Performance management is one of the seven dimensions with which hybridresourcing measures an organization's maturity when it comes to AI. This dimension is about something specific: can you determine whether work supported or taken over by AI actually becomes better, faster or cheaper? Many organizations start pilots without being able to answer that question, and therefore remain stuck in a phase of trying without being able to demonstrate what it delivered.
Performance management can only mature if other dimensions already provide something to build on. Without reliable data from data management, a performance measurement is an estimate. Without clear responsibilities from the organization dimension, as elaborated on the page about what it costs to raise organization by one level, no one is accountable for the outcome. And without an IT infrastructure that can record and repeat measurements, as discussed on the page about what it costs to raise IT infrastructure by one level, every measurement remains a one-off. That is the reason why, in the assessment, this dimension can only move to a higher level once the fundamental dimensions already carry something.
Baseline. There is no fixed way to assess whether AI-supported work performs better than the old way of working. Assessments are incidental and person-dependent. If someone leaves, the only view that existed disappears with them.
Foundation. There are separate measurements, usually per project or per pilot. They are rarely compared, and there is no fixed moment at which someone looks back to check whether the intended effect actually occurred. The measurement exists, but does not function as a learning process.
Activation. There is a recurring rhythm: measurements are repeated at fixed moments and fed back to the people doing the work. The organization is starting to see patterns, even though these are not yet systematically recorded or compared between departments.
Insight. Measurements are used to make adjustments, not just to report. There is a shared picture of what does and does not work, and that picture leads to changes in how AI is deployed. Different teams can compare their outcomes with each other because the way of measuring is the same.
Intelligence. Performance management is embedded in the way decisions about AI are made. Outcomes from earlier deployment directly feed the choices for new applications. The organization can substantiate why something is or is not scaled up, rather than basing that on gut feeling or on who argues most forcefully for a next pilot.
The level of this dimension is not found in an instrument or a dashboard alone. It lies in the behavior surrounding it: is there a look back after a pilot, or does everyone move straight on to the next one? Can someone in your organization show, on request, what an AI application has actually delivered, apart from an assumption made beforehand? Is a disappointing result discussed, or quietly forgotten? These are not technical questions. They are about whether a habit has formed of looking, not just of doing.
Within this dimension, considerable spread often arises between those who fill in the plot round. An operational manager sees the separate measurements from the activation phase on a daily basis and scores the organization higher than reality justifies, because those measurements are tangible to him. A CFO who only sees the quarterly report scores lower, because he gets no ongoing view of what has been measured in between. Both pictures are true for the person giving them. The spread itself is the information: it shows that there is no shared, organization-wide view of performance, which in itself already says something about the level.
A step from baseline to foundation mainly requires discipline: choosing a fixed moment to look back and not letting that moment disappear under the pressure of the next pilot. A step to activation requires a repeatable process, which in turn touches on how mature the underlying infrastructure is. A step to insight or intelligence requires measurements to be reliable enough to act on, and that leads back to data quality and to who within the organization is authorized to make adjustments based on those measurements. Dimensions such as ethics and privacy and security also play a role here: a performance measurement that ignores the question of whether an outcome came about responsibly is an incomplete measurement. What a step precisely costs, in time, attention or resources, differs per organization and depends on where the other dimensions stand. That is why hybridresourcing works with a plot round instead of a fixed step-by-step plan: first making visible where you stand and where opinions diverge, before anything is laid down about what is needed.
This page is about whether your organization can determine what AI delivers, regardless of which specific work it affects. That is a different question from which tasks can be taken over and to what extent. Anyone who wants to know that, per task and based on what the work itself entails, will find it in the work scan of FTE TO AI. Performance management and the work scan connect to each other: one measures whether you can assess an outcome, the other calculates which part of the work there is to assess.
Vraag maar wat er moet staan voordat AI in uw organisatie kan landen.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.