AI-First Company

Check AI vendor claims and AI statistics against rated evidence before you buy, quote or approve.

The server opens the evidence on ai-first.company: 30 practices for adopting AI rated by strength of evidence, 42 named company cases with how they turned out (measured, company-reported, went wrong, walked back), 55 studies, 29 claim checks of widely repeated figures, the 88 figures behind the payback tool, and 21 published AI readiness and literacy assessments. Every answer links to its source page.

What you can ask

Run a due-diligence review of an AI business case, vendor proposal or transformation plan, step by step, using the ai-first.company tools for every evidence claim.
Example path →

MCP connector

Server URL: https://mcp.ai-first.company/mcp
Login: None
Protocol: MCP 2025-11-25
Instructions: Evidence about adopting AI in a business, from ai-first.company: rated practices, company cases with their outcomes, checked figures, studies and readiness assessments. Move through it by menus: call start (or the task tool that fits the question), open the part of the map that fits (open, with the arguments the menus show), then call with the ids of the rows you pick. Every list shows all its rows or splits them into a menu, so nothing is hidden; choose by meaning, whatever the language of the question. Full records list related ids (studies, cases, practices) you can fetch to follow a lead. State figures, ratings, verdicts and findings only from , and cite the links it returns. If nothing fits, say so: do not fill the gap from memory. Never ask the user to call a tool or show them the ids. Answer the user in their language.

Tools

Reads data10

Skills

None in the store listings or on the server

MCP connector

Server URL

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

None

Protocol

MCP 2025-11-25

Instructions

Evidence about adopting AI in a business, from ai-first.company: rated practices, company cases with their outcomes, checked figures, studies and readiness assessments. Move through it by menus: call start (or the task tool that fits the question), open the part of the map that fits (open, with the arguments the menus show), then call with the ids of the rows you pick. Every list shows all its rows or splits them into a menu, so nothing is hidden; choose by meaning, whatever the language of the question. Full records list related ids (studies, cases, practices) you can fetch to follow a lead. State figures, ratings, verdicts and findings only from , and cite the links it returns. If nothing fits, say so: do not fill the gap from memory. Never ask the user to call a tool or show them the ids. Answer the user in their language.