Content-Based Smart E-Mail Dispatcher Using Large Language Models

K. Paramesha, K R Sriram, Sujan Shetty, Shamanth Kishore and R.Tejaswini submitted a paper titled "Content-Based Smart E-Mail Dispatcher Using Large Language Models" to arXiv on 25 Jun 2026 proposing an automated system that reads email text and routes messages to the appropriate WhatsApp groups of students across semesters.

How does the dispatcher work?

The paper describes an agent framework that takes email content as input, queries large language models to analyze the text, and uses the LLM output to decide which WhatsApp groups of students across semesters should receive the message. The system feeds the raw email content into a structured agent prompt containing instructions and context, the agents query LLMs for classification and routing decisions, and the dispatcher forwards messages and attachments to the relevant WhatsApp groups.

The authors position the agent framework and LLMs at the center of decision-making. The paper states the system "harnesses the capabilities of LLMs in analysing the textual contents for decision-making." It emphasizes that the prompt includes email content plus instructions and context so the agents can map messages to the correct semester or program groups.

What exactly did the authors build and test?

The submitted manuscript outlines a targeted dispatcher for an engineering college environment that routes emails to WhatsApp groups of students from various semesters, and it does so without using labeled datasets. The authors present the design as an automation layer that replaces manual perusal and cross-posting of email contents and attachments to instant messaging channels.

The abstract frames the problem as a productivity and error issue in large organisations, noting manual forwarding is "error-prone and time-consuming" and creates "undue stress." The proposed system aims to reduce that cognitive load and improve the flow of information by automating content-based routing to student WhatsApp groups.

Why it matters

The approach matters because it replaces repetitive human triage of administrative emails with an LLM-driven agent workflow, potentially cutting time spent reading and forwarding messages and reducing routing mistakes. The paper highlights a practical use case: dispatching email announcements and attachments to semester-specific WhatsApp groups in an engineering college, which targets a clear operational burden where small improvements in routing efficiency can scale across many students and staff.

The authors also call out a methodological point: the system does not rely on labeled datasets, which simplifies deployment in settings that lack annotated training data and accelerates adaptation to new group structures or message types.

What the paper documents (concrete details)

The submission appears on arXiv as arXiv:2606.26593 with the version recorded on 25 Jun 2026. The author list is K. Paramesha, K R Sriram, Sujan Shetty, Shamanth Kishore and R.Tejaswini. The abstract and submission metadata indicate a PDF was uploaded; the submission history records the file size as (805 KB).

What to watch

Look for a follow-up that provides evaluation metrics or deployment results showing how the system performs on real mailstreams, and whether the authors publish code or runbooks for the agent prompts and routing rules. Concrete signals will be an experimental section with accuracy or routing-error rates, or a public demonstration of the dispatcher integrated with WhatsApp groups.