Last year I gave two presentations, one at the DLM Forum Triennial and one at the IRMS conference where I developed a fictional case study of an organization that decided to apply machine learning and analytics to emails.
In my case study, a public sector organization:
is concerned about the low capture of emails in its recording system (SharePoint) and launches a machine learning application program to fix the deficit;
uses machine learning to apply its existing policy of moving important e-mails to a recording system.
attempts to apply machine learning capability to all e-mail accounts across the enterprise.
In reality, I think that the attitude of public sector organizations to applying analytics and machine learning to e-mail is very different from the organization's attitude in my case study. I assume that public sector organizations in the UK:
is reluctant to apply machine learning to e-mail accounts because of the associated risks.
is also concerned about the prospect that the use of machine learning could lead to very large amounts of emails being recorded in their recording system, such as the existing under recording of emails as records;
would use machine learning (or an analysis function) to search for specific types of correspondence that are relevant to the organization in particular accounts rather than applying machine learning / analysis to all accounts across the enterprise.
Do not move e-mails that are considered important or valuable to the corporate records system, but leave them in e-mail accounts and either reset them to prevent deletion or move them to an e-mail archive ,
Here's the video of my monologue explaining how the organization that applies machine learning to all of their email accounts came about: