Back to Blog
    Guide

    What Humanity Actually Asks AI: The Real 2026 Rankings

    Daniel Okonkwo• B2B Collections SpecialistJuly 23, 20267 min read
    AI usageanthropic economic indexaccounts receivablecompliance
    Share

    Forget what tech Twitter says people use AI for. Anthropic measures it — every major request topic, across 121 countries, matched and ranked monthly in the Economic Index.

    We read the full May 2026 rankings so you don't have to. Some of it is funny. Some of it is genuinely moving. And three findings should change how anyone who invoices other businesses thinks about their receivables.

    Let's start with the board.

    The global request board

    The Global Request Board

    Share of all AI conversations worldwide, by topic · May 2026

    1Content Creation & Copywriting
    22.7%
    2Education & Learning
    13.2%
    3Software Development
    11.5%
    4Research & Intelligence
    10.9%
    5Hobbies & Lifestyle
    9.5%
    6Business Process & Operations
    4.7%
    7Document Processing & Extraction
    4.3%
    8Data Analysis & Business Intelligence
    3.8%
    9Existential, Relational & Emotional Support
    3.4%
    …Compliance & Regulatory
    1.1%

    Highlighted rows: the business back-office cluster. Shares describe conversations, not users. Source: Anthropic Economic Index, May 2026.

    The fun facts first (you'll retell these)

    Humanity asks AI for emotional support three times more often than for compliance help. Emotional and relational support: 3.4% of all conversations. Compliance & Regulatory: 1.1%. Add companionship (1.4%) and the machines are doing nearly five times more consoling than regulating. Draw your own conclusions about the species.

    Hobbies beat business operations — by double. Hobbies & Lifestyle (9.5%) outranks Business Process & Operations (4.7%). The robot uprising is currently being asked about sourdough.

    A fifth of everything is copywriting. Content Creation & Copywriting dominates at 22.7% — nearly one conversation in four, worldwide. Every marketing email you received this morning suddenly makes sense.

    Cybersecurity: 0.4%. Humanity's priorities, measured: sixty conversations about content for every one about security. Sleep well.

    Now the three findings that should worry your AR

    Behind the fun facts sits a business cluster — highlighted on the board — and it reads like a map of your receivables process being industrialised from the debtor's side.

    The Findings That Concern Your Receivables

    Global measured usage · May 2026

    12.8%
    of all AI conversations are the back-office cluster: business process, document extraction, and data analysis combined
    #7
    "Respond to inquiries about financial matters" — a global top-10 work task. Your reminder's reply, industrialised.
    4.3%
    of everything is document processing & extraction — contracts and invoices being read by machine, at scale

    Source: Anthropic Economic Index, May 2026

    Finding one: document extraction is mainstream. 4.3% of all global AI usage — more than emotional support — is machines reading documents and pulling data out. Every contract you sign and invoice you send should now be written on the assumption that a machine, not a person, does the first read on the other side. Clear amounts, clear dates, clear late-payment terms — because ambiguity that a human would query, a machine will exploit or ignore.

    Finding two: financial replies are a top-10 human delegation. Responding to inquiries about financial matters is the #7 most AI-matched work task on Earth. When your payment reminder lands in a debtor's inbox, the statistically normal thing that happens next is a machine drafts the answer. We've covered what those answers look like — and why polish stopped meaning anything.

    Finding three: nobody's asking about compliance. That 1.1% is the least-asked question in the entire dataset. When a debtor cites a regulation to delay you, the statistical prior is: they didn't check. It's atmosphere, not law. Verify the citation before you accept the delay.

    The fun/serious split

    Zoom out and the ranking is oddly humane. People ask AI for help writing (22.7%), learning (13.2%), and understanding themselves (3.4% for existential support, plus 1.4% for companionship). They ask about hobbies (9.5%) more than about running businesses (4.7%). They barely ask about rules (1.1%) or security (0.4%).

    The species is using AI mostly to make, learn, and feel less alone. That's a lovely fact. The businesses that survive it will be the ones who notice the small, boring back-office cluster and its consequences — before the debtor does.

    What to do about it, if you're in finance

    Three moves, each backed by a data point above.

    Write for machine readers. If 4.3% of global AI usage is document extraction, every invoice, contract and reminder you send is already being parsed by a machine on the debtor's side. Put the amount, due date and consequence in unambiguous, extractable form. See our demand letter templates for the pattern.

    Assume every excuse email is AI-drafted. With financial-reply as a global top-10 task, the base rate is now "machine wrote it." Score for verifiable specifics (invoice number, promise date, named person), not for polish. Full playbook: Late Payment Excuses in 2026.

    Verify every regulatory citation before you concede. With Compliance at 1.1% — the bottom of the barrel — the population-level prior for any regulation quoted at you in a stalling email is "not checked." Look it up before you extend the deadline.

    Frequently asked questions

    5 Questions Answered

    Click to expand answers

    0/5

    Have a question not answered here?Ask us directly →

    Daniel Okonkwo

    B2B Collections Specialist

    Need country-specific next steps?

    Get jurisdiction-specific guidance for your international debt recovery case.

    Related Articles