AI doesn’t simply add work; it modifications work in ways in which are actually empirically simple. The HBR article “AI Doesn’t Reduce Work—It Intensifies It” validates what I known as the “AI Tax” practically a 12 months in the past: AI will increase the amount, velocity, and ambiguity of labor until organizations deliberately design in opposition to that end result.
When the Analysis Catches Up with the Ground
Within the AI Tax put up, I argued that AI doesn’t arrive as merely as a productiveness dividend; it arrives as six classes of latest work: juggling and gear sprawl, vetting, knowledge readiness, relevance and security, the burden of failed initiatives, and perpetual studying and relearning. These classes emerged from conversations with groups already utilizing AI in apply, customers toggling amongst instruments, reconciling outputs, and cleansing knowledge moderately than doing the “higher-value” work they have been promised.
The HBR piece by Aruna Ranganathan and Xingqi Maggie Ye affords a uncommon longitudinal have a look at that actuality, following roughly 200 workers at a U.S. tech firm over eight months to see how generative AI truly modified their work. Their conclusion is blunt: AI instruments didn’t scale back work; they “persistently intensified it.” Staff labored at a quicker tempo, took on a broader scope of duties, and prolonged their work into extra hours of the day, typically with none supervisor asking them to take action.
Put merely, the research offers the ethnography for the AI Tax’s classes of labor.
Three Methods AI Intensifies Work
The HBR analysis identifies three major patterns of intensification that emerge as soon as AI instruments transfer from demonstration to each day use.
- Activity enlargement
As soon as AI is offered, folks don’t simply do the identical work quicker; they start to do extra varieties of labor. Product managers and researchers start writing and reviewing code; workers tackle duties that may beforehand have required new headcount; and people reclaim work that had been outsourced, deferred, or just prevented. At one stage, this might be perceived as empowerment. A deeper dive exposes engineers who discover themselves mentoring colleagues on AI-assisted code, reviewing a flood of partial pull requests, and fixing low-quality “work-slop” that arrives of their queue dressed up as completed work - Blurred boundaries between work and non-work
AI makes it straightforward to “simply strive one thing” within the margins of the day: a fast immediate throughout lunch, another refinement earlier than heading to a gathering, a late-night concept examined in mattress on a cellphone. These micro-sessions don’t really feel like additional work, however over time, they erode breaks and restoration, making a steady sense of cognitive engagement. Employees within the research reported that, as prompting turned their default throughout downtime, their breaks now not felt restorative. - Elevated multitasking and cognitive load
Staff run a number of AI brokers and threads in parallel, let AI generate various variations whereas they write, and preserve half an eye fixed on outputs whereas making an attempt to give attention to one thing else. The presence of a “accomplice” that by no means will get drained encourages fixed context switching: checking, nudging, re-prompting, and reconciling. The result’s an ambient sense of being at all times behind, at the same time as seen throughput will increase.
For those who learn my AI Tax put up, these themes will really feel very acquainted—as a result of they’re the lived expertise behind the classes.
How the AI Tax Explains Intensification
In “The AI Tax,” I described six methods AI creates extra work than it saves when deployed with out design. The brand new HBR analysis slots cleanly into that framework.
- Juggling with AI: multi-tasking, switching, sprawl
The research’s third sample, elevated multitasking, is the human expertise of juggling throughout AI instruments, brokers, and metaphors of interplay. In my put up, I wrote about toolchain sprawl: one AI for scheduling, one other in e-mail, a 3rd hidden in a CRM, every with a distinct interface, set of capabilities, and quirks. The result’s a workday that appears like a perpetual reconciliation train, with consideration sliced into dozens of skinny duties. - Vetting: oversight and the hallucination downside
Activity enlargement sounds environment friendly till you do not forget that each AI-generated draft, be it a doc, snippet of code, or advertising marketing campaign, requires vetting. The HBR research paperwork engineers who begin spending vital time reviewing AI-assisted work produced by colleagues exterior their self-discipline, typically by casual Slack exchanges and favors. That’s the AI Tax’s “shadow labor,” actual work with no line merchandise in a undertaking plan, absorbed by folks already at capability. - Knowledge science and readiness: hidden work uncovered
AI makes knowledge issues seen. When workers eagerly broaden their scope: writing analyses, reviews, or prototypes they might not beforehand have tried, they shortly collide with scattered, mislabeled, or outdated knowledge. That collision forces them into advert hoc knowledge wrangling: reconciling codecs, attempting to find authoritative sources, and studying simply sufficient concerning the group’s knowledge structure to be harmful. - Relevance and security: governance lagging adoption
As AI disseminates content material extra shortly, questions of tone, bias, confidentiality, and regulatory danger develop into each day issues moderately than edge circumstances. The HBR article hints at this not directly, however the connection to my AI Tax class is direct: when governance lags behind adoption, every step ahead requires a detour to confirm compliance and appropriateness. That friction doesn’t present up in vendor demos, however workers really feel it instantly. - Failed initiatives and abandonment cycles
The research depicts enthusiastic early experimentation: folks “simply making an attempt issues” with AI. In my put up, I warned that this sample typically evolves right into a cycle of pilots that don’t hook up with actual workflows, bots that die on the sting of a promise, and technical debt that somebody has to scrub up. When each failed experiment leaves behind deserted prompts, partial automations, and skeptical customers, the AI Tax compounds over time. - Studying and relearning: AI as a shifting goal
Lastly, each the HBR article and my AI Tax put up converge on the churn of studying. Each mannequin replace, interface change, and new characteristic, not to mention the arrival of solely new instruments, forces folks again into coaching mode. Add in social FOMO (“Have you ever tried the newest mannequin?”) and also you get a tradition during which staff are anticipated to maintain up with a consistently shifting AI panorama whereas additionally sustaining their current obligations.
The purpose isn’t that AI can not create worth. It’s that worth and complexity scale collectively, and complexity arrives first.
The Free Time Mirage
When AI works, when it truly hastens a job or simplifies a workflow, a distinct query emerges: what occurs to the time that’s freed? Within the AI Tax article, I argued that this isn’t a technical query however a management and coverage problem. With out intentional design, freed time will get reabsorbed into:
- Extra duties, typically vaguely outlined as “strategic work” or “innovation.”
- Casual expectations that people will tackle additional obligations as a result of “the instruments make it quicker now.”
- Delicate stress to take care of or improve output moderately than use time for restoration, studying, or collaboration.
The HBR research makes this dynamic seen. Staff used AI to shave time without work duties, then crammed the margin with new work: serving to colleagues, experimenting with further prompts, or extending their obligations into areas beforehand out of scope. They felt extra productive, however not much less busy. Over time, the preliminary thrill gave approach to exhaustion and cognitive fatigue.
That is the core of the AI Tax argument: if organizations don’t explicitly resolve the best way to deal with time saved by AI, the default will at all times be intensification, not liberation, and in lots of circumstances, substitution moderately than augmentation.
Designing Towards Intensification
The HBR authors counsel that organizations want specific “AI practices” to forestall intensification from changing into the default: norms about when to make use of AI, when to not use it, and the best way to handle AI-enabled work sustainably. The AI Tax framework aligns with that decision and affords concrete beginning factors.
Listed here are a number of design strikes leaders could make, knowledgeable by each the analysis and the AI Tax:
- Standardize the AI stack
Cut back toolchain sprawl by selecting a small variety of platforms and constructing round them. Consolidation lowers cognitive switching prices, simplifies governance, and makes it simpler to design coaching that sticks moderately than chasing each new characteristic. - Make vetting seen and accountable
Cease treating oversight as invisible heroism. Assign vetting obligations, observe the time it takes, and issue that point into undertaking plans and ROI claims. This isn’t simply truthful; it generates the information wanted to resolve the place AI genuinely helps and the place it merely redistributes labor. - Put money into knowledge earlier than scale
Most of the frustrations uncovered within the research,, akin to partial outcomes, complicated outputs, and reliance on “vibe” coding, stem from poor knowledge, unclear requirements, or lacking context. Cleansing, tagging, and aligning knowledge are unglamorous, however they’re important if AI is to provide outputs that scale back work moderately than create further cleanup work. - Run time-bound pilots with actual endings
Organizations ought to deal with AI pilots as experiments with specific timelines and choice gates, moderately than as everlasting, half-adopted options. On the finish of a pilot, both commit and make investments, or shut it down and doc what was discovered so that you don’t repeat the identical errors later. I additionally repeatedly argue that AI requires knowledge management, however accelerated AI adoption too typically overwhelms its implementation. - Defend human time as an asset
Maybe most significantly: resolve, prematurely, the best way to reclaim free time with function. Some portion ought to be explicitly allotted to relaxation, reflection, mentoring, and exploration, moderately than being harvested as a shadow productiveness achieve. If AI is to be a colleague, it ought to create situations for higher human judgment, not merely better throughput.
From AI Tax to AI Observe
The convergence between the HBR analysis and the AI Tax is encouraging as a result of it suggests we’re shifting out of the speculative section of AI and right into a extra empirical, design-oriented one. We now have a rising physique of proof that, left to its personal gadgets, AI doesn’t scale back work; it lowers friction and invitations extra work.
The duty for leaders is to deal with these realities as design constraints moderately than as inconveniences. The AI Tax identifies the place prices accumulate; the HBR article exhibits how these prices manifest in an actual group over time. Between them lies the chance to construct “AI practices” that honor human limits, shield time, and be sure that depth is a alternative moderately than an accident.
His work has been featured in Harvard Enterprise Overview, Quick Firm, Wired, NASA Ask, and dozens of different magazines and web sites. His evaluation on the way forward for work could be discovered at SeriousInsights.internet. He’s additionally the creator of Understanding Synthetic Intelligence, Cyberlife, Rethinking Good Objects, and Empower Enterprise with Generative AI.
Rasmus repeatedly speaks on the way forward for work at occasions akin to Comedian-Con Worldwide, Wondercon, Influence, Enterprise AI World, Worktech, CLO Symposium, KMWorld, AAAI, Computer systems in Libraries, Microsoft’s Constructing the Future, Educause, Expo Capital Humano, Devlearn, Web Librarian, CAMEX, EduComm, and Sourcing Summit Europe.
As an Affiliate Teacher, Rasmus teaches state of affairs planning on the College of Washington and is an Affiliate Adjunct Professor at Bellevue Faculty.
