Meridian Claims Engine
An illustrative read of an autonomous financial-decisioning system that reaches the top of the scale. It is the deliberate counterpart to FinCo: the same system shape, the opposite governance, and the opposite result.
Meridian is not a real organisation and this is not a real assessment. It is built to show what evidenced governance looks like, and that the result discriminates: a standard that can only describe failure has not shown that it can recognise governance done well. A DSI 001 result can be issued only by an assessor authorised by Decision Standards Institute under the DSI 001 scheme, against the methodology.
The system
An autonomous claims-decisioning system that settles and pays straightforward claims at machine volume, the same autonomy shape as the FinCo credit engine.
The Meridian Claims Engine assesses incoming claims, decides straightforward matters, sets payment within a parameter range, and commits the insurer's funds. It runs at Level 3 operational autonomy, deciding the large majority of standard claims without per-decision human authorisation. What separates it from FinCo is not the autonomy. It is everything the deploying organisation built around the autonomy before it went live.
On the 1.0 to 5.0 scale, where lower is stronger, the composite sits at or below the 1.75 threshold for Certified, the top classification, which also requires contemporaneous Tier 1 evidence and ongoing surveillance.
Two reads, both strong
DSI 001 reads a system twice. The GBI asks whether governance is adequately designed. The Governance Coherence Index asks whether it is exercised in practice, the way an operating-effectiveness audit tests whether controls actually run, not just whether they exist. A strong design with a failing coherence read means governance was drawn up and is not being exercised. Meridian is strong on both.
The dimensional profile
Where FinCo concentrated risk in contract infrastructure and liability architecture, Meridian's deployer closed exactly those gaps before go-live.
| Dimension | What the assessment found | Posture |
|---|---|---|
| D1 Autonomy | Level 3 execution with documented scope boundaries, defined commitment limits, and a scope-restriction protocol that engages a human path outside the boundary. | Strong |
| D2 Data | Claims and identity data with documented provenance and a recorded legal basis for each category; training-data lineage held. | Strong |
| D3 Contract | AI-specific provisions across the material agreements; audit rights over the model provider; customer documentation that represents how decisions are made. | Strong |
| D4 Liability | AE3 recognised explicitly, allocated in contract, and matched by insurance that addresses the actual exposure rather than excluding it. | Strong |
| D5 Leverage | The claims line depends on the engine, and that dependency is governed: scope-restriction protocols and a fallback path are defined and tested. | Strong |
| D6 Stability | Retraining is a governance event; reassessment triggers are defined; the two-layer evidence model is operative and the record is traceable across the window. | Strong |
Same autonomy. Same machine volume. The difference is governance, and the result reflects it.
The evidence behind the result
Certified is not a design opinion. It requires that the governance can be demonstrated from contemporaneous records, not asserted after the fact. Under EIS-01 Meridian's evidence is Tier 1 across the lifecycle: infrastructure-generated, contemporaneous, and anchored for integrity by infrastructure independent of the system that produced the record.
The runtime log carries the governance package identifier at each consequential decision, the package registry is maintained, and the path from a log entry to the governance version active at execution is queryable. That traversal is what makes the record admissible rather than merely present. The DAIOS example sets out this criterion in full.
Why this passes where FinCo did not
The two systems are the same archetype: high-volume autonomous financial decisioning that commits an organisation's funds. The classification diverges on the governance the deployer built, not on the technology.
| What decided it | FinCo Credit Engine | Meridian Claims Engine |
|---|---|---|
| Contract (D3) | No AI-specific provisions; no audit rights | Allocated across the chain; audit rights held |
| Liability (D4) | AE3 unrecognised and uninsured | AE3 recognised, allocated and insured |
| Evidence | Deployment and outcome absent; reconstruction risk high | Tier 1 across the lifecycle; traceable |
| Stability (D6) | Retraining not a governance event; 22 months unreviewed | Retraining governed; reassessment triggers defined |
| Result | GBI 3.42, Not Compliant | GBI 1.6, Certified |
What the result tells each audience
| Audience | What the one result tells them |
|---|---|
| Board | An independently evidenced basis to authorise continued operation and scale, and a record of what was governed and how, against the duty to examine reliance rather than assume it. |
| Insurer | Level 3 autonomy with the AE3 exposure recognised, allocated and insured, and Tier 1 evidence behind it. This supports an underwriter's own process toward standard review, subject to that process. It does not determine cover, terms or price. |
| Investor | Governance that an acquirer inherits intact, with a portable classification and a contemporaneous record, rather than a remediation liability discovered in diligence. |
| Procurement | A system-level classification that the vendor is suitable for a regulated buyer's reliance process, with the evidence to support reliance rather than a name on a trust page. |
The point
FinCo shows that DSI 001 can read a live autonomous system and say why governance fails. Meridian shows the other half: that the scale discriminates, that the top classification is reached by building contract, liability, evidence and stability, not by asserting them, and that the standard recognises governance done well as readily as it identifies governance that is not yet there.