Week 45, updated 2 Nov 2026

AI use cases Top 10

Top 10 AI use cases

Ranked by public usage data from 5 publishers, not by our tool scores. Reviewed 16 Oct 2026.

  • Direct evidence, 2 points
  • Indirect (adjacent) evidence, 1 point

Each publisher counts once, with its strongest finding. Equal points? More direct evidence ranks higher, then more publishers, then more cited findings, then A to Z.

Points from cited studiesreviewed 16 Oct 2026
  1. 1AI pair programming in the terminalAnthropic, McKinsey, Stack Overflow6
  2. 2AI web research reports with sourcesMcKinsey, OpenAI; indirect: Stack Overflow5
  3. 3AI customer support botMcKinsey, Microsoft4
  4. 4AI debugging with an agentAnthropic, Stack Overflow4
  5. 5Business process automation with AI agentsMcKinsey, Microsoft4
  6. 6Write cited, long-form articles with AIIndirect evidence onlyindirect: Anthropic, McKinsey, Microsoft, OpenAI4
  7. 7A company knowledge assistantMcKinsey; indirect: OpenAI3
  8. 8Build AI agents in Python or TypeScriptStack Overflow; indirect: McKinsey3
  9. 9Long-term memory for AI agentsSingle publisherStack Overflow2
  10. 10Chat with your documents (RAG)Indirect evidence onlyindirect: McKinsey, OpenAI2

One finding per use case

The featured cited finding for each. Open “All evidence” for the rest.

  1. 1

    AI pair programming in the terminal

    Coding remains the most common use on Anthropic's platforms: tasks tied to Computer and Mathematical occupations were 35% of Claude.ai conversations (February 2026).

    Anthropic, March 2026. Direct evidence.

    All evidence (3)
    • Direct: Developers partly using AI used it for writing code (59%), searching for answers (56%), learning new concepts (47%) and debugging or fixing code (47%). Stack Overflow, December 2025
    • Direct: Respondents most often report scaling AI agents in IT, knowledge management and software engineering. About two in ten are scaling software coding agents (31% at larger enterprises), and 32% decided against buying software because agentic coding tools could build it in-house. McKinsey, August 2026

    How people rate it: in the same survey 52% of developers say AI tools or agents improved their productivity, but 66% name "almost right, but not quite" output as their top frustration and more distrust AI accuracy (46%) than trust it (33%).

  2. 2

    AI web research reports with sources

    "Seeking Information" was 24% of ChatGPT messages in July 2025, and the paper describes it as appearing to be a very close substitute for web search.

    OpenAI, September 2025. Direct evidence.

    All evidence (3)

    How people rate it: Stack Overflow found developers most satisfied with using AI to search for answers and learn new concepts.

  3. 3

    AI customer support bot

    Among individual use cases, respondents most often report using AI to capture, process and deliver information through a conversational interface, for marketing content support, and for contact-center or customer service automation.

    McKinsey, November 2025. Direct evidence.

    All evidence (3)

    Adoption evidence is from company surveys, not from customer satisfaction data. We found no sourced figure for how customers rate AI support.

  4. 4

    AI debugging with an agent

    The single most common task in November 2025, modifying software to correct errors, was 6% of Claude.ai usage.

    Anthropic, January 2026. Direct evidence.

    All evidence (2)

    How people rate it: 45% of developers say debugging AI-generated code takes more time (Stack Overflow 2025), so review every fix.

  5. 5

    Business process automation with AI agents

    Leaders named customer service, marketing and product development as the top three areas for accelerated AI investment, and 46% say their companies use agents to fully automate workflows or processes.

    Microsoft, April 2025. Direct evidence.

    All evidence (2)

    Agents are not mainstream yet: Stack Overflow found 52% of developers do not use agents or stick to simpler AI tools.

  6. 6

    Write cited, long-form articles with AIIndirect evidence only

    "Writing" was 24% of ChatGPT messages (July 2025) and the most common work topic at 40% of work-related messages; about two thirds of writing messages edit, critique or translate existing text.

    OpenAI, September 2025. Indirect (adjacent) evidence.

    All evidence (5)

    Evidence here is adjacent only: the studies measure AI writing in general. Our use case is the narrower research-backed, cited variant.

  7. 7

    A company knowledge assistant

    Agent use is most commonly reported in IT and knowledge management, with service-desk management in IT and deep research in knowledge management named as quickly developing examples.

    McKinsey, November 2025. Direct evidence.

    All evidence (4)
  8. 8

    Build AI agents in Python or TypeScript

    Among developers who build agents, Ollama (51%) and LangChain (33%) are the most-used orchestration and agent frameworks.

    Stack Overflow, 2025. Direct evidence.

    All evidence (2)
  9. 9

    Long-term memory for AI agentsSingle publisher

    For AI agent memory or data management, developers who build agents most often use Redis (43%), alongside the vector-native databases ChromaDB (20%) and pgvector (18%).

    Stack Overflow, 2025. Direct evidence.

    All evidence (1)

      Counts are tool usage among developers who already build agents, a small group.

    • 10

      Chat with your documents (RAG)Indirect evidence only

      …Respondents most often report scaling AI agents in IT, knowledge management and software engineering…

      McKinsey, August 2026. Indirect (adjacent) evidence; full finding under #1.

      All evidence (2)

      Evidence here is adjacent only (information seeking and knowledge management in general).

    How the points work

    2points for a finding that measures this exact job (direct)
    1point for a finding about a nearby job (indirect)
    1×each publisher counts once, its strongest finding
    2+points needed to be ranked at all
    Full method
    1. We collected findings from published usage studies and surveys, preferring 2025 and 2026. Every finding below links to its source, and a finding without a link was left out.
    2. Each finding was mapped by hand to the AI use cases on this site. A finding is "direct" when it measures that very job (for example debugging code) and "adjacent" when it measures a nearby job (for example AI writing in general, for our research-backed article use case). That mapping is our editorial judgement and the one place where we could be wrong.
    3. Scoring is deliberately simple: 2 points for direct evidence and 1 point for adjacent evidence, counted once per publisher, so one organisation cannot add points by publishing twice. Ties are broken by direct evidence, then number of publishers, then number of cited findings, then title A to Z, so every rank is unique. The cited-findings step only orders items with equal points; points still count each publisher once.
    4. Only use cases with at least two points are eligible. We do not convert shares from different studies into one number, because the studies measure different things (messages, conversations, developers, companies).

    Where the evidence comes from

    10 studies from 5 publishers.

    Jan 2025Jul 2025Jan 2026Jul 2026 Stack Overflow2025 Stack Overflow Developer Survey: AI, 2025Tell us what you really, really… do not want to spend time working on (2025 Developer Survey analysis), December 2025McKinseyThe state of AI: How organizations are rewiring to capture value, March 2025The State of AI: Global Survey 2025, November 2025The state of AI in 2026: On the road to ROI, August 2026Microsoft2025 Work Trend Index: The year the Frontier Firm is born, April 2025OpenAIHow People Use ChatGPT (NBER working paper w34255, with OpenAI), September 2025AnthropicAnthropic Economic Index report: Economic primitives, January 2026Anthropic Economic Index report: Learning curves, March 2026Anthropic Economic Index report: Cadences, June 2026
    One square per cited study, placed by publication month (studies dated by year only are placed in January). Squares link to the source.
    Limits of this ranking
    • These sources disagree about what people do most. Anthropic reports coding as the most common use of Claude, while OpenAI reports that programming is a relatively small share of consumer ChatGPT use. Each lab sees only its own users, and Claude's user base skews technical.
    • Surveys of developers and companies measure adoption and intent, not quality. Stack Overflow found more developers distrust AI output than trust it, so "most used" is not the same as "best".
    • Counts of independent publishers are small by nature (five publishers in this pass), so a one-point difference is not a strong signal. Treat the order as a guide.
    • Figures are as reported at publication. We did not re-run any study.
    What we left out
    • Learning, tutoring and everyday practical advice: the biggest category in ChatGPT use (practical guidance, 29%) and a top developer use (learning new concepts, 47%). We have no use case page for it yet, so it is not ranked.
    • Code review, test writing, documentation generation, refactoring, local models and media: we found no sourced usage share for them. Stack Overflow reports that 39% of developers do not plan to use AI to document code, which argues against ranking documentation high.
    • Deloitte, Gartner and other consultancy surveys: none were verified in this pass, so none are used.
    • Secondary blog summaries of these studies: left out where the primary figure could be read directly. Two summaries of the Stack Overflow survey quoted different percentages for searching for answers, so we used the primary Stack Overflow figure (56%).
    • Any figure without a public URL.
    How people rate AI at work
    • McKinsey (2026): 80% of respondents say AI improved their individual productivity, but only 37% report any enterprise-level EBIT impact.
    • Anthropic Economic Index Survey (June 2026): users report gains in speed (86%), scope (82%) and quality (69%). Respondents were Claude users, mostly technical.
    • Stack Overflow (2025): 52% of developers report a positive productivity effect, while trust in AI accuracy is lower than distrust.
    Introduction

    Which AI use cases do people actually use most, and which ones do they rate well? This list answers with public numbers only: usage studies from AI labs and surveys of developers and companies. It is a ranking of use cases by cited public usage data. It is not our own score, and it says nothing about how good any single tool is (tool scores live on the tool pages).

    Get the board every Monday

    One email a week: the new top 10, the biggest score changes and every safety label that changed.

    Free. Double opt-in: you confirm by email first, unsubscribe in one click.