Skip to main content

Which layer is stalling AI at your institution?

AI programmes in financial services rarely fail on the model. They fail on data nobody could lawfully use, a validation gate nobody designed, or an operating model nobody built. Sixteen questions, four layers, and a result you can take into a model risk conversation.

Length
About 5 minutes
Sign-up
None — nothing is sent
Context
RBI · DPDP · model risk

What the result actually means

The scoring is fixed and stated up front, because a score you cannot recompute is a sales toy.

01

Four layers, sixteen questions

Data maturity, infrastructure readiness, governance design and organisational readiness — the four a roadmap depends on before a single use case is chosen. The questions assume a regulated lender, not a generic enterprise.

02

Your weakest layer sets the result

The overall band is capped one step above your lowest layer. A programme is gated by its weakest dependency, not its average, so strong infrastructure cannot offset data you may not lawfully use.

03

Every gap links to the work that closes it

Each layer is scored on its own and points at the service that addresses it, ordered weakest first — what gates the rest, not what is easiest to start.

Answer honestly — the result is only useful if you do

Every option is a state somebody could recognise as theirs. There is no right answer to pick and nobody is watching you pick it.

  1. DATA
  2. INFRASTRUCTURE
  3. GOVERNANCE
  4. ORGANISATIONAL
STEP 1 OF 4

Data maturity

What sits in core banking, the warehouse and the departmental systems around them — how clean it is, and whether DPDP lets you use it for the purpose you intend.

Where does the data a first model would need actually sit today?
If you pulled the customer and transaction data today, what state would it be in?
Under the DPDP Act, can that data lawfully be used for the purpose you have in mind?
How much usable history exists for the outcome you want to predict — default, fraud, churn, whichever it is?
0 of 4 answered

You can move on with questions unanswered — they score zero, and the result will say so.