At a previous gig, the hype around generative AI was at fever pitch, and every product manager, engineer, and executive was throwing “AI” at anything and everything. There were good ideas — great ideas, in fact — but so many of them were, well, short-sighted. One example I ran into:
Say that as an account executive, you want to take over a project or account from a subordinate who has less experience with high risk projects. Everything is in a project management system which serves as the source of truth.
In a normal situation you might:
- Go into the system and assign the project to yourself.
- Reach out to the subordinate to say hey, I’m taking over this project, sorry, it’s just way above your paygrade now. (Alternatively, you put the message in the status update so they can see it.)
The system would then generate some sort of notification (sent by email, Teams, or otherwise) confirming to the two of you the transfer of ownership.
The AI-infused flow proposed by the product leaders went something like this:
- Go into the system and assign the project to yourself.
- An AI chatbot would help you compose the email message so your voice and tone is right.
- You then copy the resulting AI-generated blurb into your outgoing email to the subordinate.
There are two obvious problems here:
- You’re making the manager do the work of creating and sending the email
- Isn’t this what notifications are for in the first place?
Is this solving the problem between the manager and subordinate? What problem are we even solving here? No problem. This is a facile solution in search of a problem.
In 2025, so many of the AI solutions we’re seeing in the market are essentially the sort of “bolted on” solutions we see with the magic email generator. They’re solutions that use AI. But are they solving a real problem? And if they are, is it solving the problem better than what’s already incumbent? More elegantly?
The AI boosters would tell us, this is all part of early iterations! We need to try every little idea! The kinks will get worked out! Don’t question, just bolt it on! Be cutting edge!
When the Kindles first came out, I bought one. I wanted to embrace the future. And yes, maybe look cool.
As someone who wrenched his back in school from carrying too many books in my pack, a simple, electronic device that could store hundreds of books was a godsend.
One day I stopped into Starbucks on my way to the book club. Somehow, I fumbled my Kindle while I was standing in line. To the concrete floor it went. Those first generation models were made with fragile glass and couldn’t take being dropped on concrete. Device ruined.
Now, my book club had taken note of me waving around this Kindle, with some derision (though they liked that I could mark up passages and refer to them without covering the thing in highlighter ink). In the middle of said meeting, one of the attendees said to me, in front of everyone, “I now know how my paper book is better than your Kindle.” He proceeded to drop his book on the floor. And yes, the book was just fine.
Sometimes the cutting edge cuts you.
I’ve compared the current fervor around AI to sun-dried tomatoes. In the 1990s, they were in everything. Salads. Salad dressing. Pizza. Tapas. If you could put a sun-dried tomato in it, it went in. By the end of the 1990s, though, sun-dried tomato fervor was dying out. Nowadays seeing a sun-dried tomato as an ingredient on a restaurant menu is a rare curiosity.
Gartner describes this sort of rise and fall as the “hype cycle.”
- Something emerges as a new, cool thing.
- The people, and the money, flow into making the new, cool thing work everywhere, until it’s being added into every product despite a lack of consideration of whether it even belongs there. (Looking at you, Microsoft and your “It’s not Clippy” AI features that are as annoying as Clippy was.)
- At some point, we hit the realistic limits of what the new, cool thing can do, whether it’s ethical or moral or technical or monetary. The hype fades — something Gartner dubs the “trough of disillusionment.”
- The money flows out, the bubble pops, and the hype moves elsewhere.
- Eventually, as people figure out what the formerly new, cool thing does well, the real solutions emerge, driving the idea into the mainstream. Or, sometimes, they don’t and fade away (Web3, for example).
We know the AI bubble will pop. People will get disillusioned with AI, particularly when it isn’t solving their real problems. The hype will wear off, the money will move on to the next frothy thing, and everyone pushing hard for AI right now will be Mariah Carey going “I don’t know her” to any question about LLMs or agentic AI.
So what comes after the disillusionment?
The 2000-01 Dotcom Bust was an extinction-level event for web startups. Amazon barely hung on, and they were among the lucky survivors. By 2005, though, we were talking about Web 2.0, full of its design gimmicks (wet floors!) and pushing Javascript to its limit (AJAX!) The money came back.
What happened? Some people didn’t give up on the idea of the web, the browser, and what it could deliver that was fundamentally different from selling software on CD-ROMs. YouTube, Uber, Flickr, and many others were founded after everything fell apart in 2001.
Remember when Alexa was all the rage? Voice assistants were supposed to solve all our problems! Amazon loaded up on voice designers, but they struggled to make it a going financial concern for the company, and eventually they started laying those designers off.
We still have Alexa and other voice assistants. They’re not trying to solve All The Problems with bolt-on solutions anymore; they have found their purpose in the use cases they’re best used for. If I need to set a timer while elbow deep in making dinner, Siri can do that. If I need a quick thought recorded before I forget to write it down, the Alexa will hold onto it.
Sometimes the idea gets written off, only to re-emerge when a problem it’s perfect for solving becomes evident. QR codes were all but dead in the US (despite uptake in Asia) until the COVID pandemic made touchless, online restaurant menus a necessity.
Once the hype wears off, new tech ideas find what they are truly useful for.
After the AI bust, what will we see? It won’t be a screen to help you write an email when sending an automatic notification makes more sense. Nor will it be the second coming of Clippy in MS Word or Google.
What we will see are products that find the right balance between augmentation and replacement of humans. Much of the focus during the bubble has been on replacing people with bots. The replacement has gone, well, terribly. Between the slop content it produces and the constant inaccuracies and hallucinations, AI comes across as an eager, obsequious intern you shouldn’t let out of your sight.
What could you use an eager intern for? Or more, what could you use a pattern-matching, predictive language model for to augment instead of replace?
Let’s return to that sales organization where the boss is taking over the account from a junior salesperson. Imagine if:
- You could identify the accounts where the executives might need to get involved and flag them in advance for the account executive to watch
- You could suggest to the salesperson how to better manage such an account using known and tested best practices
- You could identify the customers that need “high touch” vs the ones where quarterly check-ins may be enough to keep them happy, so you can focus your energy where the escalations may lie
- You could help prioritize the accounts that will most help the business — and most help the salesperson obtain the best possible commission payouts
- You could compile the salesperson’s notes along with others from history to help understand the players, their concerns, and possible pitfalls along with potential areas to probe for upselling — in a way where anyone could step in and learn the needed info quickly
If the system could tell you farther in advance when and where the boss may need to step in, or help the salesperson solve their problems without involving the boss directly, then perhaps they would never need to take the account away using an AI bot to help them write that email. They’d never send that email, in fact. The process and the people have been augmented by a system that gives them control over the situation and helps everyone avoid surprises.
Out beyond the hype and the bolt-on solutions we’re already seeing what generative AI and large language models can do — spotting tumors in MRIs that a technician may miss, for example. That space out there, where innovation, augmentation, and solving real problems lies, is where the product design world, the UX community, needs to be focusing right now. The bubble’s going to burst. Money will race away from current AI solutions while the financial and technical dead ends continue to pile up. Something newer and cooler will lure the VC money away. And like a meteor wiping out the megafauna, the ones who are small, nimble, and different will have the opportunity in the ruins to evolve into the new megafauna.
It is essential design is out there in that space because being farsighted is something product designers are exceptionally good at. Every innovation lab, every design thinking workshop, every argument over product roadmaps and user pain points is about designers thinking beyond the right now. There’s a good reason designers are often seen as Cassandras in organizations — we spot the problems with facile solutions and get in fights over how anti-user, anti-business, anti-human those facile solutions are in the long term.
Design needs to stop hoping and waiting for the AI bubble to pop. It needs to start worrying about what happens once the bubble pops. UX has immense potential to, if not own, influence the narrative around AI, machine learning, and large language models.
Up until now, delivering the pixels has been a core part of a design’s job. That narrative is starting to change. When AI-based site generators like Replit can spit out an OK website for product folks to walk around and engineers to build on, the pixels will matter less and less. Design’s skills at understanding the problem and the user needs will matter more and more.
Being a designer isn’t, and shouldn’t be, about the pixels. It’s always been about the user experience perspective we bring to the table that others do not have — or don’t have enough time to focus on. We have an entire user research practitioner community to look to. Some of us have service design in our arsenals. More than a few of us have picked up product management skills.
The future of UX is in being farsighted, being small, being nimble, and making it through the extinction level events that hit our industry. We can own the future, so long as we are willing to look beyond the present.
