Staging preview. Search engines are asked not to index this deployment.

Specialist assistants

Embed guidance where the work happens-without pretending a platform is live.

Oddly Capable treats assistants as a product pattern: a defined job, approved guidance, known boundaries, human review and evidence that the work improves. Phase A explains that pattern and the current availability.

Configured availability: In development. No public assistant, chat or account feature is live on this site.

Product pattern

A specialist assistant should have a job, not a personality.

The useful question is not whether AI can answer. It is whether a bounded assistant can help people use trusted guidance in a repeatable piece of work, with the right review and ownership around it.

  1. Define the work.

    Name the task, decision, audience, owner and outcome before choosing any model or tool.

  2. Use approved sources.

    The assistant should draw from guidance the organisation is willing to stand behind.

  3. Set boundaries.

    Make the assistant's role, escalation points, data limits and failure modes visible.

  4. Keep human judgement.

    The person using the answer remains responsible for context, consequence and final action.

  5. Measure the work.

    Success is clearer, faster or more reliable work, not the existence of a chat interface.

Current Phase A availability.

Phase A

Available now

The public site includes free browser tools, the product catalogue, service journeys and the explanation of how specialist assistants would be designed.

Use the tools

Platform boundary

Not live yet

There are no live assistants, chat controls, uploads, accounts, subscriptions or checkout flows on this site. Those need separate product, security and provider decisions before they are built.

See consultancy options

When an assistant might make sense.

A specialist assistant may be useful when people repeatedly need the same guidance, the source material can be trusted, the work has a clear owner and mistakes can be reviewed before they matter.

Where those conditions are missing, the better first move is usually a diagnostic, framework, handover or workflow improvement.

Start with the work you want to improve.

Tell us what you would like to make clearer, easier to use or more reliable. You will get an honest view of whether a free tool, a focused session or deeper work is the sensible next step.