technical people love to do. And um, it also requires advancing slowly, step by step. So basically respecting the human learning process. And um also it's really important to connect the technology to real situations in everyday work. And I feel like in a lot of companies, we're we're not there yet. We're basically on the theoretical layer talking about what happens, how we build agents and what happens inside them. And it's um for a certain slice of the audience, this is useful information and they're able to digest it. But then there are also other people who it just doesn't resonate with at all. Technical expertise, it explains how AI works, but uh not how people come to understand it. So I'd say when a training is grounded in everyday work, people will actually start to understand faster and it's easier for them to start using the system more confidently.

SPEAKER_00 19:36

And I do believe that this distinction is really important because knowing something very deeply, very inside of us, and being able to help others understand it are completely different skills. And that actually makes me wonder if we've been framing artificial intelligence adoption the wrong way altogether. Because why should we start thinking about artificial intelligence adoption as a learning challenge instead?

SPEAKER_01 20:04

So, because adapting AI requires people to basically develop a lot of new habits that are not necessarily needed with the software tools that we've used in the past. So that's a big difference. So people need to learn how to outright clear instructions, how to review results really carefully, and uh decide when the system should or should not be used. So these are new behaviors that uh we have never had to have in the past, or I mean, at least largely when we're using our regular software tools. And these are behaviors that develop over time through practice and feedback. And I feel like organizations that succeed usually create regular opportunities for people to learn together. So not just you being left alone with your chatbot and doing your experiments, and then at some point we'll see what happens, but actually working with this in teams and with other people. So sharing examples and discussing mistakes and really refining your approach, this completely new approach to work. And that is the kind of steady process that normally builds confidence and trust. And that's the way that adoption becomes stable when this type of learning is made part of every team and every employee's routine.

SPEAKER_00 21:33

It's like success stop being about or just about usage or efficiency. It's kind of becoming something much more structural, right? So, what does meaningful artificial intelligence adoption actually look like inside of an organization that quote unquote gets it?

SPEAKER_01 21:50

I think these types of organizations start with the system overview. So they help people understand how AI fits into the Organization's goals and workflows. And it's not just that everyone needs to use it whatever way they do, but really they show how it ties into their own goals and ambitions and the actual practical workflows. So then these companies also provide role-specific guidance on how to use AI in the actual work instead of just these one size fits all technical trainings. Then meaningful adoption normally starts to happen when people use AI regularly for defined tasks. So not everything, but certain group of tasks that are clearly defined. They document what works, they share examples, or they do knowledge sharing within the organization, within the team, and they build this shared knowledge about what the technology can actually do. The company is really intentionally and thoughtfully redesigning work, not just adding AI on top as this extra layer that we already talked about, or forcefully pushing AI into every single area of that work. So they're really having to analyze their own processes to see: okay, here is where we can see AI really helping us, whereas then there are areas where maybe that's not the case. And uh in these organizations, people also continually assess whether the AI-generated output um is is good, or even if they were validated. And uh there are also discussions on where automation would be helpful, where it should be avoided. And this is something that happens on a regular basis. And uh there, I think people understand the difference between a tool's convenience, so the speed or the efficiency that we can we can get versus its reliability, somewhere with the speed is a real benefit, and then there are areas where it it really isn't. It can actually work against us. And once a company is doing most of these things, I think AI has become a normal part of how an organization operates.

SPEAKER_00 24:09

At the end of the day, it's always deeper than adoption, right? It's about, or it's almost like, and bear with me, like a transformation on how people think or decide or even interact with these systems, like you were saying. And then that brings it back to the human side of all this, right? So, what is the one thing organizations should invest today if they want their people to not just use AI but truly, truly understand it?

SPEAKER_01 24:39

So I think uh designing learning environments where thinking and understanding can happen. That's that's basically the biggest investments that organizations should do. Because if people don't understand a system, there is no tool that will fix that. And uh companies need to help people learn how to think with AI by investing in these environments where where people can actually learn through real work, by doing their everyday work. And that also means giving employees time to experiment with their own tasks and review the results and discuss what they learned with others and maybe kind of learn together in an iterative way. And uh, it also means recognizing and understanding that all of this develops gradually. It doesn't all just happen at once. We can start using the tools like this, but the real learning and understanding, since we're humans, it will take some time. And if people understand the system that they're working with, the technology actually becomes useful. But if they don't, even the best tools that they have at hand will just remain unused.

SPEAKER_00 25:56

We are getting to the end of thank you. We are getting to the end of the episode. But before we do, I would like to ask you what advice would you give to people that are listening and would like to learn more about the subjects that you discussed here today?

SPEAKER_01 26:11

Yeah, so I think there are lots of really good resources out there, and not only it it doesn't need to be these technical trainings or learning to how learning how to use a specific tool, but maybe if you have a certain area of interest, try looking for podcasts, try um try looking for talks or courses on on LinkedIn and other platforms. There are also a lot of interesting tech nonprofits that are writing on Substack that are doing podcasts. So, yeah, if there's a if there's a specific thing that you're interested in, just um try looking for resources. There are plenty of things available, whether in the podcasting world or places like Substack, even social media. So that's where I would start.

SPEAKER_00 27:04

Kaiser, thank you for this conversation. What really stands out to me is that like the challenge of artificial intelligence in organization isn't just about access, like we said, or like you said, or even if capability while it's about this understanding. And understanding isn't something you can deploy, it's something you have to build through people who think, who do, who they'll who learn, how organizations support that process. And if artificial intelligence is going to reshape the way we work, then the real question isn't just how we use it, but whether we truly understand what we are interacting with. So thank you so much for joining us on Human X Intelligence and for everyone that is listening. And if you enjoyed this conversation, don't forget to follow UNX Intelligence for more conversations at the intersection of people, systems, and intelligent technology. Thank you once more, guys.

SPEAKER_01 27:58

Thanks so much, Madalena. It's been a pleasure.