Corporate Communication Companion (CCC): An LLM-empowered Writing Assistant for Workplace Social Media

TL;DR

This paper tackles the challenge of enabling individualized yet professional writing for workplace social media by presenting Corporate Communication Companion (CCC), an LLM powered writing assistant. CCC decomposes post creation into Co-Outline and Co-Edit steps, enriching prompts with user job status and history to produce tailored outlines and edits with adjustable tone attributes. Through offline data analysis and a two phase user study, CCC improves writers' experience and audience perceived quality without increasing cognitive load, while revealing diverse personalization strategies among users. The work offers design principles for context aware, mixed-initiative LLM writing assistants in corporate communication and highlights avenues for future multi modality and retrieval based improvements.

Abstract

Workplace social media platforms enable employees to cultivate their professional image and connect with colleagues in a semi-formal environment. While semi-formal corporate communication poses a unique set of challenges, large language models (LLMs) have shown great promise in helping users draft and edit their social media posts. However, LLMs may fail to capture individualized tones and voices in such workplace use cases, as they often generate text using a "one-size-fits-all" approach that can be perceived as generic and bland. In this paper, we present Corporate Communication Companion (CCC), an LLM-empowered interactive system that helps people compose customized and individualized workplace social media posts. Using need-finding interviews to motivate our system design, CCC decomposes the writing process into two core functions, outline and edit: First, it suggests post outlines based on users' job status and previous posts, and next provides edits with attributions that users can contextually customize. We conducted a within-subjects user study asking participants both to write posts and evaluate posts written by others. The results show that CCC enhances users' writing experience, and audience members rate CCC-enhanced posts as higher quality than posts written using a non-customized writing assistant. We conclude by discussing the implications of LLM-empowered corporate communication.

Figures (7)

• Figure 1: A visualization of users' post writing varying along Use Cases and Writing Style. Each column represents posts related to a specific topic, and the width of each column corresponds to the proportion of that topic among all posts. For example, the first column represents posts about announcements, which is the most frequently written topic among all posts. The numbers within each cell indicate the percentage of a specific writing style within a particular topic. For instance, the cell in the upper left corner shows that $52.85\%$ of the announcement posts have a formal tone. Note that the clustered styles are not mutually exclusive; for example, an announcement post can be both enthusiastic and use slang in writing. Hence, the sum of percentages within each column does not necessarily add up to $100\%$.
• Figure 2: An illustration of the CCC system. The LLM-empowered system (1) matches the user input with its relevant historical posts that share the same use case as the input (i.e., professional or casual post). (2) establish user profiles based on job status and relevant historical posts (3) help users Co-Outline and Co-Edit based on user input with user profile and customization taken into account. The examples show (1) the LLM-summarized job status and writing styles of participants from their historical data (2) and customized attributions set by participants. This information was utilized by CCC in prompting for enhancing the system's writing assistance.
• Figure 3: An illustration of the interface. The example shows Alice writing an introduction to a work about some security product that her team is working on. She decides to make the tone of Co-Edit to be more (tailored, conversational, creative, detailed).
• Figure 4: Participants' self-reported user experience (\ref{['fig:all_tasks:experience']}), perceived task load (\ref{['fig:all_tasks:load']}), and trust (\ref{['fig:all_tasks:trust']}) in using CCC and standard writing assistance. Results show that people felt the writing experience with CCC is more engaging and collaborative, and the posts written with CCC are more unique and complete. As a result, people perceived that CCC is more competent, and they are more willing to use CCC. Such enhancement did not cause extra cognitive loads. Error bars represent the standard errors of the mean.
• Figure 5: Participants' evaluation of posts composed by other participants using CCC and standard writing assistance. People felt the posts written by others under the help of CCC were more informative, engaging, appropriate, and thus overall in a better quality. Error bars represent the standard errors of the mean.