Which Company's Case Study?
This case study focuses on Slack Technologies, the provider of "Slack," a business communication platform used by many companies. To effectively utilize the vast internal knowledge accumulated daily in conversations and files and to improve employee productivity, Slack developed and implemented "Slack AI," a platform-native generative AI feature.
Challenges They Aimed to Solve
As conversations accumulate in Slack, catching up on unread messages and finding past information can take time. Slack addressed this with conversation summaries and natural-language search as initial Slack AI features.
How AI Was Used
Slack introduced multiple native AI features that securely leverage customer data, such as conversations and files accumulated within the platform, to generate contextually relevant answers. Specifically, these include features for automatically summarizing channel and thread conversations and a search function that finds relevant information simply by asking questions in natural language. These features utilize large language models (LLMs) and employ Retrieval-Augmented Generation (RAG) technology to generate answers based on internal knowledge and suppress hallucinations (plausible but false responses).
Estimated Architecture
The following is an estimated architecture based on publicly available information and common configurations. It does not definitively describe the actual internal structure.
- ▸Confirmed: Slack AI summarizes and searches conversations that the requesting user is authorized to view.
- ▸Confirmed: It uses RAG to retrieve relevant information and pass it as inference context, and customer data is not used to train LLMs.
- ▸Confirmed: LLMs run in Slack-controlled cloud infrastructure and inherit existing access controls and security requirements.
- ▸Estimated: Public material does not establish the exact model selection, evaluation process, or operational monitoring design.
Implementation Effects and Key Takeaways
- ▸Slack reports that 90% of Slack AI adopters reported higher productivity than non-adopters. This is a self-reported result and does not guarantee the same outcome for every user.
- ▸Summaries and search are available inside Slack, reducing the need to switch tools.
- ▸Users should compare generated output with the underlying messages and keep people responsible for important decisions.
What Japanese Companies Can Learn
Many Japanese companies also accumulate vast amounts of information in internal chats and document management systems. Slack's case study offers concrete hints on how to unearth these "buried assets" with generative AI and link them to operational efficiency. The approach of integrating AI into tools that employees use daily to maximize implementation effects will be particularly valuable. Applying AI to specific challenges that frequently occur in certain departments, such as information retrieval or summarizing minutes of regular meetings, can be considered the first step towards success.