Data

Data

The lists of company cases, assessments, practices, studies, claim checks and payback figures as files you can download and reuse.

The files

Each list comes in two formats. CSV opens in any spreadsheet. JSON is for programs. Both hold the same rows.

Company cases

Named companies that used AI at work, one row per case. Each row gives the company, what it does, the country, the period, the size band, the kind of work AI was used for, how the case turned out, who produced the figures, the practices the case bears on, and the public sources.

39 rows · version 17 October 2026 · CSV · JSON · CC BY 4.0 · See the list on the site

Assessments

Published instruments for assessing where a company stands with AI, one row per assessment. Each row gives the question it answers, its type, how well it has been tested, who answers it, its length, licence, cost and languages, and links to the provider and the source publications.

21 rows · version 13 October 2026 · CSV · JSON · CC BY 4.0 · See the list on the site

Practices

Management practices for building and running a business with AI, one row per practice. Each row gives a summary, the theme, the stage, the company sizes and industries it fits, when to do it, the effort, the evidence level and flags, the studies for and against, and the company cases linked to it.

30 rows · version 17 October 2026 · CSV · JSON · CC BY 4.0 · See the list on the site

Studies

The studies the practices and assessments rest on, one row per study. Each row gives the main result, the authors, year and publication, how the study was done, whether it was peer-reviewed, who took part and where, and the practices it supports or limits.

53 rows · version 17 October 2026 · CSV · JSON · CC BY 4.0 · See the list on the site

Claim checks

Figures about AI that are often quoted, each checked against the source it is credited to, one row per claim. Each row gives the claim as it is usually repeated, the verdict, why, the sentence to cite instead, and the sources.

7 rows · version 17 October 2026 · CSV · JSON · CC BY 4.0 · See the list on the site

Payback figures

The figures behind the payback tool, one row per figure. Each row gives the figure, its unit, who or what it covers, what kind of number it is (measured, self-reported, calculated, reported), the line of the result it feeds, the source with a quote, and its caveat.

88 rows · version 13 October 2026 · CSV · JSON · CC BY 4.0 · See the list on the site

How the files are built

  • The files are made from the same source as the pages, each time the site is published. They always match the site.
  • Each row has the address of its page on the site. The full text is on that page, not in the file.
  • The version is the latest review date among the rows. The payback figures take the review date of the payback page.
  • In the CSV files, a field with several values separates them with a semicolon and a space.
  • Studies, cases and practices are named by the last part of their page address, such as generative-ai-at-work.
  • The files are in UTF-8.

Use with AI assistants

An AI assistant can search these lists for you. Connect it to the site’s server, then ask in plain words, for example “Is it true that 95% of AI pilots fail?”. Each answer links to the page it comes from.

Server address: https://mcp.ai-first.company/mcp

  • Claude: go to Customize → Connectors, click “+ Add”, then “Add custom connector”. Paste the server address, choose “No sign in” and click “Add”.
  • ChatGPT: on the web, open Plugins, click “+”, then “Add custom MCP server”. Paste the server address, choose “No authentication” and select “Create as a plugin”. In a company workspace, an admin may have to add it.
  • Cursor: add these lines to .cursor/mcp.json:
    {
      "mcpServers": {
        "ai-first": {
          "url": "https://mcp.ai-first.company/mcp"
        }
      }
    }
  • Claude Code: run this command:
    claude mcp add --transport http ai-first https://mcp.ai-first.company/mcp

The server never receives your questions: your assistant sends it only choices from its menus. It counts how often each tool, filter and record is used, and by which assistant app, and sends the totals to PostHog every ten minutes. It keeps no log of requests and stores no IP addresses or user IDs.

How to cite

Name the author, the list, the version and this page. For example:

Szczęsny, P. (2026). Company cases of AI use, with outcomes and sources [Data set]. Version 2026-10-17. AI-first company. https://www.ai-first.company/data/

Szczęsny, P. (2026). AI readiness assessments: questionnaires, tests, task maps and interview guides [Data set]. Version 2026-10-13. AI-first company. https://www.ai-first.company/data/

Szczęsny, P. (2026). Practices for running an AI-first business, rated by strength of evidence [Data set]. Version 2026-10-17. AI-first company. https://www.ai-first.company/data/

Szczęsny, P. (2026). Studies of AI at work, with how each was done [Data set]. Version 2026-10-17. AI-first company. https://www.ai-first.company/data/

Szczęsny, P. (2026). Checks of widely repeated AI figures against their primary sources [Data set]. Version 2026-10-17. AI-first company. https://www.ai-first.company/data/

Szczęsny, P. (2026). Figures on the time AI frees, the time put to use, its cost and its risks, with sources [Data set]. Version 2026-10-13. AI-first company. https://www.ai-first.company/data/

Licence

Data: CC BY 4.0, Paweł Szczęsny, ai-first.company.

The data files on this page are under the Creative Commons Attribution 4.0 licence (CC BY 4.0). You may copy, share and adapt them, also for commercial use, if you credit Paweł Szczęsny and link to this page.

The licence covers these data files only. It does not cover the text of the site.