ProHance’s AI Adoption Index separates tool use from evidence of delivered work

ProHance’s [October 29 announcement](https://www.prnewswire.co.uk/news‑releases/prohance‑launches‑ai‑adoption‑index‑to‑show‑what‑ai‑is‑changing‑at‑work‑302923688.html) presents its AI Adoption Index as a scorecard connecting approved‑tool use, application to work and delivered results. The company says it maps adoption depth across teams and detects unapproved AI use with an audit trail. An approved tool is one an organization permits people to use. That status says something different from whether a person used it on a task, or whether the resulting work was accepted. A team can have permission without adoption, activity without a usable result, or a useful result that still needs human revision. For a buyer, the useful starting point is therefore the definition behind each measure. Ask whether adoption means opening an application, using an AI feature, or applying its response to a work item. Ask which employees and tools are included. Otherwise, a change in coverage can look like a change in behavior. An example of why these definitions matter appears in [our report on Cognizant’s current AI disclosures](https://opentools.ai/news/cognizant‑q3‑ai‑assisted‑code‑credentials‑revenue‑measurement). A share of AI‑assisted code, a count of credentials and total company revenue have different units and populations. That example concerns Cognizant; it is not a ProHance customer result or a comparison of the products.

The [release](https://www.prnewswire.co.uk/news‑releases/prohance‑launches‑ai‑adoption‑index‑to‑show‑what‑ai‑is‑changing‑at‑work‑302923688.html) describes a combined view of activity, governance and output. It does not publish the scoring formula, weights, comparison window or independent customer results. Those missing details matter when interpreting a score. A concrete evaluation could follow a bounded work queue through a stated period. For example, an organization considering the product could ask how an AI‑assisted draft is connected to its later review and acceptance. That is a proposed evaluation question, not a capability OpenTools verified in ProHance. The question exposes a distinction between generating an item and delivering it. If an output is rejected or rewritten, a count of generated items tells a different story from a count of accepted work. A buyer can ask who makes that judgment, how revisions appear in the record and whether a score retains the underlying evidence. The time window and population also need to travel with the result. A weekly measure for one team should not silently become a company‑wide adoption statement. An evaluation can record missing coverage explicitly instead of treating absent data as no use.

The announcement gives a current product proposition, rather than evidence that a particular customer gained a financial return. An organization evaluating that proposition can keep its questions separate: what activity was observed, what work resulted, what resources were used and what would plausibly have happened without the change. A higher adoption score by itself does not establish that AI caused better work or reduced costs. Different case complexity, staffing and review standards can change what gets delivered. The practical next step is to ask for definitions and traceable examples before converting a product score into a claim about business value. No ProHance account, integration or customer deployment was tested for this report. AI‑generated conceptual illustration of permission, work in progress and human inspection of a finished artifact. The key, papers and booklets are metaphors, not ProHance screens, customer work or measured results. Original artwork by OpenTools Team, informed by [ProHance’s announcement](https://www.prnewswire.co.uk/news‑releases/prohance‑launches‑ai‑adoption‑index‑to‑show‑what‑ai‑is‑changing‑at‑work‑302923688.html).