25%
01
Business Value Fit
Is AI pointed at work that matters?
Business outcome, current pain, leadership priority, measurability and decision relevance.
Method
Five dimensions. Distinct evidence. One traceable diagnosis.
GrundMind is a human-designed advisory method for diagnosing why AI activity does — or does not — translate into useful work and business value.
Cognitive Fit is a distinctive part of the method, but no single dimension is treated as the universal cause of AI success or failure. The diagnosis depends on how the five conditions interact and on the strength of the evidence available.
The diagnostic system
Business Value Fit01
Workflow Fit02
Cognitive Fit03
Governance & Trust Fit04
Role & Capability Fit05
Core principle
The dimensions support each other. A serious weakness in one can cap the value created by the others.
Methodology
Human-designed method, AI-assisted interpretation.
The scoring, evidence rules and reporting layers are predefined. AI can assist with a draft interpretation, but it does not determine the deterministic scores or independently publish client findings.
Predefined scoring
Five weighted Fit dimensions are calculated from structured survey evidence using versioned rules.
Evidence thresholds
A dimension is shown only when its numeric and subdimension coverage requirements are met. Missing evidence is never treated as zero.
Human publication
AI-assisted narrative drafts remain editable working material until an authorised human explicitly publishes the client-facing interpretation.
GrundMind uses a practical advisory scoring model. Its operational thresholds are not presented as validated psychometric cut-offs, and Human–AI interaction archetypes are not personality types, intelligence measures or employee-performance rankings.
The five fits
The score is a map of conditions, not a verdict.
The current overall weighting is Business Value 25%, Workflow 25%, Cognitive 20%, Governance & Trust 15%, and Role & Capability 15%. Each dimension must independently have enough evidence before the overall result is reported.
25%
02
Workflow Fit
Does AI improve the actual flow of work?
Task suitability, context, process integration, handoffs, verification and whether AI removes rather than adds effort.
20%
03
Cognitive Fit
Can people work effectively with probabilistic AI?
Ambiguity tolerance, trust calibration, control preference, cognitive engagement and verification discipline.
15%
04
Governance & Trust Fit
Can teams use AI safely and confidently?
Data-policy clarity, risk boundaries, review logic, accountability, escalation and psychological safety.
15%
05
Role & Capability Fit
Are roles and support evolving with the work?
Task reallocation, role evolution, practical capability, support structures and manager enablement.
Missing ROI
The method distinguishes why value is missing instead of assuming “low adoption.”
01
Value never created
The use case or task does not create meaningful value in the first place.
02
Value created but consumed
Workflow friction, verification or rework absorbs the gain.
03
Value created but not captured
The work improves, but the organization does not convert the gain into changed capacity or operating practice.
04
Value created but invisible
Useful change may exist, but measurement cannot demonstrate it.
Missing ROI is a deterministic interpretation of mapped survey evidence, not a calculation of financial ROI lost. The four diagnostic signal strengths do not need to sum to 100, and insufficient or unmapped evidence remains explicitly unavailable.
Manager–Team Alignment
Different evidence sources stay different.
Managers and employees may be describing different parts of the same operating system. GrundMind compares their perspectives instead of overwriting one with the other.
Employee source
What the work feels like in practice.
Access, friction, support, trust, context, handoffs and day-to-day AI use.
Manager source
What management believes is available and happening.
Priorities, policy, enablement, role expectations and management visibility.
The gap is interpreted, not averaged away.
Alignment patterns can include aligned positive, aligned negative, manager overestimation, employee overestimation, definition mismatch and insufficient evidence. A gap is a prompt to investigate — not proof that one side is wrong.
Human–AI interaction
Four interaction patterns. None is a performance ranking.
The archetypes describe how people tend to engage with probabilistic AI and what support may help the work. They are context-sensitive advisory signals, not permanent identities.
01
Co-thinker
Uses AI as a thinking partner and is comfortable exploring possibilities through iteration.
Risk if unsupported
May over-trust plausible output when verification discipline is weak.
Useful support
Keep exploration, but make uncertainty, verification and high-stakes boundaries visible.
02
Calibrator
Engages deeply with AI but wants evidence, reasoning and verification before relying on output.
Risk if unsupported
Verification overhead can consume the productivity gain.
Useful support
Provide trusted sources, transparent reasoning and efficient verification inside the workflow.
03
Controller
Works best with explicit rules, checkpoints, ownership and clear human control over consequential decisions.
Risk if unsupported
Unclear boundaries can cause under-use or unnecessary manual work.
Useful support
Define sanctioned tasks, approval points, review rules and responsibility clearly.
04
AI-averse
Is skeptical of uncertain AI output and may require narrow, demonstrated value before changing established practice.
Risk if unsupported
Broad adoption pressure can deepen disengagement without improving the work.
Useful support
Use bounded, low-risk experiments with observable task evidence and preserved human ownership.
Evidence & provenance
Every result is tied to a defined evidence snapshot.
Traceability matters because a useful diagnosis must show not only what it concluded, but what evidence, scope and version the conclusion came from.
Immutable result snapshots
Generated results retain their source campaign, scope, sample, evidence and generation context rather than silently recalculating later.
Reporting layers stay separate
Individual, Team, Manager, Manager–Team Alignment and Corporate Results answer different questions and preserve their source boundaries.
Interpretation follows evidence
When a mapping is unavailable or coverage is insufficient, the method does not infer a conclusion from question labels or a nearby score.
Foundations
Research-informed, operational by design.
Cognitive science
Trust calibration, ambiguity tolerance, cognitive engagement, verification and human judgment around probabilistic systems.
Human–AI interaction
Observed differences in how people explore, verify, control, reject and incorporate AI output into real work.
Organizational diagnosis
Business value, workflow design, governance, role evolution, evidence quality and the operating conditions around technology adoption.
Discovery call
Find out where AI value is getting lost.
If AI is already in use but the business impact is unclear, GrundMind helps you see where value is being created, where it is leaking, why, and what to fix first.