Automation and Autonomy
Different AI accounting assistants are capable of different levels of autonomous activity and may be described on a scale of 0-5.
As a business function, accounting already benefits from much low-level automation, such as the generation of reports and integration to other line-of-business systems.
Here we distinguish autonomy from automation in that autonomy implies an element of judgement, discretion or decision making which has up to this point been the unique responsibility of a human operator.
Autonomous accounting activity does not imply action taking outside of established parameters or policies.
Taxonomy of levels of autonomous accounting
Level 0: No accounting autonomy.
A human being is entirely in charge or all accounting actions and decisions. This does not preclude the use of accounting tools such as spreadsheets, and data processing and recording software such as traditional ERP systems.
Level 1: Assisted accounting autonomy.
Well defined tasks requiring an element of judgement such as purchase invoice coding or debt-chasing communications, may be controlled by the assistant. The work of the assistant is subject to supervision and approval by a human.
Level 2: Partial accounting autonomy.
The AI accounting assistant is able to handle tasks within parameters, but is required to be supervised while it is in operation.
Level 3: Conditional accounting autonomy.
The AI accounting assistant can manage tasks under certain conditions, but a human supervisor is expected to take over when requested to do so by the assistant.
Level 4: High accounting autonomy.
The assistant is fully-autonomous requiring no human intervention while it is operating. However in the event of an error condition, the system is designed to to shutdown and can only be restarted after human intervention.
Level 5: Full accounting autonomy.
As the highest class of accounting autonomy, level 5 means the AI accounting assistant can perform accounting tasks in all conditions without any human intervention, including pre-emptive avoidance of, or recovery from, error conditions. The only human involvement is to set the parameters or policies within which the assistant operates.
Scope
Accounting tasks can be broken down to quite a granular specialist level, depending on the needs of the organization. An AI accounting Assistant can be similarly specialized. Our overarching approach to developing AI accounting assistants is to initially develop specialized assistants, expert in their own domain, and then incrementally assemble them into larger co-operating units.
Current Status
Bmbix is developing AI based accounting assistant technology. Uniquely, Bmbix AI accounting assistants make significant use of the Bmbix structured messaging architecture to co-operate with other Bmbix AI accounting assistants across organization boundaries.
Acknowledgements
The above structuring of levels of autonomy was inspired by the work of SAE International in respect of autonomous vehicle control: “Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles J3016_202104” revised 2021-04-30.
