Vesperance for UX Researchers
Nostalgic sunset for a UX researcher, the reluctant dawn of an unavoidable future.
Leveraging AI in UX Research
tl;dr — I invested 2023 in a sabbatical focused on leveraging AI to enhance UX research. I’m sure you’ve seen its impressive contributions to UX research. Attempting to keep current has proven exhausting and my time invested also seem misunderstood or undervalued due to products with “bolt-on” AI. I studied AI agent swarms, which is where I believe UX research is going.
I spent 2023 on a sabbatical focused on leverage AI to enhance UX research. AI’s capabilities are certainly impressive and UX research role is forever changed for the better with it. The time I invested into AI was certainly informative but I’m not certain it was as valuable as I hoped for a couple of reasons.
First, the plethora of products with “bolt-on” AI. The process I went through to learn AI is too easily misunderstood, and thus devalued as simply using a “bolt-on” AI tool. I didn’t simply chat with a built-in AI tool, ask it to analyse data, or create a slide deck of findings. I worked with and compared multiple AIs, local LLM, and what I see as the future of UX research, AI agent swarms. They are groups of AIs completing tasks instructed by a swam manager, the UX researcher. Here’s a tutorial you can try, with a ChatGPT+ account, in “Research agent 3.0 — Build a group of AI researchers” — Here is how” video by AI Jason.
Secondly, if I’m honest, I still have reservations around deploying them. This space is changing so fast, what is learned one week is nearly irrelevant the next. The foundation of knowledge is still valuable but its “Good by” date is much shorter than I anticipated.
So, would I do it again… No, probably not as it’s likely to be a moot point, which is why I wrote this article.
A Case for More UX Professionals with AI Experience
tl;dr — Jakob Nielsen stated, “”…at least a million companies worldwide will scramble to hire UX professionals with at least two years of AI usability experience…” I welcomed that statement last June, but respectfully, I now disagree. Due to AIs rapid advancements and nearly instant concept production.
In June 2023 Jakob Nielsen wrote an article in his UX Tigers newsletter (which I recommend you subscribe) “UX Needs a Sense of Urgency About AI”, forecasting by 2025 “…at least a million companies worldwide will scramble to hire UX professionals with at least two years of AI usability experience… there will be a scant 2,000 such experts”. I was extremely optimistic when the article published, already months into my own immersive AI learnings.
But now, less than 12 months away and with all the respect I can show, I now have to disagree. The reasons I breakdown in depth below but at its core is the amazing contributions AI is already making in UX research, the ability for anyone to produce working products in minutes, and the on-going strained relationship between UX Researchers and Product talents.
Current State of AI
tl;dr — AI in 2023 is like dial-up internet but it won’t take 20 years to for broadband to reach the masses. AIs ability to go from a single sentence — HD video, music, or interactive products in under 60 seconds from hand-drawn sketch. Next… AGI!
Throughout most of 2023 I was saying, “we’re still in the dial-up stage of AI but it won’t take 20 years for most people to have access to the broadband equivalent, it’ll take months.” When I suggested months I was thinking 24, maybe 36 months. I may have been too conservative as wow have things progressed!
It’s unlikely many of us thought, just 12 months ago, a single sentence prompt could result in a one-of-a-kind HD video so lifelike you would swear it is real or from a sentence music could be produced to rival what’s on the charts. I certainly didn’t think we would have the ability to hand draw an interface, add a few instructions and AI spit-out both the UI & code to a working interactive product in less than a minute, but that is possible now.
We can expect even greater advancements moving forward as AIs have started training other AIs like Anthropic’s Claude training itself, expediting AGI (Artificial General Intelligence), which I’m leaning towards happening this year. Without argument it will profoundly impact most fields, UX included.
Digital Products in Seconds
tl;dr — tl;draw’s Make Real changes everything! Stop reading this, go check it out… but come back. Can research be justified, or should I ask will it be justified?
Have you seen tl;draw’s blog post on its beta product Make Real? If not, take a moment to watch the short video examples… It is a GAME 👏 CHANGER!!! 👏
These no-code tools enable non-designers and non-developers to quickly create, test, and iterate products in minutes. Being able to iterate on living products, not just prototypes, in real-time using a ‘no-code visual canvas’ changes the speed at which you can test assumptions and validate solutions. That will result in less incentive for upfront research and rigorous testing.
Although geared towards marketing, instant persona, competitor & market analysis can already be provided by AI products like Osum. A UX equivalent tool could easily be created, you probably know of one.
Reality of Synthetic Users
tl;dr — Using AI generated synthetic users is a thing and it’s unstoppable. AIs drawing on real user’s digital footprints from tech giants’ user data are being created and will be accepted by most researchers when it closely mimics real people. Current major players: Meta, Google, and Syntheticusers.
I know what you’re thinking. ‘You cannot have user research without real users’, and I am of the same opinion. However, you have to add “yet” to the end of that statement. This is an unavoidable reality we are seeing manifested even before 2024 arrived, dodgy I know but the horse has bolted!
There are now projects and platforms building bios/profiles of both imaginary users full of assumptions, as well as those based on real users’ history of online activities and communications in order to recreate AI archetypal users. You may have heard of them labelled as “Synthetic users” or described as “highly personal artificial intelligences”.
These synthetic users cloned from actual identities will reduce real users from being tested as they will closely represent a businesses target archetypes. As these AIs are perfected, they will eventually become accepted by all but the most stubborn researchers as they will be that good at mimicking real research participants and do so at scale.
There are 3 significant products already in the works: Meta’s Metaverse avatars, Google’s Ellmann, Syntheticusers
Meta’s Initiative
Meta’s AI Studio sandbox will allow anyone to build AIs by ingesting all of their historical digital interactions — posts, messages, photos etc. It is then used to build an avatar encapsulating their writing style, vocabulary, opinions, preferences and personality. The end goal is to mimic users for the Metaverse.
From there I can easily see future abilities to include speech patterns, knowledge areas, emotional affect, one’s reason/logic, preferences and more — to create a highly personalised AI that acts as their digital doppelgänger. Zuckerberg talks about it at Metaconnect 2023 event, covered in this YouTube video (from 5:52 to 7:31).
Google’s Initiative
Google has an initiative codenamed “Project Ellmann”. Using its Gemini AI to analyse users’ Google Photos and search histories, and then develop an overview (profile) of the user’s life. More in this YouTube video.
A leaked presentation used the example “Like ChatGPT, but it already knows everything about your life” further described in this article.
Continuing
Using these highly personal artificial intelligences, synthetic users cloned from actual user identities via Meta and Google systems could feasibly reduce reliance on recruiting real people… one day. Especially if the synthetic individuals mirror specific archetypes, instead of personas, a business targets.
As these AIs are groomed and perfected, it allows segment representation to mimic real research participants. This will likely be accomplished at a scale human researchers could never perform and the outcome of said research will eventually be in orders of magnitude in quality, all of which is repeatable at any moment, for every iteration, without repeating users, autonomously.
Challenges for UX
tl;dr — Future of UX is precarious. An already strain relationship with Product, surge in faux research catnip, and management wallowing in it, without significant change quickly we are left to 🤞
It is my opinion UX is left with a precarious future. Along with the ability to produce digital products in seconds, the potential of supplementing real users, when we are already battling a current threat to the UX researcher profession via the rise of Product & design talents’ absorbing research. A recent post by Dr Nick Fine, which is technically one of many by him and many other UX thought leaders, delves deeper into the rise of Product and the challenges of UX researchers not working well together.
The result of the shift is less emphasis on high-integrity research. Research that is thorough, rigorous, and requires an investment in time has been replaced, and now fully accepted, with cutting corners and using faux research methodologies. Unfortunately, management has been easily convinced it is acceptable research due to the research cost savings. With the faux research being the norm, I have no doubt synthetic users will be fully adopted even before they’re perfected.
To be clear, I’m not suggesting AI is going to be able to fully perform the jobs of UX researchers at a high level of integrity. I’m suggesting it doesn’t have to as the digital product culture is primed for the acceptance of AIs output when performing its version of UX research tasks. Results based on assumptions, bias, cutting-corner methodologies, etc. will tick the box.
As Dr Fine has previous stated, it will be “the automation of assumptions.”
ROI of UX
tl;dr — UX researchers have been parroting “ROI of UX” for nearly 20yrs but many benefits are soon to be a moot point. Product users shift from pre-internet generations to internet generations, less emphasis on usability and more acceptance of half-baked products when AI is included.
For years all of us have been parroting the “ROI of UX” and how it shows what our profession can do for products. Let’s take a look at a few of the statements commonly heard:
- “You can solve big problems early when it is cheap!” Not any more. To incorporate UX researcher to solve problems early is going to be cost prohibitive when you can iterate your assumptions in seconds with fully working concepts using tools like Make Real.
- “UX research conducted before design & development can help prevent fires instead of calling on UX researcher after the product is designed and built to put out fires.” I’m confident a real-time AI product monitoring system is going to be able to prevent and put out fires before we ever knew there was a fire as it’ll be able to take action on a spark. That is after the AI agent swarms performed all of its tasks upon the conception of the product or feature.
- “Without facts you’re basing your decisions on assumptions, and assumptions equal risk.” I recently came across a new talking point in this space and it makes the argument, ‘companies aren’t risk-averse, they’re loss-averse.’ That is really what UX should have been selling all along. Companies take risks all the time but they will not accept loss! They don’t like losing revenue, losing market share, and for large corporations, losing stock value. Risk is how companies separate themselves from the competition and what causes some to fail. One of the primary ways companies achieve this is by cutting costs. Soon they will be doing it by deploying AI for UX researchers task.
- “With UX, you can help reduce support tickets.” With real-time engagement, updates, improvements, education, and fixes performed by AI… how does a UX researcher help???
- “You are risking brand damage without a fully researched product.” With or without AI, I believe we are approaching a time when usability is going to be less of a concern.
Long gone are the days when users needed their hand held as they use digital products. When users no longer needed “Click here” links to use websites, we stopped using them. They also don’t need new tabs/windows opening when they click on links, as most know how to perform that action if they want it, but why we still have to deal with them is one of the last mysterious evils of the digital world.
As the pre-internet generations are decreasing in numbers our internet generations are increasing in numbers. There will already be less need to cater to the needs of former by making things usable/intuitive… and designers will rejoice far & wide as product decisions are made when they proclaim, “it looks better!”
Have you noticed someone young learning how to use something the first time? They are happy to just try to figure it out. It’s us older users who complain about usability and intuitiveness. My point being, half-baked products will be on the rise. For those who will still need the assistance, AI will be there to assist.
Most of the UI research needed has already been done anyway. We know what works and what doesn’t, we know what is ethical and what is dark, in principle anyway. Really, what we know just needs to be applied, then through the benefit of AI it iterates.
Finally, as AI becomes more prevalent, users will become more accustom to using half-baked products, if the AI built into them provides value. Hell most products are half-baked as it is without AI and miss the target on real value… other than visual catnip.
Remember the value-chain of products: Useful >> Usable >> Aesthetics Time to reconsider that, perhaps: AI >> ummm, then whatever…
Ask Yourself
I’m sure there are still areas where you contribute as a UX researcher that you cannot see an AI performing. Let me ask you, given its abilities previously described and the rapid pace in advancements in AIs abilities:
- What task/s are you performing that is irreplaceable, will continue to be valued, and remain unable to be accomplished by AI?
- What about it prevents AI from performing the task/s you see as irreplaceable, within a few months?
- Thinking about where AI was 12 months ago, and the rate at which it is evolving, where will AI be 12 months from now?
- Will your current employer see the value in you performing those task/s while AI capabilities keep progressing?
- Where will you be in 12 months career-wise? Where will I be? Where will most UX professionals be?
UX Predictions
tl;dr — Possibly up to 85% of UX researcher have <18 months. The balance in slow moving industries like government, will have more time. Some will transition into UX-AI Managers as that becomes a thing, briefly. AI agent swarms and MoE will dominate research.
I was very hesitant to write this article as I do not want to be counterproductive. I just want to share a few talking points which you may not have heard anywhere else. 12 months ago we were all telling each other, “AI won’t take your job but someone using AI will.” Perhaps in some industries but not for UX researchers.
Here are my predictions for UX researchers:
- UX research has a shelf-life <18 months for most UX researchers, maybe up to 85% of the current talent pool. [a] An article on an MIT study, Rethinking AI’s impact: MIT CSAIL study reveals economic limits to job automation stating, “findings show that currently, only about 23 percent of wages paid for tasks involving vision are economically viable for AI automation.” I’d say that is going to have a significant negative impact on us. [b] Andreessen Horowitz released an article last September, Commercializing AI in Healthcare: The Jobs to be Done, suggesting AI could perform JTBD in the healthcare sector. Will it be limited to only that one methodology? Not likely. Note: this article is a call to those who use the methodology to create a consultancy offering the service to healthcare industry.
- The remaining ~15% will likely be in slower moving industries e.g. Government, enterprise, education, health, corporates destine to fail due to moving too slowly
- We will see a rise in AI agents as AI assistants performing tasks using AI agent swarms.
- MoE (Mixture of Experts) technique will surge in use. Essentially that a technique using multiple experts from different domains working together to perform tasks.
- Some UX researchers’ will be fortunate enough for their role to evolve into “UX-AI Managers” or a similar title. I can see these researchers being the talents to fill the needs Jakob is referring to in his article but far fewer than we think.
- That too will be on a short leash, a temporary bridging role, before it is unneeded like a lamplighter, knocker-up (human alarm clocks), street manure sweeper, etc.
- Oh, and accessibility… DONE!
UX Predictions Came From???
tl;dr — My sabbatical on AI assistants, nearly instant AI product tools, observing the ever increasing tensions between Product & UX talents, and acceptance of faux research.
From my nearly year long sabbatical learning AI and how it can be incorporated into UX research. Observing the ever increasing tensions between Product & UX talents, and the continued digital talent lay-offs partly due to AI now with massive lay-offs due to AI in the months to come, and understanding the future of automation research through AI agents.
As Dr Nielsen says, ‘You don’t stand a chance against AI.’ He’s right, we don’t. We also don’t stand a chance against management’s vision of increased profits and willingness to accept the inherent risks of research short-cuts.
Using the previously mentioned Make Real, and similar AI tools allowing anyone to make the case it is ok to work with assumptions. Deploy AI agent swarms to conduct UX research with synthetic users, allowing for quick checks and iterated within a single session.
Honestly, with real-time product builds and testing results in minutes, I can see an even larger shift by the remaining team members into the thinking assumptions are fine.
With AI agent swarms providing other laborious tasks like compiling research findings, analysing very large datasets, monitoring production products in real-time for issues, engaging with the product’s users via bot to assist and educate, implementing improvements immediately, and generating & presenting anything anyone needs on any relevant UX topic that speaks to their individual needs in seconds after it is requested. What will our job be?
What else from your irreplaceable list of contributions is either not included or no longer needed as a result of what I’ve listed? I’m sure there are many tasks one could list but will they really be valued by your employer or clients?
Future Employment, or Lack of
tl;dr — Digital talent lay-offs of ~30K in January alone. Forecasts for significantly more in the coming months. UX research seen as a luxury while increase in accepting faux research.
In January alone, there were just over 30,000 tech lay-offs and another 15,000 in February, according to Layoffs.fyi. All of the big-boys are doing it e.g. Meta, Amazon, Microsoft, Alphabet, TikTok, Salesforce, Zoom, PayPal, eBay, etc. That is much lower than January of 2023 estimates between *5K — 100K but this year will be much different.
Considering how UX research is seen as such a luxury, how many new UX research talents are going to be looking for work? After a quick look into Spotify list of laid-off employees last month, I was able to find ~45.
In a recent slide deck from , it highlights the following:
- “Technology, Information and Media” (which I guess that’s what UX researchers fall under) GAI (Generative Artificial Intelligence) will likely: Augment 41% Disrupted 36% Insulated 23%
This signals a very questionable situation for us UX researchers.
Impossible, Ridiculous, Never Gonna Happen
- YOU: No way my employer could implement those changes within your 12–18 month timeline.
ME: Probably not, but AI can. It can do it at your employer and all the others who want the change… at the same time! And then lay-off the team that deployed the AI solution… - YOU: AI cannot replace the human factor, it will only be able to create rough unpolished products which won’t appeal to users.
ME: Have you seen the images and video AI is producing today from a single sentence? Have you seen the product currently being shipped by humans without UX research? e.g. ChatGPT chat UI - YOU: Where will new ideas come from for new products? How will products improve?
ME: I can envision a future where AI is going to be doing a lot of thinking for us. Most of our ideas/work we take credit for today is already at least partially involved AI, if not the idea holistically. AI will continue to at least contribute to ideas and new products. We would be putting ourselves at a disadvantage not to use AI. - YOU: How can you call it user research if you’re not using real human users, DAMMIT!!??!!??
ME: In addition to what has already been shared above, maybe we don’t call it user research anymore. Maybe we won’t need humans any more, we only need the synthetic user. I don’t know, we’ll see…
Each of those responses may sound far-fetched, my guess is you were just as sceptical about AI creativity not long ago. Considering the rate of innovations materialising, I say it is realistic to expect AI to arrive sooner than we think.
Remember, we’re still on the dial-up end of the AI progression…
Grab the Straitjacket
tl;dr — We will all have custom digital solutions provided by our on personal AI assistant. Products will no longer be created and sold to us, our personal AI will provide what is needed/wanted, in fact, we won’t even use a browser for very much longer. Our AIs will use ‘virtual AI communities’ for new ideas. Potential commerce models derive from those products and selling us “life experiences”.
Could we be on the verge of custom digital solutions for each one of us? That is something only AI could possibly deliver. If so, why would there need to be research? Where/how would UX researchers fit into that scenario? In my mind, we wouldn’t. Fortunately for us, we won’t have to try as we simply cannot in any form cater to every individual’s, individual needs, in digital products.
Here’s how our future digital world could look:
- If not already secretly achieved, AGI will be achieved.
- We will each have our own personal AI. It will fully understand us and our needs so intimately, so accurately, it can actually represent us if needed as a synthetic you.
- We will benefit from our AI helping us engage with others, educate us using a learning approach that is optimised for the individual way each of us learn. It will help identify our “tribe”, maybe a perfect partner. In turn, it will help address one of the modern epidemics in Western cultures, loneliness. It can do this as it will know us unlike we can know ourselves.
- Perhaps there won’t be fully designed & developed products offered to us in the future. Maybe products are intentionally given to us as half-baked, vague concept, that our personal AI optimises to our needs. Then products would just be delivered as a skeleton, a scaffolding, which in turn each of us customise to suit our individual needs, wants, likes, workflows, and most importantly, System thinking (System 1 or System 2) & Neuro-type (Neurotypical or Neurodiverse), while meeting individual accessibility needs.
- So how/where will new ideas come from if everything is customised by our own personal AI that caters to only one individual and their very specific needs & wants. Perhaps AIs will be linked together, exchanging ideas, improving products & systems, in “virtual AI communities”. With those new insights they would have the ability to offer new ideas to their human… “owner”???
Future Revenue Models
What future revenue models might there be, where will people work:
- Potential half-baked products I just described come to mind. I can imagine a form of sustainable commerce where there will be social networks, enterprise solutions, communication tools, entertainment, health, government, and learning platforms. However, all the other one-off independent whizz-bang ideas that built and have thrived on the internet since way back in the 1900’s, likely gone, as our personal AIs can be a substitute.
- Maybe there won’t be products created for us at all. Instead our personal AI will intuitively develops what we want, how we want/like to use it, to solve the problems we think need to be solved the way we want them solved. But, before our AI has the ability or license to develop personalised solutions/products it has to be purchased, like modules, which are available within the “virtual AI community”. This idea was recently backed up by the original author of Open Interpreter Killian Lucas, an AI model designed to automate your desktop computer, in a YouTube video “ChatGPT Code Interpreter But 100% Open-Source”. If that happens, so goes browsers.
- Then there is access to the “virtual AI community”. This is how AIs will know what to seek, what to tell its owner to buy, what others are using, have new ideas & solutions, new ways to conform to new social norms? Maybe there will need to be engagement networks where AIs can communicate between each other and this access has a cost.
- Creating and selling life experiences. There is already significant work on AI mind reading and triggering lucid dreams research. What if you could purchase a “life experience”? An experience to find out what it feels like to be a rock star on-stage with a stadium full of people screaming your name? Or, perhaps walking on the moon? Being able to… you get the point.
Systems and platforms such as those may be a way forward to continue employment for some. But that is just my simple list, other opportunities will pop-up as humans are cleaver. Humans with the power of AI should be super-human cleaver.
Finishing Up
I am notoriously wrong when predicting what’s to come. So much so, if I predict it, it’s not going to happen. So I (~70%, the balance several different AIs’ for review & artwork) wrote this article in hopes of warding off such a trajectory — you can thank me later — but then again, efforts such as this one rarely work for me, too… Oh well, I tried! 🤞
Leave you with one more image. Amazing how close the illustrator was to getting it correct, just 1 year short.
