How People Are Really Using AI in 2026

Harvard Business Review · article · 2026-06-01 Reader 🖥️

AI Slop Cognitive Surrender Emotional Impact of AI Work Enterprise AI Adoption & Readiness Agency Measuring AI Impact

Across a third annual edition, generative-AI use is widening into an ever-broader and more intimate range of tasks — best read as shifts in emphasis, not ruptures — with people increasingly outsourcing both their thinking ('thinkslop') and their emotional lives to AI, even as, in business, the technology so far yields mostly marginal efficiencies rather than game-changing transformation, leaving the survival of human agency as the open question.

theoretical · Expository data/study piece (third annual edition of the 'AI in the Wild' longitudinal study) with a normative frame — descriptive findings about usage carrying an argument about cognitive/emotional risk and a 'marginal-value' verdict, plus a short practical tips coda. — Primarily theoretical/descriptive: it argues what is TRUE about how people use AI (a taxonomy of 12,637 use cases). The thinkslop and agency sections add a normative overlay (this is cause for concern; here is what to do), but the core is empirical description. The evidence is social-listening on public posts, not a survey — a classification-relevant caveat.

Structure

Part	Role
Framing & methodology	Establishes the empirical basis (12,637 use cases from ~50,000 records, Mar '25-Feb '26, mined via 'social listening' from Reddit/Quora/articles plus new LinkedIn/TikTok/YouTube; hybrid human-AI coding) and the interpretive rule that year-over-year change is 'shifts in emphasis, not stark ruptures' because the user pie is growing.
Headline claim + three concerns	States that usage is widening, then previews the three worries: cognitive surrender (thinkslop), emotional over-reliance, and marginal-not-game-changing business benefits.
'Thinkslop' section	Argues that ~a quarter of top use cases involve asking AI to do our thinking, and lays out four mechanisms of erosion (lost intention, outsourced problem-solving, stopped writing, false rigor via AI sycophancy) — then concedes AI-as-sparring-partner can sharpen thinking.
'Emotion' section	Documents therapy/companionship as the #1 use case (11% of dataset, up from 5%), anthropomorphizing (naming AIs, assigning genders, grief at model updates), and high-stakes risks (AI psychosis, romance, heartbreak, death), with expert caution that chatbots aren't a substitute for clinicians.
Work section (efficiencies / growth / transformation)	Shows AI at work is bottom-up and often covert ('shadow usage'), then walks the value ladder: efficiencies (time saved) dominate; growth cases exist but explicit ROI is rare and mostly anecdotal; transformation is early-stage, unquantified, and confined to SMEs.
Closing on agency	Reframes the whole survey as a question about how much human agency can survive a powerful, ubiquitous, always-on service — placing the answer mostly with users, not AI companies.

Key terms

thinkslop
The lazy, sloppy thinking engendered by excessive use of AI — outsourcing to the model the cognitive responsibilities humans both need to own and thrive at. Operationalized as four mechanisms: (1) losing track of our intentions (prompting before we've thought), (2) outsourcing our thinking (denying the brain a fresh solve; the cited 'cognitive debt'), (3) stopping writing (skipping the drafting-is-thinking process, producing 'workslop'), and (4) a false sense of intellectual rigor (AI sycophancy flattering us into stopping too early).
shifts in emphasis, not ruptures
The rule for reading year-over-year change: because the total user base is growing, a use rising in the rankings does not mean people abandoned another; changing proportions within a growing pie, not populations switching behavior.
social listening
The methodology: inferring AI use cases by mining public online posts (Reddit, Quora, articles, and newly LinkedIn, TikTok, YouTube), then using a hybrid human-AI system to identify bona fide use cases. The author acknowledges 'pros and cons' and cites a separate 2025 piece detailing them.
marginal vs. game-changing benefits
Marginal = modest, uncontroversial efficiency wins (time saved, clarity gained, painful tasks automated) and some sales/marketing growth; game-changing = fundamentally rethinking business processes to do new things in new ways. The verdict: business activity is producing the former, not the latter, 'so far.'
AI sycophancy
AI being optimized to engage and keep us engaged, lavishing congratulatory praise that induces users to stop fine-tuning and to return for more flattery — the engine of the 'false sense of intellectual rigor.'
shadow usage / shadow AI
Covert, unsanctioned individual AI use at work, adopted despite governance/reputational headwinds ('nobody knows I'm using AI'); a consequence of tailwinds (licenses, training, C-suite encouragement) being weaker than headwinds (IT/AI governance, restricted LLMs, fear of cheating/job loss).
autonomous agentic operations
AI going ahead and DOING rather than just talking/advising — a new #6 entry, but the examples are experimental and small-scale, with a large share being conversion of voice notes to other work media.

Propositions

explicitPeople are adopting generative AI for an ever-widening and increasingly intimate range of uses; the 2026 edition analyzed 12,637 use cases, an order of magnitude more than the two prior editions.conf: high
explicitTherapy/companionship is the #1 use case for the second consecutive year, growing from 5% to 11% of the dataset (more than 1,400 entries) — i.e., in both absolute and relative terms.conf: high
explicitYear-over-year changes should be read as shifts in emphasis within a growing user base, not as ruptures where users abandon prior uses — 'the pie of users is growing.'conf: high
explicitIn at least a quarter of the top use cases, people are asking AI to do part of their thinking ('thinkslop'), and these are precisely the activities for which humans need to take responsibility and at which the species thrives.conf: high
explicitdevelopedIn business, generative AI is producing mostly marginal efficiencies (time saved, clarity, automating painful tasks) rather than game-changing transformation; where transformation occurs it is early-stage, unquantified, and confined to SMEs.conf: high
supportsThe Organization Is the Bottleneck · p1 — Population-scale social-listening data (12,637 use cases) empirically corroborates Fowler's claim that faster AI activity is not obviously delivering faster/greater organizational value — HBR finds mostly marginal efficiencies and few fundamental process rethinks.
refines ←The Hidden Realities of AI Adoption · p6 — Offers a mechanism for the 'marginal efficiencies, not transformation' finding: individual/ungoverned AI access doesn't yield measurable productivity, while enterprise-scoped process redesign does — explaining why breadth of individual adoption hasn't translated into business transformation.
explicitcorroboratedExplicitly quantified ROI is rare in the data; business growth from AI is more often claimed anecdotally than demonstrated with numbers.conf: high
supportsAgentics / Tech Things: Tokenmaxxing is dead, long live tokenmaxxing · p12 — Independent real-world corroboration of tokenmaxxing's gap that AI's value/ROI is asserted more than demonstrated: HBR reports explicitly quantified ROI is rare in the data and business growth is mostly claimed anecdotally.
supports ←The Hidden Realities of AI Adoption · p16 — Corroborates from the organizational-culture side why quantified AI ROI is rare: most organizations lack a disciplined measurement/experimentation culture in the first place, independent of the social-listening methodology critique.
explicitMost workplace AI use is bottom-up, individual, and often covert ('shadow usage'), driven by workers themselves rather than top-down corporate initiatives, because governance and reputational headwinds outweigh organizational tailwinds.conf: high
explicitcontestedAutonomous agentic operations entered the top 10 (#6) but remain experimental and small-scale, with a large share of the 500+ entries being conversion of voice notes into other work media rather than consequential autonomous action.conf: high
contradicts ←Research: How AI Agents Broaden the Scope of Knowledge Work · p9 — That piece, reading public social-media posts, found autonomous agentic operations still 'experimental and small-scale' in 2026; this piece's internal-telemetry field study finds agent use already reshaping 23-40% of task content within one company's product — a sharp contrast that may reflect a real maturity gap between consumer-facing agent features and a dedicated agent product, or the very different (and equally non-independent) methodologies each study relies on.
explicitWhether AI erodes or sharpens thinking depends on the user's posture: used as a crutch it engenders thinkslop, but used as an intellectual foil/sparring partner (to challenge assumptions and poke holes) it can sharpen thinking.conf: high
supportsCognitive Surrender · p7 — Near-verbatim parallel: HBR's population-scale data independently corroborates the posture-dependence thesis — the same AI can hollow out or sharpen thinking depending on whether the user thinks WITH it (sparring partner) or lets it think INSTEAD (crutch).
implicitThe methodology assumes that what people voluntarily post on public online platforms (Reddit, Quora, LinkedIn, TikTok, YouTube, articles) is a valid proxy for how people actually use AI — i.e., that vocal, self-selecting posters represent the broader population.conf: high
implicitRanking use cases by frequency of online mention conflates how often something is talked about with how often it is actually done; emotionally charged uses (therapy, grief, romance) may be over-represented because they generate more posting.conf: medium
implicitBreadth of adoption is treated as the headline finding, but breadth of use implies little about the depth of value derived — counting distinct use cases measures variety, not impact.conf: high
implicitThe 'marginal rather than game-changing, so far' verdict presupposes that transformation is still coming, yet the study offers no mechanism, threshold, or leading indicator for what would convert marginal efficiencies into game-changing value.conf: high
implicitThe thinkslop harm is inferred from vivid user testimony at population scale, not measured: the study counts use cases where thinking could be outsourced, not the cognitive damage actually incurred, nor the ratio of crutch-use to sparring-partner use.conf: high
implicitConsumer/individual usage data is being read as a signal about enterprise value, conflating individual-scale use with organizational transformation.conf: medium
implicit'Shifts in emphasis, not ruptures' functions as a near-unfalsifiable interpretive hedge: because the user base is always growing, any pattern — a use rising, falling, or vanishing from the top 10 — can be absorbed without disconfirming the framework.conf: medium

Arguments

In business, generative AI is so far delivering marginal efficiencies, not game-changing transformation. strength: medium

Explicit premises:

Most workplace use makes existing processes more efficient (time saved, clarity, automating painful recruitment/summarizing/templating tasks).
Growth cases exist (e.g., a 20-30% conversion lift from AI-optimized email A/B testing) but explicit ROI is rare and growth is mostly claimed anecdotally.
Transformation examples are early-stage, the benefits unquantified, and the beneficiaries SMEs; where major change occurs it often comes with cynicism about quality ('replaced an entire analytics function. Is it better? No.').

Implicit premises:

The social-listening dataset is a representative cross-section of real business AI use.
The absence of quantified ROI in public posts implies the absence of large ROI, rather than reticence to disclose value publicly.
'Marginal, so far' implies transformation remains genuinely possible and pending.

Inferences:

This is the third consecutive edition reaching a version of the same 'transformation still coming' verdict; the recurring lag is itself evidence that the bottleneck may be structural (organizational absorption, legibility) rather than temporal — an inference the author does not draw.
If the biggest wins are bottom-up and covert (shadow usage), organizations may be structurally unable to capture even the marginal value on their own books.

The widening of AI use is causing people to surrender cognitive responsibility ('thinkslop'). strength: medium