Live missions.
What the case files add up to. Explainers, analysis and coverage of AI operations and automation. Argument, never proof: nothing on this page carries a badge, and every figure carries a source you can click.
medicare ai prior authorization in 2026: what the court record actually shows
Everyone is arguing about how accurate the algorithm is. The two federal courts that have ruled on Medicare Advantage AI denials asked a different question: did the insurer keep its own promise about who decides.
ai agent fraud in 2026: what the adjudicated record actually shows
Banks are warning that AI shopping agents will get you scammed. The AI-agent fraud that regulators have actually charged runs the other way: the agent itself was the fraud.
Generative engine optimization: what the controlled evidence actually shows in 2026
One 252,000-trial experiment, a 45-study survey and four crawler-log and correlation studies, read from the source tables. The factors that decide a citation are not the ones being sold, and the one most often sold measured as no effect.
AI in retail examples: what the documented record actually shows in 2026
Most public AI-in-retail examples are enforcement actions, not deployment wins. Here are the ones that survive the record, and the few clean ones.
Does agentforce work? The agentic customer-service record before you buy, 2026
Salesforce sells Agentforce as agents that resolve cases and hand off to humans only when needed. The last big autonomous customer-service deployments say the dial you set matters more than the demo.
AI drug discovery examples: what the verified public record shows in 2026
The pitches promise new medicines. The public record shows AI doing narrower, upstream work, and here are the examples TIN checked itself.
ai coding agent: what the measured record shows about developer productivity (2026)
The benchmarks rank the model. The field trials measured real developers, and the range runs from 26% more work to 19% slower. Here is what the evidence supports.
AI search optimization: what ChatGPT's retrieval stream shows, and the numbers nobody can source
A viral analysis of ChatGPT's server-sent events says the engine ranks passages, not pages. We checked it line by line. The mechanism holds up across four independent observers. Two of the numbers being quoted from it are not in it.
ai regulation moratorium in 2026: what the enforced record shows while Congress stalls
A moratorium would freeze new state AI statutes. It would not touch the fines, bans and court orders already landing on AI systems under laws that predate the debate.
ai regulation: what the enforced record already shows in 2026
Everyone is waiting for the AI bill. Meanwhile regulators fined Clearview 30.5 million euros, banned an AI storefront scheme, and let a nationwide age case against Workday proceed, all under laws that already exist.
Deepfake fraud: what the adjudicated record actually shows in 2026
The biggest deepfake fraud numbers are victim-reported or projected. What an independent regulator and a court have actually put on record is smaller, specific, and more useful.
Will AI replace underwriters? What the documented record shows in 2026
AI already makes underwriting calls at scale. The record shows the accountability, and most of the job, staying with a person.
ai chatbot lawsuit: what the adjudicated record shows in 2026
Three courts on two continents have now answered the same question: when a chatbot gets it wrong, who pays? The consistent finding is the operator, not the bot.
Algorithmic management: what the enforced record shows for gig workers in 2026
Three European regulators fined delivery and ride-hail platforms for managing workers by software. Read together, the fines punish one thing: the missing human.
AI fraud detection: what the documented record shows in 2026
The loud numbers on AI fraud detection are vendor-reported. The two cases in the independent public record tell a harder story: one model that saved money, one that a regulator ruled unlawful.
ai claims processing in 2026: what the record actually shows
The demos promise a claim that files, checks and pays itself. The public record shows something narrower: computer vision speeding a motor insurer's review, document automation inside healthcare billing, a rule-based benefit process that erred in almost every case an auditor sampled, and an AI denial tool now arguing its case in federal court.
agentic ai in banking in 2026: what the measured record actually shows
Every vendor deck promises autonomous agents running the bank. The public record of the last AI wave in finance says something colder: the headline numbers were single-source, one was reversed, and regulators fined the firms whose AI turned out not to exist. Here is what to demand before you buy the autonomy.
automated decision making: what the enforced record shows (2026)
The debate treats automated decision making as a question about how smart the model is. The enforced record treats it as a question about one thing: was there a human who could review the decision and a way to contest it. Five cases show what happens when the answer is no.
facial recognition lawsuits: what the enforced record shows (2026)
The headlines say facial recognition is being banned. The enforced record says something narrower: six actions across four jurisdictions, and every one turned on how the tool was deployed, not on the technology itself. Here is what each actually ordered.
does ai increase productivity: what the measured record shows (2026)
AI clearly saves workers time. Whether that time becomes output is a separate question, and the measured record splits: a central-bank null on one side, specific audited wins on the other.
ai in healthcare in 2026: does clinical ai improve outcomes, on the peer-reviewed record
The strongest AI-in-healthcare results of 2026 are peer-reviewed and real. They also share a feature the marketing leaves out: the AI only counted after a human confirmed it.
ai copyright lawsuit: what the adjudicated record actually shows in 2026
Three courts, three different answers. A $1.5 billion Anthropic settlement, a rejected fair-use defense in Delaware, and a Getty loss in London show that the outcome turns on what was copied, not a single rule.
ai contract review in 2026: what the documented record shows
Vendors sell ai contract review as read, extract, flag, route, and faster than a lawyer. The most-cited enterprise deployment, JPMorgan's COIN, backs the speed but not the accuracy. This post says which claims carry a source and which never did.
ai workflow automation in 2026: what the documented record shows
AI workflow automation is sold on huge hours-saved numbers. The deployments that hold up trade the big number for a small one you can check: who approves, what got routed, how long it took.
ai inventory management in 2026: what the documented record shows
AI inventory management is sold as stock down and margin up. The one case that holds up ties its accuracy claim to a test the business can check: did the item sell.
supply chain automation in 2026: what the documented record shows
Every vendor sells supply chain automation as headcount out and savings in. Two cases TIN verified show a narrower, more useful truth: automate the task with a success test, not the job.
ai supply chain in 2026: what the documented record actually shows
The vendor pitch for AI in the supply chain is savings and autonomy. Two cases TIN verified, and one fresh company disclosure, show what actually holds up when you check who measured the number.
ChatGPT stopped citing Reddit and kept buying it: what the drop says about how to get cited by ChatGPT
Reddit's share of ChatGPT citations fell about 86% in four days. In the same month Reddit reported data licensing revenue up 24% with OpenAI as a joint-largest buyer. OpenAI did not give up Reddit. It stopped crediting it, and that distinction decides which substitute platform is worth anything.
What is reward hacking? In 2026 it broke the benchmarks, then it breached a live company
Reward hacking is an AI moving the score without doing the work. The 2026 record is measured, dated and worse than the coverage suggests: a third of runs on some benchmarks, every frontier model tested by the UK's safety institute, and one evaluation that ended in a real intrusion at another company.
Agentic AI customer service: what the verified record shows in 2026
The pitch is autonomous agents that finally let you cut the support team. Four documented deployments say the durable win is honest deflection with a human handoff, not headcount removed.
ai hiring discrimination in 2026: what the enforced record actually shows
Vendors warn that AI screening will get you sued. The enforced record is smaller and more specific than the scare copy: two US cases, resolved by settlement, name who paid and what changed. Here is what each actually established, and what it did not.
ai document processing in 2026: what the documented record actually shows
The two most-cited document-processing deployments TIN has read closely, Omega Healthcare on UiPath and JPMorgan's COIN, deliver real gains. Nearly every headline number is the company's own. This post says which figures carry a followable source and which were never independently measured.
AI voice agents: what the verified record actually shows (2026)
The loudest AI voice agent numbers come from vendor case studies. The one voice deployment a regulator examined was found overstated. Here is what survives verification.
Are AI detectors reliable? The record says no, and Claude's watermark does not fix it
Anthropic now watermarks every word Claude writes. But AI detection false positives are what actually cost people their work, their grades and their reputations, and a watermark proves a machine was involved, not that it wrote anything.
ai automation companies in 2026: what the documented record shows
You cannot judge an ai automation company from a logo grid. Here are seven deployments TIN has documented, what each provider actually delivered, where the figure came from, and how to tell who delivers from who markets.
ai fabricated citations in 2026: what the court and tribunal sanctions record shows
Lawyers keep filing briefs with AI-invented cases, and courts keep sanctioning them, from $3,000 to $31,150 and up. Four documented rulings across two countries show the pattern and the one thing every panel now agrees on: the AI excuse makes it worse, not better.
the ai job-losses myth in 2026: what Zalando's own 2018 announcement actually shows
'Algorithms replaced 250 marketing jobs at Zalando' is one of the most-cited AI-job-loss examples, and it's missing half the sentence. The same 2018 announcement carried a plan to hire 2,000 more people, and total headcount kept rising for years after.
ai washing in 2026: what the FTC's $25M Ascend Ecom order actually shows
Ascend marketed 'AI-powered' e-commerce storefronts as passive income. The FTC alleged clients earned virtually nothing, and a federal court entered a $25M judgment and a permanent ban. The AI label wasn't the violation, the unsubstantiated income claim was.
does ai customer service cut jobs in 2026? what Octopus Energy's record shows
Octopus Energy's Magic Ink handles a third of support queries and the company kept hiring. It's a documented counter-example to the assumption that AI customer service means headcount cuts, with the honest caveat that a person still reviews and sends every reply.
klarna ai layoffs in 2026: the 700-agent number and the 2025 reversal, explained
Klarna never announced AI layoffs. It said its assistant did the modeled work of 700 agents, then in 2025 reversed the AI-first hiring freeze and started re-recruiting humans after its own CEO said quality had suffered. Here's what the record actually supports.
ai back office automation in 2026: what the documented numbers actually show
Three of the most-cited back office AI deployments, Omega Healthcare, JPMorgan's COIN, and Manulife's MAUDE, read closely. Real gains, but nearly every headline number is the company's own. This post says which figures survive a check and which do not.
ai agents in banking in 2026: what the most-cited numbers actually show
The two AI figures quoted most in banking, JPMorgan's 360,000 hours and Klarna's 700 agents, are both single-source company claims. One was never independently measured; the other was reversed. Here is how to read them before you buy.
ai claims automation in 2026: what one insurer's numbers really support
The most concrete ai claims automation figure on the public record, PZU taking detailed review from about 20% of body-shop claims to nearly all, comes from a single 2020 vendor release. This post argues from that one case and is honest that TIN holds it as unconfirmed.
ai customer service automation in 2026: what the record actually shows
Vendors promise deflection and speed. Three deployments TIN checked against the public record show where the numbers hold, where they moved, and where a company walked the whole thing back.
ai finance automation in 2026: what one well-documented case actually reports
Most finance automation numbers are vendor slides. One deployment, Omega Healthcare on UiPath, is documented well enough to read closely, including two hours-saved figures that should never be merged. This post rests on that single case, and says so.
A post here is an argument. A case file is proof. Only case files carry a verification badge, and nothing on this page ever will. That separation is the point: the badge is worth something because it is scarce.