An AI skill is a packaged set of instructions, structure and guardrails that teaches an AI assistant how to complete one specific job task to a professional standard. Instead of typing a fresh prompt each time, you install the skill once into ChatGPT, Claude, Microsoft Copilot or Gemini and run it whenever the task comes up. A Business AI Skills skill defines the inputs it needs, the steps it follows, the format of the output and the limitations a human reviewer should check.

A prompt is a single message. An AI skill is a reusable capability: it carries persistent instructions, worked examples, an output template, quality checks and version history. Prompts vary in quality every time someone retypes them; a skill produces the same structured result across a whole team, which is what makes it usable in regulated work like apprenticeship reviews, care records or HR casework.

A custom GPT or Claude Project is a container you configure inside one platform. An agent is software that takes actions on its own across tools. An AI skill sits between the two: it is the portable know-how for a single task. You can load the same skill into a custom GPT, a Claude Project, a Copilot agent or a Gemini Gem, and it behaves consistently because the instructions travel with the skill rather than the platform.

No. Every skill on Business AI Skills is written for the person who does the job, not for a developer. You copy the skill into your AI assistant, paste in your source information — a session note, a job description, an observation record — and review the structured draft it returns.

In ChatGPT you can use a skill in three ways: paste it into a new chat for a one-off run, save it as a Custom GPT so colleagues can reuse it, or add it to Project instructions so every chat in that project follows the skill. Business AI Skills publishes each skill in a ChatGPT-ready format with the instruction block, the input checklist and an example output you can compare against.

Anthropic's Claude supports reusable skills and Projects that carry standing instructions and reference files. Our skills map directly onto that model: the instruction block becomes the skill or Project instructions, the reference material becomes Project knowledge, and the output template keeps Claude's responses consistent. The same skill text works in Claude on the web, in the desktop app and in Claude Code.

Claude Code is Anthropic's terminal and IDE-based coding agent. It reads project-level instruction files, so a skill can be dropped in as a repository-level file and invoked on demand. That is most useful for technical skills — documentation, test plans, data transformation, migration checklists — where the work happens in a codebase rather than a document. Business skills such as a progress review assistant are better run in ChatGPT, Claude or Copilot, where the person doing the task already works.

Codex is OpenAI's software engineering agent that plans and writes code across a repository. It is a platform, not a skill. An AI skill is the reusable instruction set you give a platform like Codex, Claude Code, ChatGPT or Copilot so it completes a defined task the way your organisation wants it done. The skill is portable; the agent is the engine that runs it.

Yes. Each product page lists the platforms the skill has been tested on. Microsoft Copilot works well where the source data already lives in Microsoft 365, and Gemini Gems work the same way as Custom GPTs. The instruction block is written in plain structured language so it transfers between assistants without rewriting.

Choose the assistant your team already has licensed and governed. ChatGPT and Claude suit document-heavy professional work, Microsoft Copilot suits organisations standardised on Microsoft 365, and Codex or Claude Code suit engineering teams. Because our skills are platform-portable, changing assistant later does not mean rebuilding your library.

Every skill produces a draft for a qualified person to review. Skills do not make decisions, and they should never be the sole basis for a record, an assessment judgement or an employment decision. Each product page lists usage limitations and the review points a professional should apply before the output is used.

Follow your own organisation's data policy and your AI platform's enterprise settings. As a rule, minimise personal data, remove identifiers where the task does not need them, and use an enterprise or business tier of ChatGPT, Claude or Copilot where your inputs are excluded from model training.

Skills are versioned. Each product page shows a version history with release notes, so you can see what changed and when. When an AI platform changes behaviour or sector guidance moves, the affected skills are revised and republished.

Yes. Custom AI skills turn in-house expertise, templates and quality standards into private skills for your team only. Message us on WhatsApp or email isa@cotalentai.com to talk through scope.

The marketplace is in pre-launch. Message us on WhatsApp at +44 7900 297464 or email isa@cotalentai.com and tell us the industry and task you need covered; early access places are allocated by sector.