Findability Index
The IA Readiness ScoreiA single, scientifically-calculated
score that tells leadership how ready their data infrastructure is for AI — and
gives engineers a precise to-do list to close the gap.
Sources connected3databases & file stores
Objects assessed10tables, views & files classified
Records catalogued654,344counted, never read
Columns profiled227types, keys, sensitivity
Sensitive columns11detected & typed
Below threshold2need human confirmation
Bronze
5
avg Findability Index 0/100
Silver
3
1 below threshold · avg 0/100
Gold
2
1 partial metadata · avg 0/100
Objects & their grade · expand any row for the evidence
Findability IndexMonthly saving
—
nothing selected yet
One-off to build
—
no promotions selected
Findability Index
0
from 0 today
ceiling
—
Every figure is derived from the assessment — nothing to enter.
Promote a layer up
pick a target tier on any cardDormant objects — not assessed
nothing selectedRecommendation runs
Each run maps one business process to the data that supports it
| Run | Status | Step | Updated | Actions |
|---|
CoreWH.dbo.dim_customer
Silver0.94table · 41 columns · 139,822 rows · nightly
Why this layer
–
Columns
| Column | Type | Key | Nullable | Sensitivity |
|---|
Lineage
Upstream
Downstream
Classification history
| When | Change | Method | Confidence |
|---|
–
Where your data comes from
read-only · metadata only| System | Platform | Status |
|---|
– connected · credentials held as vault references
Sync status
idle
last run succeeded
Objects synced
10
active objects cached
Last synced
12 min
ago
Errors
0
none recorded
Connected sources
| Database | Server | Auth | Objects | State | Last sync |
|---|
Cached projection of every active object · synced 12 min ago
Objects cached
10
active · retired excluded
Columns
–
types, keys, sensitivity
Sensitive columns
–
detected & typed
Below threshold
–
confidence < 0.75
Traced edges
–
– direct · – inferred
Contradictions
–
flow downward
Filters
–Objects
click any row for columns, lineage & classification history| Object | Type | Layer | Confidence | Rows | Freshness | Sensitive |
|---|
Traced edges
–| From | To | Evidence | Confidence | Flow |
|---|
–
Confirmed
decisions are appended, never overwritten| Object | Outcome | Layer | By |
|---|
run 8f3c1adraft
Describe the business process
grounded answers only · every claim cites a catalog objectThe assistant runs a tool-use loop over the catalog. It can search, read column and lineage detail, check a classification, and list sensitive columns. It is instructed never to invent an object name or claim it has seen business data.
BPC Co-pilot
Describe the process you need data for, in your own words. I will search the catalog and propose candidate objects, citing each one.
Candidate objects
hybrid search · vector + lexicalRanked by hybrid retrieval over ingested schema text, then annotated with each object's assessed layer and confidence. Untick anything that does not belong. The choice is recorded in the audit trail.
| Object | Assessed layer | Confidence | Why it matched |
|---|
Lineage confidence
–%
edges flowing upward
Key integrity
–
orphan references
Freshness
–
from job schedules
Layer mapping
validated server-sideGovernance scan
blocks approval until acknowledgedSensitive columns detected
| Object | Column | Detected type |
|---|
Not assessed
Ownership no data source in the model. Reported as null.
Retention no data source in the model. Reported as null.
Cost estimate reported as 0 rather than modelled.
These are returned empty by design. A fabricated owner is worse than an admitted gap.
Approval is blocked until the governance scan is acknowledged. The server enforces this. It is not a UI-only check.
Audit trail
chain verified| # | Event | Actor | Hash |
|---|
Ask Metadata
all databases4 toolsBPC Co-pilot
I have read the schema and job metadata of every connected database and assessed each object into a medallion layer. Ask why an object landed where it did, what feeds it, or which columns are sensitive. I cite the objects I draw from.
I only ever see metadata names, types, keys, SQL definitions and job steps. Never the business data itself.
I only ever see metadata names, types, keys, SQL definitions and job steps. Never the business data itself.
Sessions
Each session keeps its own history and scope, so a line of questioning can be resumed.
Guardrails
- ✓Four typed, read-only tools: search, table detail, classification, sensitive-column report
- ✓Never reads business data. Schema and job metadata only
- ✓Instructed never to invent an object name; unknowns are declined
- ✓Answers stream as they are produced, with citations attached
Profile
Email is the login identity and cannot be changed here.
Roles are assigned in User Management by an administrator, never self-selected.
Change password
API keys
| Name | Prefix | Created | Last used | Action |
|---|
Roles and per-domain access · enforced at the API, not in this interface
| User | Role | Domain access | Status | Actions |
|---|
Roles the platform issues
| Role | Seats | Scope | Reach |
|---|