It can be fine to keep some of the core software that works in COBOL, and write other software in different languages.
This is probably not the reason it's hard to estimate the recoverable gap. Instead, I assume the IRS has a lot of returns they would like to audit, but they can't know how much unpaid tax they will find in un-audited retu…
It can be fine to keep some of the core software that works in COBOL, and write other software in different languages.
This is probably not the reason it's hard to estimate the recoverable gap. Instead, I assume the IRS has a lot of returns they would like to audit, but they can't know how much unpaid tax they will find in un-audited returns until they actually spend the time and headcount to audit them.
What the IRS could do, and would be useful, is: "out of $ X spent on auditing today, how much $ Y unpaid tax was found and $ Z tax recovered". To show how much hopefully (Y > X) or (Z > X).
I do also suspect that AI could substantially improve auditing and recovery of unpaid tax, if the IRS is allowed to investigate this.
It can be fine to keep some of the core software that works in COBOL, and write other software in different languages.
This is probably not the reason it's hard to estimate the recoverable gap. Instead, I assume the IRS has a lot of returns they would like to audit, but they can't know how much unpaid tax they will find in un-audited returns until they actually spend the time and headcount to audit them.
What the IRS could do, and would be useful, is: "out of $ X spent on auditing today, how much $ Y unpaid tax was found and $ Z tax recovered". To show how much hopefully (Y > X) or (Z > X).
I do also suspect that AI could substantially improve auditing and recovery of unpaid tax, if the IRS is allowed to investigate this.
My experience with COBOL, having written it, is that is very solid if there are minimal system changes. COBOL-based systems can't turn on a dime.