Where the model that cannot be explained does not ship.
Building an analytics function from zero at a commercial bank, and earlier at JP Morgan Chase across risk and finance.
What is different here
Banking has had model governance for decades, which is an advantage and a trap. The advantage is that nobody has to be convinced that a model needs an owner, a version and a review. The trap is that the existing framework was written for models that do not generate text, and applying it unchanged either blocks everything or waves everything through.
What usually needs fixing
A team is hired, a platform is bought, and the queue of requests turns out to be reporting rather than analysis. Starting from zero is easier than fixing an operating model that already has defenders.
Two functions, two definitions, both defensible, and a leadership meeting that starts with reconciliation rather than a decision.
The useful demand is an explainable decision, and the two are not the same thing.
Where I have done it
1 of the seven case studies is in this sector.
The work this usually involves
Six services, and most engagements touch three of them. Which three depends on where the chain is breaking rather than on what you came in asking for.
Not sure where your money is going? Seven questions, three minutes, no email.
Take the 7D Index Request a 45 minute call