Findings Worksheet: Atlas Legal Group

Use this as input to the Module 4 scoring exercise. The findings below are what a competent assessor would have written after Session 1, the artifact review, and the SDC Agents SMB introspection runs on the case described in case_atlas_legal.html.

Dimension: Schema Integrity

Working score: 2 Evidence: - Clio Manage enforces a fixed schema for matters, contacts, and time entries (Level 3 evidence on this system in isolation) - Access database clients table has typed columns but no documentation, no version control, and a case_number format change in 2019 that was never migrated (format_drift flag confirms two coexisting formats) - Three shared Excel workbooks have header rows that drift between sheets; one has 6 columns, the same workbook on a different tab has 8 - The same logical entity (client) is modeled with different fields in Clio, QuickBooks, and the Access DB - No data dictionary exists for any system other than Clio's vendor docs

Confidence: High Notes: The Clio side could be argued as Level 3, but the assessment is for the firm's data estate, not for Clio. The estate-level score is dragged down by the unmanaged components.

Dimension: Constraint Enforcement

Working score: 2 Evidence: - Clio enforces required fields and some pick lists - Access DB has NOT NULL on client_id and intake_date only; no CHECK constraints; intake_date contains 3 rows of 00/00/0000 which means the NOT NULL is meaningless in practice - Excel workbooks have no enforcement at all; conflicts log has free-text in fields that should be categorical - Cross-system constraints (e.g., a client in Clio must exist in QuickBooks before billing) are enforced manually by the bookkeeper

Confidence: High Notes: The bookkeeper is the constraint engine. When she retires, this score drops to 1 unless the constraints are externalized.

Dimension: Semantic Identity

Working score: 1 Evidence: - No shared identifier across Clio, QuickBooks, and the Access database - Quote from Dana: "A client's name is spelled three different ways across Clio, QuickBooks, and the Access database" - Introspection near_duplicate_identifier flag fired on Access client_name column with 47 instances of the same client appearing under 3 spellings - Matter codes in Access are paralegal-assigned and human-readable, not stable - No CUID2, UUID, or URI-formed identifiers anywhere in the estate

Confidence: High Notes: This is the floor that caps every derived dimension. This is also the easiest score for Dana to understand because she lived the pain in the discovery call.

Dimension: Provenance

Working score: 2 (capped to 1 by floor) Evidence: - Clio captures created_by and updated_by on matter notes (good) - Access DB has no audit log - Excel workbooks have no audit; SharePoint version history exists but the firm has never used it for an investigation - The state bar audit took 4 days to compile communications because lineage across systems is manual - No cross-system provenance graph

Confidence: High Notes: In isolation Provenance scores 2, but the floor constraint caps the effective score at 1 because you cannot trace what you cannot identify. Report both: working score 2, effective score 1.

Dimension: Interoperability

Working score: 2 (capped to 1 by floor) Evidence: - Failed Clio↔QuickBooks integration in 2022; abandoned because client records did not match (this is a Semantic Identity failure presenting as an Interoperability failure) - No standard format (JSON-LD, FHIR, OData) used for any exchange - Bar association exports are manual PDF compilations - Calendly→Outlook is the only working automated sync

Confidence: High Notes: The 2022 integration story is the canonical example of how the floor constraint manifests in real money lost. Use it in the Session 2 walkthrough.

Dimension: Governance

Working score: 2 (capped to 1 by floor) Evidence: - No data steward; no documented data quality policy - State bar audit finding is an external pressure forcing reactive change - Partners share informal accountability ("we'll talk about it at the partner meeting") - Patricia (bookkeeper) is the de facto governance layer for financial data; her retirement is a single point of failure - No metrics, no SLAs, no review cadence

Confidence: High Notes: Dana will want to start with Governance because the state bar finding is on her desk. Module 4 covers how to redirect this without losing the engagement.

Summary table

Floor: 1 (Semantic Identity) Effective ceiling on derived dimensions: 1 Single highest-leverage intervention: Establish stable cross-system identifiers via SDC Agents SMB catalog + SDCStudio component minting. This unblocks Provenance, Interoperability, and Governance simultaneously.