Introduction

I want you to think about something.

AI, did you see how fast it's developing lately and how almost every week something new comes up and it's really hard to keep up with everything?

Also, how easy it has become to build stuff. I'm building my own tools with AI and although I have sort of a technical background, I'm by no mean a developer or anything

and I still can do it.

When production becomes cheap, intent becomes scarce

so it's fair to say that the AI has collapsed the cost of production like building something is cheaper and faster than ever so then if the cost of building something collapses and we can build whatever we want then what has value so value comes from what's scarce right and that's intent like what do we want to want to build and that's kind of the theme of the presentation

I have tonight how to frame your AI ideas with a problem framing canvas and I'm going to introduce you to a very specific tool that we're using with with the teams we work with so a little bit of an

About the speaker and the work

introduction I am John I am the CEO of Design Sprint Academy Dana who was supposed to be the speaker is my co -founder she's also my life partner so we're doing this together and our

work is about helping organizations helping teams make better decisions around ai and for that we use workshops as building blocks but don't think of workshop as one of events where people are just playing with post -its, we actually use workshops as a system, as an operating model inside organizations to help make these decisions.

So what we do, we train internal AI facilitators, people who are able to do this at scale within the organizations. We also run them with teams working on real AI projects. We're doing two types of workshops, problem framing and design sprints.

I'm going to talk about both of them tonight, but one of them essentially is about identifying use cases. The other one is like finding a solution and validating it.

And for other organizations who are now more intentional in 2026 around AI transformation, we're setting up AI labs. Think of them as sandboxes where they can explore all sorts of crazy AI ideas while the rest of the organization is taking care of the day -to -day business so that at one

point some of these ideas can move into the main business. So that's kind of what our work is about.

We're working mostly with large organizations from all over the world. That's why I'm on a plane or in a hotel constantly.

Luckily today Deutsche Bahn was on time so I could make it from Berlin. And yeah, that's about us.

Good, so I'm going to start my presentation now. It's going to be around 15 minutes I've done this last week so I know I can be roughly on on time so

Where AI ideas come from—and why it gets messy

where do I ideas come from I guess everyone has ideas nowadays with AI but

typically in an organization you have bottom -up where employees are coming up with ideas they're identifying low low repetitive low value type of task they see bottlenecks in their daily workflows and processes right so they're coming up

with ideas to fix or address those then you have top down more and more actually you have the leadership mandates a lot of time they're vague like let's do something with ai but sometimes they have lighthouse projects or they envision a future where eco function is almost entirely driven by ai so they have all sorts of ideas like this then of course there is external inspiration

you're looking in in your ecosystem you're looking at your partners what they're doing maybe you're inspired by reports or consultants come in and say like you should be doing this right in the end you end up with a bunch of ideas right

and then here you're sitting I don't know dozens sometimes hundreds like literally I have companies like John we have a hundred and thirty ideas we have 300 ideas which one should we do because you do have limited budget you can't

just build everything right and they all come come in different shapes different language different assumptions and then they're so different some could be very broad some could be very narrow future or immediate abstract granular some more like strategically aligned some like not some high risk low risk some start from the problem some start from from a solution so they can really be all over the place so then organizations need to decide how are we going to to do this

Why “apples to oranges” comparisons break decision-making

And you can't realistically look at all of these different ideas and then compare them and say like you're making a good decision when you're looking at apples and oranges. It doesn't work like that.

So that's why what happens in this instance, there's a little bit when you have too many things to do and to decide, you get tired. And at one point, it's like you just decide something, you pick something. Okay, I'm going to do this.

It seems like the most simple thing or the safest thing I could be doing.

Or politics take over, and whoever has the most authority or presents in a best way, not necessarily the best idea, they can say, okay, we're going to do this.

Or you just delegate, it's like you take care of this, or you know that's not exactly the right thing to want to do that.

But in the end, you choose something, you do it, and it feels like progress because you start building and then there's going to be meetings, there's going to be prototypes, there's going to be demos or like all of those things around

What goes wrong when teams build first and think later

building something and then at one point what happens and I've seen many of our clients experience this is you realize that the solutions you've built it's not doable with the data you have, with the technology you have, like it doesn't integrate, it doesn't solve a real customer customer problem, it would be like super, super nice, really good.

But like it really doesn't solve something or people are not using it. There's no clear link to a business goal. What else do we have here? And yeah, it can't integrate in the systems you have today.

So many of those things could happen, one of them or all of them, depending on on the context. Or, oh, I forgot about this at the end, like something like legal comes and shuts you down or even IT,

because I know teams that have built agents within their corporation and then when they wanted to scale IT they're just like no we can't do that so it's a

end of end of game so what if you didn't realize all of these things at the end after you invested time resources money into this but rather at the beginning

and imagine you have a standardized way to collect all of these AI ideas so you can compare apples to to apples where you don't have to do all of the thinking yourself because people submitting the ideas do at least a little bit of thinking you know before sending something your way and then you

A standard way to capture and compare ideas: the Problem Framing Canvas

only review ideas that meet a certain quality bar because the rest are being disqualified before they even get there and that's why we created this this tool

it's called the AI problem framing canvas where ideas within an organization or a team they're not longer submitted in free form or one one person writes a sentence, the other one writes an essay, but in a very structured way,

and then you can compare them across the same dimensions, making it easier to make decisions, what do you want to do with AI.

You can download this tool and you can use it yourselves. I'm just going to give you an example in this thing.

We have a Miro version, which is born, anyone, you're familiar with Miro, I guess? they're here in Amsterdam so just so you know and then there's a PDF one as well like just in case you want to use it in an in -person workshop but this is the

structure I'll be here because I do need to read certain things I hope I can see

Canvas walkthrough: AI-assisted ticket triage in customer support

on this on the screen but we're starting always with the solution and then we're asking people write down the idea you have around AI and I'm going to give you

a very simple example today that's something that comes from customer customer service or customer support and say like I have this AI solution in mind like where we automatically analyze incoming support tickets to suggest category, urgency and the best team for routing. That's the idea. Quite clear right?

Classify the role of AI: assisted, augmented, or powered

Okay but then with AI we can we then categorize or label this as AI assisted, AI augmented and AI powered and what does it mean?

AI assisted is like where the AI supports the human by making suggestions, recommendations but like the human always stays in control.

AI augmented is where a workflow is changed by AI because part of it is being digitized, optimized, right, but the human is still in the loop and overseas and then we have AI powered which is completely run by by AI, right, end -to -end.

And in this case let's say that this idea we want to to make it AI augmented. We want to keep the human in the loop, but we do want to use AI for parts of the workflow.

Document today’s workflow and the real problems

And then one important thing that we're asking teams to want to do is to think of, okay, you have this idea, but what does today look like? How do you work today? What's the status quo and your current workarounds?

And then there could be things like they support agents read tickets, they interpret free text, they set category, urgency, and route to a team. this is something they do during big volumes triage quality varies between agents and the idea

is when you do this when you write them plainly you can then see the workarounds teams come up with like or how they work today and the manual work that they're overseeing uh maybe or the only unofficial rules and shortcuts they're taking in um in their world right so you get to see all of

those those things and then also like if i have an idea or a solution around ai what problems am i solving for my team or for my organization so it's really good to want to spell those out and for

example problems we have here as we said before trial speed and quality vary based on agent's experience right or urgent issues get flagged late while simple requests soak up senior people's time so those could be some of the issues that we encounter so we start here what's the problem what's the context and what are the issues that we're

Tie the idea to business goals and user needs

solving with it and then we always want to link it to a beast to the business because if you don't link it to the business especially in large organizations then it's not going to get built right so the business domain here

would be customer support or operation and then ideally you want to link it to a goal like an OKR if you're familiar with that like whatever metric the company needs to move so this would be the goal great so we have the idea we managed to link it to one of the goals

and then whatever else we have then we're moving into the customer side of things now we have a solution who's going to use it because I've seen this scenario as well where where people are actually coming up with good solutions within their organizations, but like employees are just not using that workflow.

So who is the customer or user here who's going to interact with the AI solution? And in our case, it's going to be customer support agents or team leads that are handling these incoming tickets. What's their problem? Like what can we help them do better? So this is the customer problem, not necessarily the company problem that we're solving.

now you could hear them say I just waste a lot of time figuring out what this ticket is actually about right or the problem is vague and I'm not sure what the customer wants or I don't know what happened before this I don't have the context I don't know how to answer it or reproduce it so here we have these

Check data feasibility, integration effort, and risk early

things and then with AI what's important is also to consider very early data feasibility and risk because again I've seen agents that worked beautifully as a demo but then once put in production production they failed because data was

not there so we want to look at how clean is the data how the quality of the data do we have access to it how fresh it is how structured unstructured it is so for this one we can say okay data available we have the history of all of

the tickets we have categories priorities resolution times cleanliness it's kind of mixed access via our help desk system that we have in place for years freshness tickets arrive in real time so we do get access to it but like it might be sensitive because it includes personal

information so those are like some things we know around this idea about data and then we're looking we could connect to it so that's okay but yeah this integration is required the team has limited experience with machine learning but they're like really

good at back -end so probably they could build it so maybe initial implementation is more AI assisted at AI end -to -end and then of course there's legal compliance and trust things like GDPR for the sensitive data transparency

Transparency, misclassifications, these sort of problems. So those could be some concerns around legal compliance and risk.

What changes when ideas are framed this way

Now, imagine like somebody does this, like ideas come already form because it forces them to think about everything that's important about a potential AI solution.

Now, it doesn't guarantee success, but it guarantees that you can fail early because you're considering all of this.

So what changes this for you? As the person bringing the idea, pitching it within your organization or within your startup to your teams, you show that you understand the trade -offs and you surface feasibility, data, and risk concerns really early. And then you connect your ideas to outcomes, not necessarily to, hey, I want to build an AI feature.

Like one of my clients is saying, put a bird on it. We have everything, put a bird on it, like AI.

and then as the person making the call in the organization right like what do we do with AI it saves you time because ideas come to you framed the same way and then it's very easy for you to justify why you're saying yes to this and no to that I like you you do have this this criteria and then you build consistency consistency across team you're basically scaling decision -making across your teams in a very easy way and why is this matters to to both because

while like at a junior level and senior level people are let's say evaluated and rewarded on different things at the end of the day when you come with like a well -structured ideas like this you demonstrate decision quality and that's important for both in both parts good so what do we have what do we achieve with

the problem framing canvas imagine you're using it with your teams that's like very good in large organizations everyone is submitting ideas you know you can visualize them in a standardized way and then it basically makes better

decisions possible but it doesn't filling in this canvas doesn't mean that you've made the decision right like you still need to make the decisions but like it's a very good start and what you can do afterwards like it's easy just by

looking at them to dismiss some of them maybe they're not linked to the business goals or maybe they don't have a customer problem in there or maybe there's like issues around data so it's like easy to want to make an initial triage but you still probably have a bunch of them to do and then you you want to pick your bets.

Making better AI bets through collective intelligence

How do you do that? You need to make those decisions. And that's something that we like to say.

There is no AI unicorn, somebody making all of the decisions. Maybe you have sometimes a CEO who's already making all of the decisions, but it's very rare.

What I've seen is that AI touches many functions, different stakeholders. They all need to be involved in this decision making.

The thing is, that's why decisions are about collective intelligence, right? You need to have data people, you need to have AI experts, you need to have the business functions, you need to have risk, IT, people who are building, external vendors, whatever the context is, there's multiple people that see parts of the problem and you need their input to make those decisions.

The problem is that collective intelligence and this sort of collaboration and decision making does not happen by default in any organization that I can tell you for for sure so that's why it needs to be designed it needs to be facilitated

otherwise you end up you know spending weeks month on meetings email slack threads whatever system you have for collaboration to get to a point instead what we're doing we're designing this conversation using an AI problem frame

framing workshop.

And that's kind of a recipe, a repeatable way in which you can make decisions as a team. And I'm just going to give you a quick overview of this, and then I'm done.

The one-day problem framing workshop (decision gates and outputs)

Basically, this is a one -day workshop where you get in the right mix of people, things, stakeholders, decision makers, your experts, people who need to make decisions around those AI opportunities.

And then they follow a very structured process where we start maybe counter -intuitively because it's called problem framing but we start with solutions with ideas and then we go through some decision gates like how do they link to the business how do they link to the customer problems what's the context the workflow that we're trying to to change or the customer experience and then we finish with this ai use cases that clearly state

for who is this solution who experience what sort of problem we believe there is an opportunity to use a certain AI capability to help them achieve whatever outcome they want to achieve and this opportunity is aligned this way with our business goals now this is not a hundred percent certainty it's still a hypothesis that needs to be validated through an experiment or something but at least it's an informed decision and then you get everyone on the same page and you have looked at all of the dimensions before saying yes I think this is a good bet.

So that's that's basically how we work with with many teams starting with collecting in a standard way and then doing some triaging and then in the end using this

AI discovery pods: organizing teams around problems

I call them AI discovery pods these cross -functional teams basically it's teams formed ad hoc around problems which is not how usually organizations organizations work because they're organized by function or org chart. But this is organized around problem. You just like assemble them when you need.

And then after they make the decision, everyone goes their way and they work. So, yeah, that was that was it.

Conclusion: key takeaways

What we've learned today, AI problem framing canvas helps you capture AI ideas in a structured way. OK, so you compare apples to apples. You can then do a triage just by looking at the dimensions.

see is this linked to a business goal is it not right and then finally using a workshop to make the decisions workshop why workshop because there's not one single role that can make all of the decisions and you want to get early the

input from relevant stakeholders and that was my presentation thank you so much