Your team is producing more with AI. You are not sure they are getting any better at their work.
Output is up. Something has changed in how the team handles novel problems, defends its reasoning, operates when the tools are unavailable. You cannot quantify it yet. That gap is measurable, and it is the most important question in AI deployment that almost no one is asking.
The concern usually arrives before there is data to support it. Work that would previously have required genuine thinking now gets handed to an AI tool first. People are less willing to take on problems where the AI cannot lead. Someone who used to reason through an argument independently now finds it difficult to do so without assistance. The output metric says everything is fine. What you are watching says otherwise.
This is the substitution pattern: AI absorbing not just execution work but the work that was developing judgment, building range, maintaining the capability stock the team depends on. It does not show up in productivity data because the output does not change, or improves. The capability underneath is what is contracting, and that only becomes visible when the tools fail, when a genuinely novel situation requires unassisted judgment, or when someone is asked to explain their reasoning from first principles.
By the time the problem is visible in output data, it has usually been accumulating for over a year. The point of measuring it now is that intervention at this stage is straightforward. The options narrow considerably once the gap becomes obvious.
The Kaivant-I measures two axes simultaneously. Personal Leverage Architecture covers how effectively AI is integrated into each person's workflow: whether the leverage is genuine and building. Human Capital Depth covers what is happening to the capability underneath: whether judgment independence is stable, whether cognitive range is expanding, whether people can perform at material quality when the AI tools are unavailable.
The dimension that matters most in this situation is Capability Independence and Resilience: a direct measure of whether each person's baseline capability is stable or declining. A team member can score well on AI integration while CIR is falling. That combination is the substitution signature: strong leverage built on eroding human capital. It is the pattern your instinct is detecting, and it is measurable before it becomes a performance problem.
The composite score is non-compensatory: strong AI leverage cannot offset declining human capital depth. If a team member is producing well with AI but losing independent capability, the score reflects both facts separately. You see the full picture, not an average that conceals what is actually happening.
Each team member completes the Kaivant-I individually. Results are private to the individual by design: the CIR dimension in particular is not shared with the organisation, because accessible dependency data would create the wrong incentives. What you receive at the team level is the pattern: which dimensions are strong, where the substitution signal is appearing, and whether the concern you have is confirmed or contradicted by the actual data.
The facilitated session turns that pattern into a plan. The team reads its results together, identifies the specific dimensions driving the risk, and agrees on two or three targeted changes. The most effective interventions at this stage are not dramatic. They are structural: reintroducing the types of work that develop the specific dimensions that are declining, without reducing the AI leverage that is genuinely adding value.
The goal is not to use AI less. The goal is to use it in ways that compound capability rather than substitute for it. That distinction is achievable, and the Kaivant-I shows you exactly where the current pattern diverges from it.