IonBin
All pages
Search
Question

How do I find out everything an AI tool can actually do?

Combine the vendor's current documentation with the tools, settings and permissions visible in your own account. Ask the assistant for leads, then verify important capabilities with a low-impact test. Record plan, app, date and limits so a feature announcement does not become an unsupported team promise.

Verified Oct 18, 2026 Review by Jan 16, 2027 2 sources

What it is

Capability discovery is a record of which operations are documented and available in a particular account, application and configuration. It separates a product announcement from a verified result in your working environment.

How it works

  1. Define the exact operation: read a document, edit a record, export a file or run a task while the computer is asleep. “Connects to the app” is too broad.
  2. Ask the assistant what tools are available in this conversation. Treat its description as a lead; it may omit tools or describe a capability it cannot currently use.
  3. Check current vendor documentation and release notes. Match the plan, region, app, rollout stage and administrator policy.
  4. Inspect your actual account settings and connector permissions. A model’s API capability can differ from a consumer app, a coding agent or a workspace integration.
  5. Run a low-impact test with sample data. Verify the resulting artifact or destination state, not merely the assistant’s success message.
  6. Write a short capability record: source URL, date checked, account scope, operation tested, limits and actions requiring review.

Useful examples

OpenAI’s release notes distinguish features available in Chat, Work, Codex, desktop and web, often with rollout or workspace restrictions. Anthropic’s release notes similarly describe feature-specific availability. A release announcement shows documented availability; it does not establish that your account has access.

Benefits

A small, dated capability record gives teammates a repeatable starting point. It identifies missing access or uncertain steps before someone promises a workflow, and provides concrete cases to recheck after an update.

Limits and risks

Keep permissions and actual execution separate. Approve or pre-authorize actions according to a clear policy, especially publishing, messages and purchases. Retest important workflows after a client, connector or model update.

This method discovers supported, useful operations; it cannot establish an exhaustive list of everything a changing product can do.

Best-fit stacks

Stack 06

Ops & admin automation

Hand repeat admin (receipts, logs, notices, handoffs) to workflows your agent can trigger.

Agent compatibility

Tool in this entryHow agents connect
ChatGPT Agent platform API/CLIFile import
Claude Agent platform MCPAPI/CLIFile import
Gemini (Google) Agent platform API/CLIFile import

Levels come from each tool's recorded, sourced compatibility. How the levels work.

Sources

  1. ChatGPT release notes · OpenAIofficial accessed Oct 18, 2026
  2. Claude release notes · Anthropicofficial accessed Oct 18, 2026

Practitioner views are restated in our words and labeled as views, not measured results. Product facts come from official sources dated above.

Suggest an edit

IonBin's wiki isn't user-edited. Corrections go to our Directory Curator, who checks them against primary sources before anything changes. Spot something wrong or out of date on “How do I find out everything an AI tool can actually do?”? Email us with a link to a primary source.

Suggest an edit by email or write to contact@ionbin.com