Should we continue investing in the AI programme?
The board sees a €2.4m programme, six pilots and weak adoption. It wants a go/stop recommendation within ten days.
The displayed question
What the committee asks is not always the question the evidence can answer.
What is actually preventing operational value?
Technology, data, process ownership, incentives, governance and decision rights must be separated before judging the programme.
TruthX principle: never accept the initial narrative as the analytical perimeter.
Observable facts
Facts are localized, sourced and separated from interpretation.
F01
sourcedSix pilots delivered
Only one is used weekly by more than 20% of its target population.
Source: programme dashboard · month 11
F02
sourcedNo named process owner
The sponsor owns the budget, but no executive owns the redesigned workflow.
Source: governance charter · v3
F03
sourcedData quality varies
Three business units use incompatible customer and product definitions.
Source: data assessment · Q2
F04
sourcedModel quality is acceptable
Offline tests meet the threshold agreed by the technical steering group.
Source: model evaluation · release 0.8
F05
sourcedIncentives are unchanged
Teams are measured on the legacy process the programme is meant to replace.
Source: performance policy · current year
F06
openActual cost is unresolved
Internal change effort and data remediation are not included in the headline budget.
Missing: consolidated cost baseline
Competing narratives
Click a narrative to inspect its evidential coverage. No narrative is accepted because it is dominant.
A · “The technology failed”
- Supported: adoption is weak.
- Contradicted: technical threshold was met.
- Missing: production failure analysis.
B · “The employees resisted”
- Supported: usage is low.
- Contradicted: incentives favour the legacy process.
- Missing: segmented user research.
C · “The operating model was never redesigned”
- Supported: no process owner, fragmented data, unchanged incentives.
- Compatible: acceptable model quality.
- Open: full economic baseline.
D · “The use cases were poorly selected”
- Supported: five pilots have weak recurring use.
- Open: selection criteria and alternative portfolio.
- Does not alone explain governance gaps.
Select a narrative. The interface will show coverage, contradiction and missing evidence—not a probability of truth.
System and governance map
Formal responsibility and operational control do not currently coincide.
| Actor | Formal role | Actual control | Interest / constraint |
|---|---|---|---|
| Executive sponsor | Owns budget and board reporting | Can continue or stop funding | Needs visible delivery this year |
| Business units | Expected adopters | Control local workflows and data definitions | Measured on legacy performance |
| Technology team | Builds models and platform | Controls releases, not business adoption | Optimizes technical quality |
| Data office | Defines standards | Cannot compel business-unit remediation | Limited capacity and mandate |
| Consulting partner | Delivery support | Shapes roadmap and reporting | Contract rewards pilot completion |
Structural finding: budget ownership, process ownership, data authority and adoption incentives are distributed across actors who are not governed by one decision chain.
Causal hypotheses
Indicators describe evidential coverage. They are not “truth scores”.
H1 · Model performance is inadequate
One supporting fact · one contradicting fact · two missing tests.
Coverage: limited · contradiction open
H2 · User resistance is primary
One supporting observation · one incentive contradiction · user research missing.
Coverage: limited · causal direction unresolved
H3 · Governance and operating model are misaligned
Four compatible facts · no direct contradiction · economic baseline incomplete.
Coverage: strongest current explanation · reservation remains
H4 · Portfolio selection is poor
Two compatible facts · selection documentation missing · partial explanation.
Coverage: plausible contributor · not sufficient alone
Current model: the programme is being managed as a technology portfolio while value depends on business-process ownership, shared data and aligned incentives.
Decision options
TruthX makes the decision space explicit. The human decision-maker remains accountable.
Trace the reconstruction
The educational bundle records sources, open contradictions, hypotheses, reservations and the decision frame.
What OpenProof verifies
Structural conformity, deterministic serialization and whether the bundle changed after its public hash was calculated.
OpenProof does not certify that a narrative is true or that the recommended option is legally or commercially correct.
Select “Generate educational RPO”.