Some industries already lost their moat. Yours is on this map somewhere.

A free chatbot wiped out a homework-help business worth ~$14B at its peak — and the CEO had to say so on an earnings call.

What changedChegg's moat was a paywalled library of textbook solutions and on-demand Q&A. ChatGPT answered the same questions for free, instantly, with no subscription — the value Chegg charged for became a commodity overnight.

Note — Often cited as the first public company to formally blame AI for revenue damage. Its own OpenAI-powered 'CheggMate' failed to stem the loss; subscribers kept falling through 2024–2025.

The best-funded digital freight broker automated ~90% of brokerage tasks, then shut down — proving automation alone didn't beat the economics.

What changedConvoy's automation cut the cost of matching loads and carriers, but a freight-market downturn plus heavy cash burn killed it; Flexport bought the remnants.

Note — Primary cause was the freight recession and burn rate, NOT AI displacement. Included as the cautionary 'automation isn't a moat' case rather than pure AI disruption.

Sites that scaled thin AI content to farm Google traffic were wiped out when Google's algorithm specifically targeted unhelpful, search-first content.

What changedAI made mass content production nearly free, flooding search. Google responded with the Helpful Content Update (2022–2023, folded into core ranking in Mar 2024), and many AI-heavy sites lost 90%+ of organic traffic.

Note — The same AI that lowered content cost triggered the platform response that destroyed the arbitrage. Kept general rather than naming individual victim sites.

The viral-content king bet its future on AI-written articles and instead wrote its own obituary.

What changedCommodity listicle/quiz content lost its scarcity once AI could generate it for free. BuzzFeed shut its Pulitzer-winning news division, pivoted to AI content, and kept losing audience and advertisers.

Note — By 2025 the company disclosed 'substantial doubt' about its ability to continue as a going concern. Decline is multi-causal (digital-ad collapse) — AI was the failed turnaround bet, not the sole killer.

Klarna said one AI assistant did the work of 700 agents in its first month — then admitted it had cut human support too far and started rehiring.

What changedTier-1 contact-center work (refunds, order status, cancellations) is largely scripted and automatable; the AI handled ~two-thirds of chats faster than humans, but quality limits on edge cases forced a rebalance toward premium human support.

Note — The '700 agents' / '$40M' figures are Klarna's own marketing claims (the $40M was largely avoided hiring, not pure layoffs). The 2025 walk-back is the credible part — this is NOT a clean 'AI replaced everyone' case.

The category owner shipped its own generative AI fast, yet still saw its stock cut sharply as the market doubted AI would protect its subscription moat.

What changedAdobe's moat was a near-monopoly on professional creative tooling. It embedded Firefly across Photoshop and Illustrator to defend it, but commoditized generation and AI-native rivals (Figma, Canva) put premium pricing in question.

Note — Firefly adoption and launch date are solid. The stock decline is verifiable; attributing it specifically to AI fear is the consensus narrative but not the only factor. Defending strongly — 'on the clock,' not 'gone.'

Analyst pay fell and the ranks thinned as AI, passive investing, and unbundling squeezed the research desk.

What changedAI now drafts the models, summaries, and first-pass analysis junior analysts once produced; combined with MiFID II unbundling and the shift to passive, the economic case for large research teams collapsed.

Note — Bloomberg Intelligence projected banks could cut up to ~200,000 jobs over 3–5 years. AI is one of several compounding forces, not the sole cause.

The firms whose pyramid runs on armies of junior reviewers are racing to automate the very work those juniors do.

What changedAgentic audit/tax platforms now do first-pass document review, data collection, and tax compliance, threatening the leverage-pyramid economics — but liability, regulation, and audit sign-off keep humans accountable.

Note — Reinvention is active and self-inflicted; net headcount/economic impact isn't yet resolved — hence on-the-clock rather than gone.

Free and cheap AI-assisted self-filing is quietly shaving the addressable market for paid human tax prep, season by season.

What changedSelf-prepared returns are rising while professionally-prepared returns flatten; AI chatbots in DIY software lower the perceived need to pay a preparer, eroding the core — though those chatbots have themselves been faulted for inaccurate advice.

Note — AI accelerates a long-running DIY shift rather than being the sole driver.

Freight brokerage's human-relationship middle is being squeezed as incumbents deploy fleets of AI agents to price lanes, vet carriers, and book loads.

What changedUber Freight launched an LLM-backed network of 30+ AI agents that handle pricing, carrier reliability scoring, and booking — the routine cognitive work that justified traditional brokers' margins.

Note — Disruption is visibly underway but unresolved; Uber Freight itself remained unprofitable, so the economics aren't settled.

The marketing-copy darling raised $125M at a $1.5B valuation, then watched a free ChatGPT commoditize its core product almost overnight.

What changedJasper was a UX wrapper on GPT-3; once OpenAI shipped ChatGPT (free) and the API directly, the value-add collapsed and Jasper had to pivot from copy generation to enterprise marketing workflows and brand controls.

Note — Survived by repositioning toward enterprise teams; not dead, but the original moat is gone.

The platform recruiters live in shipped an AI agent that automates the sourcing, ranking, and outreach that defined the recruiter job.

What changedHiring Assistant writes job descriptions, identifies and ranks candidates, runs InMail Q&A, and pre-vets applicants — LinkedIn estimates it cuts recruiter admin time by ~70%, putting routine sourcing under pressure from the incumbent itself.

Note — The 70% admin-reduction figure is LinkedIn's own estimate.

The poster child for 'AI replaces sales-development reps' got hit by an investigation alleging fake logos, inflated ARR, and heavy churn.

What changed11x sold autonomous AI SDRs to automate prospecting and outreach and raised from top VCs, but a Mar 2025 investigation alleged it listed non-customers and counted trials as annual revenue — undercutting the 'AI fully replaces the SDR' narrative.

Note — Claims are CONTESTED; the CEO disputed the report and investors backed the company. Included to show the AI-SDR category is real but full SDR replacement remains unproven — do not state the allegations as established fact.

A flagship Latin American IT-services firm lost roughly two-thirds of its market value as clients questioned whether they still needed billable headcount.

What changedThe staff-augmentation model prices on hours and headcount; AI coding tools compress the hours. The firm is racing to reprice around AI 'Pods' and subscriptions before the old model erodes.

Note — Stock decline is verifiable; the AI-pricing pivot is real but unproven. Securities suits also allege broader LatAm demand problems, so AI is a major but not sole driver.

The vast Indian outsourcing engine began shedding tens of thousands of jobs as AI took on the entry-level coding, testing, and support it was built to staff.

What changedThe model sold scale of cheap labor for routine dev/test/support work; AI now does much of that, so 'more people = more revenue' is breaking and hiring is being structurally cut.

Note — Layoff figures are press/analyst-sourced and partly attributed to AI plus macro/efficiency pressure — multi-causal. Treated as a sector exemplar, not one company.

The stock-photo giant is fighting AI image generators in court while racing to ship its own 'commercially safe' tool.

What changedText-to-image models now generate on-demand visuals that substitute for licensed stock photos. Getty sued Stability AI over training data and launched its own licensed generative tool to defend the franchise.

Note — Getty alleged unauthorized scraping of ~12M images. In the UK High Court ruling (Nov 2025) it largely lost on copyright and won only a narrow trademark point.

Freelance translators watched rates and volume crater as clients switched to AI with light human editing.

What changedGenerative AI made 'good enough' translation nearly free, so agencies and localization studios cut rates, moved to AI-plus-post-editing workflows, and reduced staff — squeezing professional translators' income.

Note — A 2024 UK Society of Authors survey found over a third of translators had lost work to GenAI. Larger figures circulating (e.g. '80% income drop') are anecdotal — treat as directional.

Two ex-Big-Tech execs built an AI-agent company for customer service that reached ~$100M ARR in under two years and a multi-billion valuation.

What changedAI agents now resolve customer issues end-to-end (refunds, account actions), so enterprises buy outcomes instead of seats of human-staffed contact centers — a new AI-native category challenging the old CX/BPO stack.

Note — ARR and valuation are press-reported, not audited — approximate. Decagon is a comparable AI-native reinventor in the same space.

A tiny self-funded team turned text prompts into pro-grade imagery and built a nine-figure business with no outside investors.

What changedCustom commercial imagery used to require a skilled illustrator or photographer and hours of work. Midjourney collapsed that to a prompt and seconds, redefining the bottom of the commercial-art market.

Note — Founding date and 'no VC' are well-established. Revenue figures are third-party estimates, not disclosed financials — treat as approximate.

The language-learning app went 'AI-first,' generating course content that used to require human contractors.

What changedSentences, translations, and error review that once needed human contractors can now be generated and checked with LLMs, so Duolingo rebuilt content production around AI and cut contractor roles.

Note — Cut ~10% of its contractor workforce (translators/writers hit hardest) starting late 2023; CEO later issued an 'AI-first' memo. A reinventor that also displaced workers — note the dual lens.

Instead of being eaten by AI, the course platform turned 'learn AI' into its single biggest growth driver.

What changedDemand shifted toward learning how to use AI itself. Coursera leaned in with hundreds of new GenAI courses and an AI coaching layer, making AI a tailwind rather than a threat.

Note — Reported ~3M new GenAI enrollments in 2024 with 450+ new GenAI courses from partners. Framing as 'reinventing' is directional — the long-term threat of LLMs as substitute tutors remains unresolved.

A bank's internal tool did in seconds the loan-agreement review that consumed 360,000 lawyer/loan-officer hours a year.

What changedMachine learning extracted ~150 attributes from 12,000 commercial credit agreements in seconds and cut interpretation errors sharply — showing core document-heavy finance work could be automated in-house years before generative AI.

Note — An early, pre-LLM exemplar. The 360,000-hours figure is JPMorgan's own and often cited.

A two-year-old startup became the default AI layer for elite law firms before incumbents could ship a credible competitor.

What changedGPT-class models plus a forward-deployed, ex-lawyer change-management model let Harvey embed legal research, drafting, and review into BigLaw workflows; roughly half of the Am Law 100 are customers and the category formed around it.

Note — Valuation and ARR figures are company-reported and rising fast; treat point figures as as-of-date snapshots.

The legal-research incumbent bought its way into the AI era for $650M rather than be disrupted by a GPT-4 startup.

What changedCasetext's CoCounsel — a GPT-4-powered assistant for research, document review, and deposition prep — proved generative AI could compress hours of associate work to minutes. Thomson Reuters acquired it to defend its Westlaw franchise.

An AI voice agent now sources, screens, and interviews thousands of candidates a day — work that used to need a team of recruiters.

What changedConversational AI agents conduct adaptive voice/video/phone/SMS interviews in 20+ languages, 24/7, reading job specs from the ATS and scoring candidates — redefining high-volume screening.

Note — Early-stage; throughput and CSAT figures are company/customer-reported, not independently audited.

An AI-native code editor went from zero to ~$100M ARR in about two years and redefined what a developer tool is.

What changedInstead of autocomplete bolted onto an IDE, Cursor made the LLM the primary interface for writing and editing code — a new category that incumbents had to chase.

Note — ARR and valuation figures come from press reports, not audited filings — treat the exact numbers as approximate. The 'AI-native reinventor of the category' claim is well-supported.

Rather than just defend against AI, Shutterstock sold its library as training data and built generation into its platform.

What changedIt turned its image archive into a new revenue stream by licensing it to AI labs for model training, and integrated text-to-image generation for customers, with a fund to compensate contributors.

Note — Later reported its AI/data-licensing business generating roughly $100M+ in a year; exact figures vary by reporting period.

A human-transcription marketplace turned the AI that threatened it into its own weapon — releasing an open-weight model that beats OpenAI's Whisper.

What changedRev used the data exhaust of its human transcribers — 200,000+ hours of human-transcribed speech — to train Reverb, an open-weight model with materially lower word error rate than Whisper large-v3. The human workforce became the training-data moat.

Note — Reinvention came at the cost of its human transcribers (fewer jobs, lower pay) even as the company moved up-market.

An AI-native entrant redefined transcription from a service you ordered into ambient meeting intelligence baked into every call.

What changedInstead of per-minute transcription, Otter embedded directly in Zoom/Meet/Teams as an always-on notetaker, then layered in chat and agents — redefining the category around real-time meeting intelligence.

Note — A category-redefining new entrant, not an incumbent's turnaround.

An AI-native translation startup out-translated Big Tech and became the localization industry's default tool.

What changedDeepL built neural-MT quality that professionals rated above Google Translate, then sold it as a B2B product — capturing the enterprise translation market incumbents and human agencies once held.

Note — Reported ~$185M ARR for 2024, up from ~$141M in 2023.

AI ate the grunt work, but licensure and liability kept the profession's core billing intact — for now.

What changedDocument review, first-draft contracts, and research are increasingly automated, yet unauthorized-practice rules, malpractice liability, and ABA Formal Opinion 512 (2024) keep a licensed human in the loop. The visible shift is firms hiring fewer juniors, not mass replacement.

Note — 'Insulated' applies to the licensed-judgment layer; the entry-level training pipeline is genuinely under pressure — insulated-but-eroding.