Wellness
Rewards
Recognitions
Learn more about Wellable's all-in-one
employee engagement platform
“AI Brain Fry” Is Costing Companies More Than They Think
Estimated reading time: 7 minutes
In this week’s episode, Nick and Geoff dig into a BCG study of 1,500 U.S. workers that puts a name to something many employees are quietly experiencing: “AI brain fry”. They unpack how the push to use more AI agents is driving cognitive overload, decision fatigue, and real productivity losses—and why tying compensation to AI usage metrics may be making things worse. They close with a striking real-world illustration: a solo founder who built a $1.8 billion business with AI says his only reason to hire now is loneliness.
Pressed for time? Here’s a quick summary…
- A new BCG study coins the term “AI brain fry”—acute cognitive overload distinct from burnout—affecting employees increasingly pushed to manage multiple AI agents simultaneously
- Productivity gains from AI peak at two to three tools and actually decline beyond that, challenging the assumption that more AI use always means more output
- Marketing and HR are among the hardest-hit functions, with nearly one in five employees in those departments reporting “AI brain fry”
- Workers whose managers actively answer AI-related questions report 15% lower mental fatigue scores, making manager involvement one of the most practical interventions available
- Tying compensation to AI usage metrics—like token consumption or lines of AI-generated code—drives behavior that prioritizes activity over impact, accelerating brain fry rather than reducing it
Episode Summary
What Is “AI Brain Fry” and Why Is It Different from Burnout?
Nick opens with the BCG study that anchors the episode: a survey of about 1,500 US-based workers at large companies, spanning individual contributors to managers. The study focused specifically on agentic AI use — employees setting up and overseeing agents to complete tasks on their behalf — rather than the more common experience of using a chatbot to help complete a task. That distinction matters, because it means the findings apply most directly to companies further along in agentic AI adoption, even as that group is only growing.
The researchers asked employees about the quantity of their AI use, their day-to-day experiences with AI tools, and their cognition and emotions while using them. Out of that came the term “AI brain fry”: a state of acute cognitive overload that emerges when employees are managing too many AI tools, agents, and decisions at once.
Geoff highlights why the distinction from burnout matters. Burnout is a physical and emotional phenomenon — the classic experience of feeling exhausted by work. Brain fry, by contrast, is cognitively driven. It is what happens when your attention is split across too many active threads at once, even if you are not emotionally drained. Nick notes that a manager could theoretically improve one metric while making the other worse — reducing burnout by automating routine tasks, for example, while simultaneously increasing brain fry by adding more tools and agents to oversee.
One survey respondent, an engineer, captured the feeling precisely: juggling multiple AI tools at once — one for technical decisions, another for drafts and summaries — left their thinking “cluttered” and “noisy,” like having a dozen browser tabs open in their head, all fighting for attention. The tools were not making the work faster. Managing them had become its own job.
The Hidden Costs: Errors, Decision Fatigue, and Attrition
The backdrop, Nick explains, is that large companies are increasingly incentivizing employees to build and oversee complex teams of agents — and in some cases, rewarding usage itself rather than outcomes. Meta’s variable compensation structure, for example, ties part of engineer bonuses to AI usage, including the volume of code written through AI tools. Put yourself in that engineer’s position: if part of your pay depends on AI usage, you optimize for usage, not necessarily for impact. The natural result is employees launching more agents and burning more tokens than the work actually requires.
That dynamic carries a real financial cost. According to the study, employees experiencing brain fry are 33% more likely to report decision fatigue than their peers. The same group also shows higher error rates — 11% for minor errors and 13% for major errors — driven in part by the sheer number of decision points agents constantly present. There is also a retention cost: employees pushed past their cognitive limit are more inclined to quit, which means companies risk losing their most engaged, most heavily AI-adopting talent precisely because of how that adoption was managed.
A clear leading indicator emerged from the data: the number of agents an employee runs simultaneously. Productivity increases as employees go from managing one agent to two. It continues to increase from two to three agents, but at a diminishing rate. Once employees are managing more than three agents at once, productivity actually declines. Nick draws a direct parallel to decades of multitasking research showing that splitting attention across parallel tasks — human or AI-agent-assisted — produces diminishing and eventually negative returns, even when it feels productive in the moment.
Why Marketing and HR Are Getting Hit Hardest
Geoff points out that the impact is not evenly distributed across the organization. Marketing and HR/people operations show up prominently in the data, with close to 20% of HR employees reporting brain fry. Both functions tend to operate lean to begin with and are often seen as prime candidates for AI-driven efficiency gains — which means they are also the functions most likely to be handed multiple new tools and agents without much support in how to manage them.
Nick adds that the study did not project where this trend is headed, but the underlying drivers — rising token usage, expanding agent adoption, growing usage-based incentives — all point toward an increasing, not decreasing, share of the workforce experiencing brain fry over time. A number sitting at roughly one in five HR employees today is not likely to shrink on its own.
The episode also touches on a related cost: loneliness. As AI tools replace the informal, human parts of work — walking over to a colleague’s desk for feedback, for instance — employees can end up doing more work with less social interaction, even while producing more output. Nick and Geoff connect this to a New York Times piece about a single-person company generating $1.8 billion in revenue through AI-driven marketing and customer service for a telemedicine business. When asked whether he planned to hire, the founder — who had, notably, already hired his brother — said he wanted to bring on people “because I’m lonely.” It is a striking data point on just how far AI-driven productivity can go, and what can get lost along the way.
What Companies Can Actually Do About It
Geoff and Nick land on two concrete levers for employers. The first is manager involvement. The study found that employees whose managers take time to answer their questions about AI report mental fatigue scores 15% lower than those who don’t. Too many companies, Nick notes, hand employees an AI tool or account and expect them to “go run with it” without any structured conversation about how to use it well. Manager involvement doesn’t just reduce fatigue — it also addresses some of the loneliness and lack of support that comes with working more independently alongside AI systems.
The second lever is being deliberate about what gets measured and rewarded. Geoff cautions against tying compensation or usage metrics too tightly to AI adoption itself, since that tends to drive behavior toward maximizing usage rather than outcomes — and can spiral into the kind of intense, unproductive activity the study describes. Nick agrees, arguing that core business KPIs shouldn’t change just because AI entered the picture: a salesperson is still measured on bookings, a customer success rep on retention. Token usage should never become a factor in compensation or promotion decisions. What should matter is impact — if an employee can outperform using little or no AI, they should be rewarded for that; if AI tools help another employee become a stronger performer, that’s the intended outcome. The tools are supposed to serve the goal, not become the goal.
Frequently Asked Questions
AI brain fry is a term coined in a Boston Consulting Group study to describe acute cognitive overload experienced by employees who manage multiple AI tools and agents simultaneously. It is distinct from burnout: burnout is physical and emotional exhaustion, while brain fry is a cognitive strain that comes from juggling too many active tools, decisions, and agents at once, even when the employee doesn’t feel emotionally drained.
Burnout is emotionally and physically driven — the familiar experience of feeling exhausted and depleted by work. Brain fry is cognitively driven and can occur independently of burnout. An AI tool could theoretically reduce burnout, for example by automating a repetitive task, while simultaneously increasing brain fry if it adds more decisions and tools for an employee to actively manage. The two are separate phenomena that companies need to track separately.
According to the BCG study of roughly 1,500 US-based workers at large companies, close to 1 in 5 HR professionals report experiencing brain fry, with marketing employees showing similarly elevated rates. Both functions tend to operate with lean teams and are often early targets for AI-driven efficiency initiatives, which may explain why they show up prominently in the data.
Employees experiencing brain fry are 33% more likely to report decision fatigue than their peers. The same group shows higher error rates, scoring 11% on minor error frequency and 13% on major error frequency. Companies also risk higher attrition, since employees pushed past their cognitive limits by AI tool management report greater inclination to leave their roles.
No. The study found that productivity increases as employees go from managing one AI agent to two, and continues to increase — though at a diminishing rate — from two agents to three. Once employees manage more than three agents simultaneously, productivity actually declines. The pattern mirrors long-standing research on multitasking, which shows that splitting attention across too many parallel tasks produces diminishing and eventually negative returns.
Two levers stand out in the study. First, manager involvement: employees whose managers take time to answer their questions about AI report mental fatigue scores 15% lower than those who don’t, suggesting that structured support — not just tool access — meaningfully reduces cognitive strain. Second, companies should avoid tying compensation or performance metrics directly to AI or token usage, since that incentivizes maximizing usage rather than outcomes. Core business KPIs like sales bookings or customer retention should remain the basis for compensation, with AI treated as a tool that can help employees hit those outcomes rather than a metric in its own right.
Full Episode Transcript
Nick: If you have one out of five HR people at large companies experiencing brain fry, I’ve got to think that number’s only going up, not down.
Welcome to the Wellable Weekly Podcast, where we talk about key topics and trends at the intersection of wellbeing, technology, and HR. I’m Nick, and with my colleague Geoff here. Geoff, happy Masters Friday.
Geoff: Yes, my favorite time of the year.
Nick: Yeah, I’m just rocking a little bit of Masters gear, in case you can see it. They say that tournament is “a tradition unlike any other.” I feel like you and I should harness some of that spirit in our podcast today.
Geoff: Well, it’s a podcast unlike any other, so — very fitting.
Nick: Indeed. All right, so no surprise, more AI articles. I found this one really compelling because, yes, it’s an article, but it’s really about a study I found super intriguing — one of those you have to read twice just to digest it. It’s from a Boston Consulting Group study — full disclosure, BCG is a customer of ours — of about 1,500 US-based workers at large companies, spanning everyone from managers to individual contributors.
One thing worth noting: the study focused on large companies because it was specifically looking at agentic AI use — setting up an agent to go do tasks for you — as opposed to the way most people, myself included, mostly use AI today, which is more like having a chatbot help complete a task rather than an agent completing tasks on your behalf. That’s an important distinction, because some of the results need to be read with that context in mind.
They asked participants about the quantity of their AI use, their work experiences with AI, and their cognition and emotions during those experiences, all to understand the impact of AI. They ended up with a ton of insightful findings and coined a term: AI brain fry. That’s really the focus of the article and of our conversation today.
In short, big companies are increasingly incentivizing employees to build and oversee complex teams of agents. They’re doing this by measuring outcomes and AI usage, and in some cases rewarding people based on token consumption. A good example — we talked about this a couple of weeks ago — is Meta, where part of the variable compensation structure is based on AI usage. They’re tracking things like how many lines of code an engineer writes through AI. Put yourself in that engineer’s shoes: if part of your comp is based on it, you’re going to optimize for AI usage, not necessarily for outcomes or impact.
As a result, you’re probably launching a bunch of agents, using a bunch of tokens, and employees at these companies are finding themselves really pushed to the limit — specifically the cognitive limit, being pushed to manage all these things happening at once. And it carries a significant cost to the employer: employee errors, because they’re overloaded with what to review; decision fatigue, because agents are constantly presenting forks in the road; and an inclination to quit because of the pressure to just get the job done. So the financial costs of this concept of AI brain fry are pretty impactful.
Geoff: Right, and the potential costs are disproportionately impacting certain parts of organizations. As a concept, it was refreshing to see an article really focused on the human impact of AI. We’ve seen other studies and projections about how AI could advance industries and functions — that spider chart from Anthropic made the rounds, showing potential impact across industries and where we are today, with substantial gaps. Everybody appreciates that potential, but having a study that examines who today, at an individual level, is really feeling the negative impacts of trying to increase efficiency and output with tools that are still being developed and understood — that’s valuable.
And we’re seeing some really familiar faces here: marketing, obviously the world I operate in, and HR and people operations right behind it, at just shy of 20% of employees reporting brain fry. That’s clearly impacting a part of the world we know very well. I think it’s a combination of those functional groups being ripe for efficiency gains, and also that HR folks consistently tell us they operate pretty lean to begin with. Those teams have aspirations to do more with tools that let them do that with a pretty low footprint. So it’s interesting to see how different departments are impacted.
Nick: Exactly. The study didn’t weigh in on where this is headed, but if you have one out of five HR people at large companies experiencing brain fry, I’ve got to think that number is only going up, not down, because token usage keeps rising. As that creeps into more people’s lives, you’ll likely see more of what the survey already shows: individuals who endorsed experiencing brain fry were 33% more likely to experience decision fatigue. That same group is more likely to make mistakes — the study defines these as minor errors and major errors, scoring 11% and 13% respectively on error frequency. Real, real impacts.
A big driver — or leading indicator — of brain fry was the number of agents someone is running. Not surprising, but good to have that confirmation. They broke it down: as employees go from using one AI tool to two simultaneously, they see an increase in productivity. As they go from two to three, productivity increases again, but at a lower rate. Then, as they go from three to four — which blows my mind, that people are running four agents — for those managing more than three, productivity scores actually dip. And this is separate from the brain fry concept entirely; it’s just broadly about productivity, before you even get into how people are feeling cognitively or emotionally.
It’s not surprising, honestly. There’s so much literature telling you multitasking doesn’t make you more productive, and I say that as someone who multitasks and tells myself otherwise all the time. It’s the same thing with agents — you keep adding more, trying to multitask, and you get diminishing marginal gains early on, then an actual drop in productivity.
Geoff: I think it’s an important distinction the article makes, separating burnout — a term we’ve been hearing for a long time — from brain fry, which is a newer phenomenon specifically associated with overuse of AI. Burnout measures the physical and emotional dimension of distress, actually feeling exhausted from work. Brain fry is more like acute mental fatigue — too many resources being demanded, your attention so focused you’re operating beyond your capacity.
As I was reading this, I had a moment of separating my own personal experience from what others are clearly going through at larger organizations further along in agentic AI use. I’m using AI more like an advanced chat interface, getting prescriptive answers, maybe multitasking a bit more efficiently. That was a real unlock for me — understanding that while there’s more for me personally, and even our organization, to do in advancing our own AI use, there are definitely pitfalls to be mindful of as we enter what is, in some cases, a genuinely dangerous world for employee mental wellness — not just managing a team of people, but also a team of agents.
Nick: Yeah, like opening the can of topics we’ve discussed before — there are productive ways to use AI and unproductive ways, and it’s hard to tell which is which. But at least from this study, if you’re replacing routine or repetitive tasks, that’s a productive use. It saves time and lowers burnout. What I find interesting about brain fry is that it’s truly distinct from burnout — burnout is emotionally driven, brain fry is cognitively driven. They’re two very separate things, and you can see a tool improving one while lowering the other.
If you just read the headline, like I did initially, you’d think this is about burnout. Then you realize it’s something different. I think about this in the context of what we’re being sold around AI — the optimistic pitch to every white-collar worker is that AI will give you more time for meaningful work by replacing the routine stuff. But I think, partly as a natural transition and partly because companies are incentivizing usage over impact, it’s becoming a tool that doesn’t do that. It’s not freeing people up for meaningful work — it’s stressing them out, overloading them, and in some cases, if overused, actually hurting productivity.
There’s a quote from a survey respondent, an engineer, that sums up everything we’ve talked about really well. I’ll just read it: “I had one tool helping me weigh the technical decisions, another spitting out drafts and summaries, and I kept bouncing between them, double-checking every little thing. But instead of moving faster, my brain just started to feel cluttered — not physically tired, just crowded. It was like I had a dozen browser tabs open in my head, all fighting for attention. I caught myself rereading the same stuff, second-guessing way more than usual, and getting weirdly impatient. My thinking wasn’t broken, just noisy, like mental static. What finally snapped me out of it was realizing I was working harder to manage the tools than to actually solve the problem.”
Geoff: Wow, that picture of a dozen browser tabs open really resonated. We can all relate to having way too many things going on at once in a browser — and that’s within a single application, something you’re managing yourself. That really brought it to life.
Nick: In terms of what managers can do, the article had a good stat: workers whose managers take time to answer their questions about AI had 15% lower mental fatigue scores. I think that’s really important. A lot of companies are essentially saying, “Here’s an AI tool, here’s a Claude account, here’s an OpenAI or ChatGPT account — go run with it,” without much actual conversation.
We’ve also talked about the loneliness factor. You’re being asked to interact with something that seems kind of human but isn’t a person, while doing more work. What used to be walking to a colleague’s desk for feedback is now replaced by an AI tool, and that makes people lonelier — less supported. So I think manager involvement goes a long way, both in reducing burnout and mental fatigue, and in helping people feel supported by their organization and equipped with best practices.
Geoff: That’s a good point — being really deliberate about the metrics you set. You mentioned the example of usage being tied to bonuses, and that’s going to drive behavior in a way that isn’t always productive or healthy. It can get really intense rather than focused on what matters, which is business impact. There are great ways to track usage within Claude and other tools, but it’s important that usage as a means to an end isn’t the only focus — that’s how you get into these vicious cycles of intense activity.
Nick: Exactly, I’d double-click on that. For centuries, companies have had metrics — if you’re in sales, you have a bookings quota; in customer success, you have retention targets. I don’t see any reason for that to change. You still get comp and bonuses based on core business KPIs. I see no reason for token usage to become a factor in how you’re compensated or promoted. What should matter is: what do you produce? What impact do you have? If you can be the top-performing sales agent without using AI at all, good for you — you should be rewarded accordingly. And if AI helps someone else become a top performer by speeding up and improving their responses to prospects, that’s a good outcome too.
One thing worth sharing to close on — unrelated to the BCG article, though we debated leading with it instead — is a short New York Times piece I’d encourage everyone to read, about a single-person company worth $1.8 billion. To be precise, that’s $1.8 billion in revenue, so presumably a billion-dollar company. Years ago, Sam Altman predicted there would be single-person, billion-dollar companies — I’m not sure he expected it this quickly, but here we are.
This is a guy who started a telemedicine business. He’s a marketer by trade, but he used AI agents to build the marketing strategy, deliver on it, build the website, and handle customer success. He doesn’t have his own product — there’s another company actually delivering the telemedicine and GLP-1 drug business — but he runs the front end that drives all that revenue. He did hire his brother, so he’s not entirely alone. But when asked whether he was going to hire more people now that the business has become legitimately huge, his quote was: “At this point, I want to hire people because I’m lonely.” I thought that was really telling about AI, loneliness, and burnout all at once. It’s wild that one person can build a business to that scale that quickly.
Geoff: It really was incredible. I’m sure Sam Altman wasn’t thinking about loneliness when he made that prediction — he was probably thinking about enormous productivity. Who would’ve thought a billion-dollar company and the thought of social connection and loneliness would be so closely linked.
But I think that’s a nice place to put a bow on today’s pod. Thanks as always for those who are listening. You can tune in on Apple Podcasts, Spotify, or wherever you get your podcasts, and be sure to subscribe to Wellable Weekly for all our other insights. Thank you.
-
Share
-
-
-
-
