ArkCloud Insights capabilities

Insights reads what the other products have already recorded. Every figure it shows is counted from persisted rows in the tenant asking for it — there are no sample dashboards and no illustrative numbers.

Its analytics - financial, payer performance, denials, clinical, rehabilitation outcomes, quality indicators, benchmarking and forecasting - compute measures from those rows with the same metric definitions the revenue-cycle and Rehab screens use. A rate with nothing to divide by reads “Not enough data”, never 0%.

Each states its own limit. Benchmarking compares an organisation with its own history; an external benchmark is shown only when it names its source, and there is no peer ranking, because comparing with other organisations would need their data. Forecasting projects from the organisation's own collection history and shows no confidence range it cannot support. Quality indicators are operational - not CMS eCQM, HEDIS or MIPS measures. Rehabilitation outcomes counts administrations but reports no change in score until outcome measures are published.

Each analytics screen also has an Explain with AI panel, labelled as AI. When the screen opens it shows a short summary of the figures on it, and it answers a short question about them. It is given only the totals, rates and per-payer or per-programme rows that screen already shows, with the screen's method note - no patient records. An explanation that contains a number the screen does not show, or that talks about what will happen, is withheld and says so; a figure the screen shows as “Not enough data” stays that way. A summary is reused for a few minutes rather than written again on every view. It runs on ArkCloud's own model server on its private network, not a third-party AI service. It describes the figures; it does not recommend which claims or accounts to work, and nothing on the screen changes because of it. When the model is unavailable the panel says so and the screen works as before.