Discussion about this post

User's avatar
Hisashi Miyaguni's avatar

The "Chief Assurance Officer" item is the one I'd underline. As agents take more of the processing, the scarce role becomes whoever can tell you the machine was wrong.

The field data backs it up. The Stanford GSB / MIT Sloan study published in the Journal of Accounting Research this year found the more experienced the accountant, the more they intervened - specifically when the model signalled low confidence. The value came from knowing when to step in, not from leaving it alone.

We learned the inverse expensively. We rolled OCR across our whole client book with nothing measuring quality client by client, and the failures surfaced months later as corrections that looked like ordinary bookkeeping noise.

Agree with the overall take that firms don't need a transformation roadmap. The one thing worth building deliberately is the assurance layer, because it doesn't arrive with the tool.

No posts

Ready for more?