I live in a small apartment in Chicago with no closets, which means storage is not something I can take for granted. It is an ongoing design challenge.
So, I’ve had to get creative. Every square foot has to earn its keep, while still feeling aesthetically like “me.” I built an open-shelf “closet” in my bedroom, use narrow shoe storage for sweaters, and will add funky legs to just about any boring cabinet to make it more fun.
That’s part of why I have a soft spot for IKEA. Yes, it’s affordable and quite accessible for me, but what I really appreciate is how adaptable it can be. The most satisfying projects are not always the ones where I build the piece exactly as intended. They’re the ones where I figure out how I can manipulate a piece to serve the space I actually have.
Working with AI has felt surprisingly similar. AI can make aspects of our work more efficient and easier, but like designing a space, success isn’t just about following the instructions for a prefab cabinet that looked good in a showroom. It relies on understanding the space you’re working in, knowing what the pieces can and can’t do, and being willing to assemble (and hack) them creatively.
Start with the Space, Not the Furniture
When you live somewhere with limited space and storage options, you learn not to start with the catalog or rigid plan.
You might find a cabinet you love, but that doesn’t mean it will work in your space. Before buying anything, you have to understand the actual constraint. Is the problem that there is no place to store coats anywhere near the front door? Is it that I have too much dang kitchen equipment? Is it that pile of “I’ll deal with this later” items that somehow appears whenever life gets busy?
You need to understand your constraints when working with AI too. It’s tempting to just start with the tool and ask: “How can we use AI here?”. That question will send us chasing novelty instead of function. Start with the actual problem, and weave in the delight or novelty later.
Ask the important questions, like: Where are people losing time? What work is repetitive or difficult to start? What information is scattered across too many places? What decision needs better input? What conversation keeps happening without enough clarity?
AI is most useful when it is pulled into a real need, not pushed into a workflow because it seems like the thing we’re supposed to be doing.
Inventory the Pieces
Before assembling IKEA furniture, there’s usually a moment where everything gets spread out on the floor; panels, screws, dowels, hinges, a tiny wrench…and at least one piece that looks important but is unsettlingly mysterious.
Take inventory in a similar way before you jump into using an AI tool or workflow. Sure, Claude, Codex, or whatever AI tool you prefer, can do a lot of useful little jobs. It can summarize messy notes, compare options, or help you figure out why an idea feels half-baked. Those are good pieces to set out on the floor, but they are not all of the pieces.
AI does not automatically know your users, your team dynamics, your business constraints, your project history, or the reason a certain decision is politically sensitive. It may not know which tradeoffs are acceptable and which ones are nonstarters. It will not reliably know when the technically correct answer is actually the wrong answer for the moment.
That missing context matters. If we don’t notice what’s missing from the box, we may not realize there’s a problem until halfway through assembly.
The Instructions Are a Starting Point
IKEA instructions are useful, but they assume you are building the thing exactly as designed. Sometimes that’s exactly what you need. However, other times, the standard configuration doesn’t quite work.
That’s where a little hacking needs to come in. There’s a whole genre of people using IKEA furniture for purposes it was not originally designed for. For example, a series of bookshelves can become a room divider. A kitchen cart becomes bathroom storage. A cabinet system becomes a custom-looking built-in. The parts were designed with one use in mind, but once people understand their shape, constraints, and possibilities, they start using them to more creatively meet their needs.
AI can have flexibility like this too. There are the obvious use cases: Summarize these meeting notes, write me a draft email response, review my message for any inconsistencies, etc. Those are all fine places to start. Like a good IKEA hack though, the real value often comes from understanding the pieces well enough to use them in a new way that is better for your specific scenario.
If you’re using AI to summarize your meeting notes, make sure you’ve set it up with a template, so it understands what output is most important to you. Use those notes to build up a knowledge base that can be prompted throughout the lifecycle of the project. Give other people access to it, so you’re all on the same page, all the time.
Try to go beyond the surface-level tasks and understand what AI can actually be good at, what it’s bad at, and where it can be combined with human input to make something better suited to the actual need.
Try It in the Space Before You Tighten the Screws
In a small apartment, you can measure carefully and still not know whether something works until you experience it in the room. It’s about more than just if it physically “fits,” but also how it feels and if it relieves or creates any operational friction. A cabinet might technically fit but make the entryway feel cramped. Sometimes the only way to know is to move things around, live with it for a bit, and adjust.
Working with AI is not much different. The first output is rarely the final answer, and it doesn’t need to be. One of AI’s strengths is how quickly it gives us something concrete to react to and then iterate upon. That first version helps reveal what we actually want (and don’t want). It shows us what context is missing, and gives us something to disagree with.
Like leaving screws a little loose while assembling furniture, we need to keep the work flexible long enough to see whether the pieces actually line up. Add a new constraint and ask for another version. Bring the output back to your team and see what conversation it opens up, then adjust based on that.
AI works best as part of an iterative process: gather input, generate, inspect, adjust, and generate again. Don’t tighten the screws until you know the whole thing fits.
A Good Hack Still Has to Hold Up
If my sweater storage looks nice but makes it impossible to actually get to my sweaters, I have not solved the problem. If a cabinet has delightfully fun legs, but wobbles every time I walk past it, the legs are not delightful enough to save it.
AI experimentation needs the same standard. A clever use of AI still has to be accurate, functional, and secure. Speed, polish, and novelty do not negate the need for the solution itself to be correct. This is especially important because AI output can look more finished than it really is.
That does not make the tool useless. It just means the work still needs plenty of human inspection and adjustments. AI can help us move faster and think differently, but it does not own the outcome. We do.
We decide what to trust, what to change, and when something is ready to use. If something wobbles, don’t ignore it.
Build for the Space You Actually Have
My apartment still does not have closets. That constraint has not gone away (and it’s not going to because cheap rent, baby!). However, working around that constraint has given me a fun problem to solve, and made me more creative in how I approach spatial design.
It has taught me to look for flexible pieces and test ideas in real life. Sometimes the best solution is not the one pictured in the showroom that can be assembled straight from the box. It might be the one that requires me to make too many trips to Home Depot, but eventually results in a functional piece that actually fits where I need it to.
AI tools can work similarly if we use them similarly. The best results do not come from treating AI as a magic set-it-and-forget-it solution. They come from understanding the problem, experimenting with the pieces, and staying involved long enough to shape the output into something useful, thoughtful, and sturdy enough for real life.