Accend's AI spreads financial statements and experts verify them. Banks are asked to trust the experts. This shows an eval layer at each step of the workflow, and the scorecard that turns trust into a track record.
Built for Accend · the short version
Banks are asked to trust Accend's experts. A track record would let them check.
In this sample, line-item mapping and add-backs cause most bank corrections, and each has a plain rule that would catch it.
The data is made up. To make it real, I'd need spreads with reviewer edits and client correction tickets, step by step.
Who built it
Anirudh Devella, data scientist. Before this, in production:
“I’d hire him again.” Alex Dixon, Co-founder, Bluelight (YC), managed Anirudh directly. Read it on LinkedIn
Each step of spreading a financial statement gets its own check, built from two things: context (the bank's chart of accounts, prior periods, the borrower's industry) and plain deterministic rules. Click a step to see what its check runs and an error it catches.
What the eval layer misses, a reviewer should catch. What the reviewer misses, the bank corrects. Counting that per reviewer gives banks a real track record, and gives Accend a way to route hard spreads to its best people.
The memo is what the credit committee reads. Every number in it is checked against the spread, and every section against the bank's format.
Next step