Introduction

So, thank you. I know I'm the only one standing between you and the drinks, the pizza, and the great conversations you'll have afterwards, so I'll try to be as efficient as I can with this.

Short intro, I'm John, I'm the co -founder of Design Sprint Academy, and our company helps teams make better AI decisions.

We do that by using workshops as building block for these decision systems we install inside companies and then the things that we do

we train internal AI facilitators other call them AI champions but at the end of the day are those people inside organizations who can bring together the business detect legal everyone to work on these solutions and help them make decisions fast we also work directly on AI initiatives and projects there was a

discussion about being techie or not. We're not techie, but we run these workshops to help teams identify AI use cases and then quickly de -risk solutions before they invest a lot into building them.

And then finally, for larger transformation projects, AI adoption, we run AI labs with companies. And AI labs is a place where people inside the organization, because everyone was talking, take the people with you in the AI journey.

And these AI labs, they can explore and new ideas prototype them test them and then maybe they're going to make it to the main business so that's kind of the core of our work

we're working all across the world with larger companies generally and yeah I'm going to stop here because I need to go into the content and talk about the

Where AI Ideas Come From

Bottom-up, top-down, and external inspiration

topic tonight so first I want to start with where do ideas come from in organizations right like because everyone says like we have lots of AI ideas so where do they come up so one of the things is bottom -up where employees

come up with them and they would identify low -value repetitive annoying tasks that they they want to do or also they're looking at their processes and they can identify bottlenecks inefficiencies what's not working and or skill gaps whatever they want to fix and that's one source of ideas then of

course you uh you have top down where leadership comes up with their uh brilliant ideas right they can have all sorts of mandates from something very vague like like let's do something with ai and that's it to maybe having lighthouse projects or maybe envisioning how a certain function would work purely powered by ai in a distant future right they can they can have all

of these ideas and then of course there is inspiration you look at what others do whether Whether it's the tech giants or the AI companies, you're looking into your ecosystem, see what your partners are doing, or you get inspiration from industry reports or consultants come up and say, hey, you should be doing this. So you have all of these AI ideas piling up right from all of these sources.

Why AI Idea Backlogs Become Hard to Manage

1Now, the question is, what do you do with them? Because now you're sitting on dozens, maybe hundreds of ideas, and we're having customers having these odd numbers. John, I have 127 ideas in my team. What do we do with them?

because you can only place limited bets right you can't do them all and uh the problem at this point is that these ideas are very different they come in different shapes it's a different language used uh different assumptions baked in can have broad or narrow future or immediate very abstract or granular ones strategic ideas versus very granular uh ones high risk low risk uh how was hannah saying low ceiling high ceiling right on the edge at the core so they can be all over the the place

problem driven or technology driven so how do you make decisions because i didn't do it because because you can't you can't really compare you know apples to oranges and tell yourself okay i'm now making a very smart and rational decisions when everything is so different and then what do you do when it's things are hard to compare first you get tired or exhausted of this and then you just stop evaluating them and you just pick whatever seems okay this is

the safest the simplest the easiest thing i could do so you pick that one or maybe politics take over which means the person that has the best presentation skills or the one with the more authority saying like this is what we should be doing and then that's what you're doing and finally you delegate like you know what you do it and you'll figure it out although like you know that's wrong so that's kind of what you do but at the

When progress hides feasibility and value risks

end of the day in the end you start you choose something you start building we're doing an AI project yay and then there is like this sense of progress because we have meetings we're starting to seeing prototypes things are moving there is a demo until you realize and often it's too late in the game that the

solution is not really doable with your data with your timeline with the tech stack that you have you realize that hey well this is great it doesn't really solve a real customer problem or even more like maybe customers love it but it doesn't link to business goals so it doesn't make money for us or it can't integrating to all of the legacy systems we run because our company runs on Excel whatever and then you can't bypass illegal or risk and they shut you down so all of these things can happen maybe one maybe maybe more maybe all of them

at once so the big question is what if you didn't realize this at the end or once you're invested and you're working on it but you realize this at the very

A Standard Way to Evaluate AI Ideas

beginning of your projects so imagine that you have a standardized way to collect all of these AI ideas where you don't have to do all of the thinking by yourself because people submitting ideas did their homework did a little bit of thinking and then you can only review the ideas that meet a certain quality bar because the rest were disqualified that's why we created the AI problem

Introducing the AI Problem Framing Canvas

framing canvas so we make sure that in the teams in the organizations we work with ideas are not long longer freeform and all different about their frame the same way and then when we're comparing them we can compare them across the same dimension so we can make some smarter faster decisions so this is the problem

framing canvas AI you can download it here in Miro you have the PDF version printed okay but like you can get the digital formats from from here too or

Canvas Walkthrough: Support Ticket Triage Example

you'll find it on our website but let's take a closer look i want to give you an example uh of how we're uh how we're using this so this would be on uh on miro we're supposed we're like pretending we're writing in miro so the way we're starting with ai you know like a design sprint academy

Define the proposed solution and the AI application type

we're very obsessed with problems and being problem first and so on but like with with ai this changed a little bit because people have so many ideas it's like okay you have ideas of solutions let's start there you know it's easy so we're starting here in the center like where where people can put in their own solution.

So this one would be about automatically analyzing incoming support tickets to suggest category urgency and best info routing, so ticket support system. So that's the idea about it. Quite simple, right? OK.

So the next step that we're doing, what happened here?

Then we select what type of AI application are we doing? Are we doing something that's AI assisted? That means that the AI supports the human offering maybe suggestions, insights, recommendations, but the human stays in control at all times?

Is it AI augmented where we're changing how the process works by automating part of the workflows, but humans still oversees, but does a lot less manual work? Or is this AI powered where AI has a lot of autonomy and then the human steps in only if there is a must?

So let's say this one is going to be an AI augmented where we want to automate parts of the of this support workflow then the rest

Capture the current workflow, problems, and business objectives

thing the next thing that we're doing we're looking at the status quo current workflow okay you want to do this how do things work today then you have support agents read tickets interpret the text set the category urgency and route to a team so as you can see very manual that's what they're doing during peak

volumes try ash quality varies between agents so when people do these sort of of things and they're looking at the workflows that they're trying to apply AI,

they're going to see all of the manual work that's being done, all of the repetitive tasks, all of the workarounds that the team invented to make their lives easier, right?

So that's why we're doing this, to see how things work today.

And then, of course, we're asking, good, you wanna do these AI solutions, what problems, what gaps, what needs are you fulfilling with this one? Like, what's not working?

like why do you need this solution so here we could have something try a speed and quality very right based on agents experience it's not consistent urgent tissues get flagged late simple request so cups in your time when they shouldn't duplicate tickets multi issues threads so on and so forth right like you could

have these sort of problems so already by doing this work like forces people to think not only of an idea isolated but in the context of a workflow and like to point to specific problems that they want to solve with it so that's one of

the things but then no ideas will move forward in an organization if it's not linked to a business goal a need or priority because they're just not going to get budget so that's why the next step is be can we link it to a business objective and we're linking to the business domain so this will sit in

operations customer support and then this is going to align with whatever whatever goal, KPI you already have. And by doing this already, you make this intentional, two things happen.

You find a clear connection to a business goal, which means this idea has a chance of success, or you don't find the connection, right? And maybe that's something you need to dismiss.

However, you have many ideas that are not linked to business goals. Maybe you want to revisit that, but ideally that this is why you're doing it. You want to make

Clarify users and assess feasibility (data, tech, legal)

sure it like we're addressing this business goal and then we're moving into the user part right because we want to see okay we have this ai solution who's going to use it like we're going to build it for whom so very important to talk about that and this would be in our case customer agents or support team leads who are handling these incoming tickets right and then we want

to see what is their personal problem we understand the workflow problems but like what is their their personal problems because maybe maybe they're happy triaging tickets all day long right like if they're happy to do that they're not going to use the solution as beautiful as it is so we're looking at their problems I'm going to

break something so we're looking at their problems like I waste time figuring out what the ticket is about so that's a pain or the problem is vague and I'm not sure what the customer wants right like that's annoying for me so we

We see there's also problems we can solve for the customer, not only workflow or organizational ones. So that's a good thing.

And then we're talking about AI, and AI is, we need to talk feasibility. So one important aspect of this is the how.

And I think in the previous talk, Hannah mentioned the importance of data because you can build the most beautiful solution, it will not work without proper data.

So that's like the first thing we're looking at, data. what data do we have how available it is the quality of it the freshness the accessibility so in this case we're looking data available oh we have a

historical support of tickets we have categories priorities and times cleanliness mixed all their tickets are noisy newer ones are cleaner access we can have them via our helpdesk system freshness new tickets arrive in real time so we have that and data is sensitive because it may include

personal data right so this is what we know about data and the structure is mostly text then we're looking at the technical integration and feasibility and then we can say we have existing helpdesk integration is required for our

system team doesn't know machine learning but they have back -end so so on and so forth I'm just looking at the clock and I know I have some some other things and then finally we're looking at legal and compliance right the GDPR

would be a concern because of the private data transparency is needed agents must review ai suggestions and then we have the risk of misclassification of the tickets right so those would be some of the risk we can identify now when you do this as a team doesn't guarantee

What Changes When Ideas Are Properly Framed

success by any means but it guarantees you fail early because you considered different dimensions you consider the business goals you consider your customer problems and also the feasibility right like you you already thought about all of these things so what changes for you as the person

bringing the ai idea in the organization it shows that you understand the trade -offs uh you're thinking about risk feasibility early and you connect your ideas to outcomes it's not just random thoughts and then for a leader making the call someone who need to decide where do we place

our ai bets now you're going to save time because ideas arrive to you framed in the same way you can then easily explain why you said yes to that and no to the other one then you build consistency across team right like you're scaling decision making um for them without slowing them down

and why does it matter for both because although like at junior and senior levels people are judged on uh different things as you can see here what you'll demonstrate when you come up with an idea frame like this you demonstrate that you have decision quality

From Framing to Prioritization

which is important in both in both roles so what we're doing with this with their problem framing canvas is the start we can visualize all of the ideas across a team or organization in the same way but just standard standardization doesn't

make better doesn't mean we're making the decisions so what happens next because we're connecting them to the business goals you can simply remove move those who are not helping the business in any way at the moment.

This could be one prioritization. And then the next one would be to really prioritize.

And the thing is,

in no organization, there should be this AI unicorn, one person who calls the shots, right? Because that would be very hard to do because AI touches many different functions, right?

Like you You have the leadership, you have business, you have IT, you have data, risk, so on and so forth. 1So you need the collective intelligence of these people to make those decisions.

However, you know, that doesn't happen by default in organizations. You need to design it. You need to bring them together somehow.

I'm almost done. So it has to be facilitated.

Why prioritization needs a facilitated, cross-functional workshop

And that's why we created the AI Problem Framing Workshop, which is where we prioritize all of these ideas, right?

like once we have this initial filter of ideas we we get the right people in the room for one day so think you're going to have your ai experts you're going to have the people who know the workflows your leadership the sponsor i'm going to have all of this in the room looking at the data and then they follow a very structured process and by the end of it they finish with very clear

AI use cases that tell you, okay, what is the problem for who it is? What sort of AI capability we want to use? What are the outcomes we want to get? It's still a hypothesis at this point, right?

And then you need to go through a validation process, like a design sprint, for example. But at least at the end of this process, you end up with like, this is what we're going to do with AI. You have an informed decision, and you're already talking to everyone who's relevant.

So you don't have surprises after you invest into building something, all of a sudden legal tells you it's not possible now you already got them

Conclusion

here early so that's wrapping up that's basically what we covered today the AI problem framing canvas to capture the ideas in a very structured way so you can compare apples to apples you can do an initial prioritization filtering by

business goals and then use an AI problem framing workshop with the the remaining ones, because no single role can make decisions around AI in an organization. So you bring everyone in the same decision space

so they can look at all of this information and then decide what we want to do. And I think that was my presentation. Thank you, and sorry for going over. You can connect with me here on ..