Most failed initiatives were working on the wrong problem. Are you solving the right one?

A structured diagnostic tool that applies consulting-grade problem analysis to whatever you're grappling with. You describe the problem as it arrived on your desk — messy, incomplete, as you'd explain it to a trusted advisor. Premis classifies it, surfaces the most likely diagnosis with the reasoning behind it, builds an issue tree to test that diagnosis, and designs the tests needed to prove or disprove it. The output is a structured brief you can act on or share.

Most AI tools move straight from problem to solution — which is exactly the failure mode Premis is designed to correct. Premis is built around the diagnostic step that consultants treat as the most valuable part of an engagement: defining what is actually wrong before designing any response. It shows its reasoning, names what it doesn't know, and produces output structured for a decision-maker — not a chatbot conversation.

Problems where the diagnosis is genuinely unclear — where smart people in the room disagree on what's broken, or where the obvious explanation feels incomplete. Strategy questions, organisational failures, declining performance, stalled initiatives, market decisions. It works best when the cost of misdiagnosing the problem is measured in months of effort or significant resource, not hours.

A four-stage diagnostic: what kind of problem this is and what decision actually needs to be made; the primary diagnosis and the alternatives it was ranked against; an issue tree structured to test that diagnosis; and a set of designed tests with explicit criteria for what would confirm or abandon each hypothesis. You can download the full output as a structured PDF brief.

Premis is explicit about this. Every output includes a confidence rating and a list of what the analysis was working without — the missing evidence that would change the read. The diagnosis should be treated as a structured hypothesis to test, not a verdict. That is by design. The value is in the quality of the question Premis sends you away with, not in replacing your judgment.

Premis logs diagnostic runs for rate-limiting and quality purposes. It does not store or share your input beyond what is required to process the request. For that reason, the terms ask that you do not enter confidential or proprietary information. If you need a version that runs entirely on local infrastructure with no data leaving your device, that capability is in development.

Your input is processed by the Anthropic API. Anthropic's standard data handling policies apply. Until a local inference version of Premis is available, treat it the way you would any cloud-based AI tool — suitable for real problems described in general terms, not for verbatim internal documents, unreleased financial data, or named individuals.

Each diagnostic run calls the Anthropic API, which carries a cost. The current limit of three runs per hour per user is a practical constraint during this early stage, not a product decision. If you need more runs, get in touch.

Yes, for now. Premis is in private beta. Access is free while the tool is being tested and refined. Pricing for wider availability has not been set.