Re: My paper on applying agentic AI to business workflows

Your tiered architecture addresses semantic orchestration within an
enterprise boundary.

Have you considered how reasoning traces, delegation intent, and service
invocation semantics would operate across institutional boundaries where
agent identity and authority need to be verifiable?


Dave Raggett <dsr@w3.org> 於 2026年2月24日週二 下午8:20寫道：

>
> You may be interested in a preprint of my perspective paper on applying
> agentic AI to business workflows:
>
> [image: cogai.png]
>
> cogai/papers/AI4OfficeWork.pdf at master · w3c/cogai
> <https://github.com/w3c/cogai/blob/master/papers/AI4OfficeWork.pdf>
> github.com
> <https://github.com/w3c/cogai/blob/master/papers/AI4OfficeWork.pdf>
> <https://github.com/w3c/cogai/blob/master/papers/AI4OfficeWork.pdf>
>
> The approach exploits the strength of LLMs for reasoning with the open
> texture of language in combination with symbolic approaches for efficient
> deterministic processing. It shows how Agentic AI can be used for semantic
> orchestration across a swarm of specialised agents, along with delegating
> difficult or high risk cases to human staff, who can enrich the business
> rules and precedents used by the agents.  The approach uses audit trails as
> a basis for human oversight.  Agents are required to cite which rules were
> used along with supporting quoted extracts. These lower the risk of
> hallucinations, and can be checked automatically.
>
> Dave Raggett <dsr@w3.org>
>
>
>
>