Discovery · Diagnostic
What the diagnostic actually asks
The Sextant diagnostic mixes validated psychometric instruments with hands-on skill tasks and structured perception items across all seven pillars. Every item ties to a specific decision this report needs to make — nothing is here for academic completeness. AI-scored items are graded by a model against a rubric, with a human reviewer confirming every score before it reaches a report.
The interactive walkthrough below is a short illustrative sample — not the full instrument.
Representative Sextant output — fictional company and modeled assumptions.
133
Total items
9
Sections
97
Structured items
36
AI-scored items
Question types
Two families of items, deliberately mixed
Classic instrumentation establishes comparable baselines across the workforce. AI-scored items measure demonstrated skill on real tasks rather than self-report — the two are not interchangeable, and the diagnostic uses both.
StructuredSelf-report, scale-scored
Likert scales, forced-choice, multi-select, and ranking items — fast to answer, directly comparable across respondents and cohorts.
- Likert Scale57A statement rated on a fixed agreement or frequency scale, used for the validated psychometric core and most perception items.Scored mechanically against the published instrument's anchor and reverse-keying rules.
- Multiple Choice30A single- or multi-select question, used for awareness, ownership-perception, and objective-knowledge items.Scored mechanically against a fixed answer key, including an explicit "I don't know" option where guessing would distort the signal.
- Ranking3An ordered list task, most often ranking task categories by AI potential.Scored mechanically by comparing the submitted order against the weighting the item is designed to surface.
- Constant Sum1A fixed budget (for example, 100 points) allocated across named categories — used once across the diagnostic, for the AI-investment 10-20-70 spend-allocation item.Scored mechanically; allocations that do not sum to the fixed total are flagged for re-entry before submission.
- Fill in the Blank1A short free-text list, used for the shadow-AI tool-disclosure item.Scored mechanically by matching submitted tool names against a maintained reference list, with unmatched entries routed for manual tagging.
- Open-Ended Response4A short free-text answer, used for data-quality pain points, a bad AI-tool experience, and general feedback.Not individually scored; aggregated as qualitative evidence and quoted by role, never by name, in the cohort report.
- Automation Spotting1A self-classification task where the participant identifies which parts of their own role show automation potential, feeding the interview agenda for selected participants.Scored mechanically as a structured signal and forwarded to interview planning; not individually graded for correctness.
AI-scoredDemonstrated skill, model-judged
Prompt-judgment and output-critique items where the participant reasons through a real scenario, scored against a rubric — not a self-assessment.
- Prompt Craft16A workplace scenario where the participant writes the prompt they would actually send an AI system to accomplish a task.A model grades the prompt text against a rubric for delegation clarity and task description; a human reviewer confirms every score.
- Scenario Judgment15A realistic workplace scenario — a trust-calibration moment, a governance edge case, a job-replacement reaction — where the participant chooses and justifies a response.A model grades the response and justification against a rubric; a human reviewer confirms every score.
- Output Evaluation4A piece of AI-generated output with a subtle seeded hallucination or bias that the participant must catch and explain.A model grades whether the flaw was correctly identified and explained against a rubric; a human reviewer confirms every score.
- Role Play1A single manager-tier pressure scenario used to elicit a frozen-middle response to being asked to automate part of a direct report's job.A model grades the response against a rubric for how the pressure was handled; a human reviewer confirms every score.
Topic scope
Every pillar, covered on purpose
Item coverage maps directly onto the seven pillars scored in the cohort report — nothing in the domain profile is inferred from a pillar the diagnostic never asked about.
Strategy16 items
- Strategy-artifact awareness
- Sponsor & budget-authority identification
- Buy / customize / build posture (stated and revealed)
- Portfolio mix & use-case-trap diagnostic
- 10-20-70 spend allocation (constant-sum)
- Decision-rights vignette
Data12 items
- Source-of-truth provenance
- Job-vs-organization data-accessibility pair
- Data-source inventory & freshness
- Master data management
- Unstructured-data governance
- Stated-vs-actual data-quality gap pair
Tech14 items
- Sanctioned-tool 30-day usage
- Tool-to-system integration
- Copy-paste frequency (integration signal)
- Shadow-AI raw behavioral frequency
- DLP / data-classification readiness
- Adversarial red-team (sanctioned-rollout reasons)
Risk16 items
- Policy-knowledge items
- Shadow-AI policy-violation framing
- Defend-this-position governance hot-take
- Scenario judgment: teammate AI-work misattribution
Talent16 items
- Operating-model ownership perception
- Training receipt & hours bracket
- Sponsor-fragmentation count
- Workflow-redesign synthesis sketch
- Roleplay decision (manager-tier pressure)
Culture71 items
- Validated psychometric core (TPS-7, GAAIS, AIAS-JR, S-TIAS, ALS-12, AICOS, UTAUT2)
- Behavioral & typology self-classification
- Shadow-AI tool disclosure
- Demonstrated skill (4D Fluency: prompt-craft + output-evaluation)
Value19 items
- Initiative-results & pilot-retirement discipline
- Per-pilot named-sponsor recall
- Observed workflow-redesign frequency
- Automation-spotting ranking
- Agent-readiness ladder frequency & comfort items
Validated instruments
Published measures, not invented questions
Several blocks draw on externally validated, published instruments rather than items written for this engagement — the credibility of a score depends on the credibility of what produced it.
| Instrument | Measures | Pillar | Items |
|---|---|---|---|
TPS-7 Team Psychological Safety (Edmondson, 7 base items) + 4 AI-specific extension items | Whether people can admit not knowing AI, report AI-assisted mistakes, and propose unconventional AI uses without fear. | Culture | 11 |
GAAIS-10 Negative General Attitudes towards AI Scale, negative subscale (Schepman & Rodway) | AI-related concern and skepticism, reverse-scored and reported separately from the positive subscale. | Culture | 5 |
GAAIS-10 Positive (trimmed) General Attitudes towards AI Scale, positive subscale, trimmed to the top 3 items by item-total correlation | Positive attitudes toward AI's potential. | Culture | 3 |
AIAS-JR Wang & Wang AI Anxiety Scale, Job-Replacement subscale only (trimmed from the full 21-item AIAS-21) | Job-replacement anxiety specifically — the subscale that best distinguishes management layers in prior fielding. | Culture | 9 |
S-TIAS Trust in AI Scale, short form (most recent validated adaptation) | Trust calibration — detecting both over-trust and under-trust; Trust and Distrust subscales scored separately, never collapsed. | Culture | 12 |
ALS-12 Wang AI Literacy Scale (12 items) | Self-reported AI literacy across awareness, use, evaluation, and ethics. | Culture | 12 |
AICOS AI Competency Objective Scale (Markus, Carolus & Wienrich) | Objective knowledge of how AI actually behaves — hallucination, bias, capability limits, prompt sensitivity — paired with ALS-12 to surface a stated-vs-actual gap. | Culture | 20 |
UTAUT2 (trimmed) Unified Theory of Acceptance and Use of Technology 2, Performance Expectancy + Behavioral Intention constructs only | Likelihood of continued AI tool use going forward. | Culture | 4 |
4D Fluency Anthropic 4D Fluency conceptual scaffold (rubric anchoring only, not a published psychometric instrument) | Demonstrated skill at delegating to, describing to, and evaluating AI output, scored via the prompt-craft and output-evaluation tasks. | Culture | 2 |
Time commitment
Effort scales with role, not a flat number
| Role | Commitment |
|---|---|
| General participant | ~45-minute diagnostic |
| Selected interviewee | 90-105 minutes total (diagnostic plus a structured interview) |
| Executive not interviewed | ~3 hours 15 minutes total (60-minute kickoff, 90-minute readout, plus the diagnostic) |
| Executive selected for interview | ~4 hours total |
| Internal coordinator | 3-5 administrative hours across the 28-day engagement |
SextantDiagnostic
Preparing0 / 24
Auto-saved · resume any time
Sextant Diagnostic · Step 1 of 3
Representative Sextant output — fictional company and modeled assumptions.
Before we begin — a word on what this is, and isn't.
This isn't a performance review. It's a diagnostic of your organization's readiness for AI. Your candor produces a sharper picture than anyone else's polish.
Your individual responses are private
Your CEO will see aggregate patterns and themes across the workforce — never your individual answers, attributed to you. Quotes used in reports are reviewed for re-identification risk before publication.
About 45 minutes for the diagnostic
You can save and resume any time. Most participants finish in 2–3 sittings. If you're selected for the follow-up AI Conversation, that adds separately.
There are no right answers — only honest ones
If a question doesn't apply to your role, mark it "skip." Your candor matters more than your pace.
After completion, you'll receive your own AI-readiness archetype report
Only you and the Sextant system see your personal report. You decide whether to share it with your manager or your team.
The diagnostic establishes the quantitative baseline. Selected participants continue to a live AI Conversation that probes the reasoning behind the numbers.
/interview/ →