Effective internal communication is crucial to the success of any organization, from how its employees collaborate in groups to how its managers train their direct reports. The rise of video conferencing; voice call; and instant messaging web applications such as Slack, Jive and Skype for Business has dramatically increased the amount of communication that occurs online rather than in person. These tools have rendered telecommuting a viable alternative in certain industries; in fact, some companies estimate that 50 percent of their employees will work remotely by 2020.

It may be impossible for technology to become a full substitute for face-to-face interaction with colleagues. Even video conferencing fails to capture and transmit key components of body language. Yet online workplace communication provides an often unsung, increasingly relevant benefit: the ability to collect data about interactions as they take place between employees. Analyzing this data, in turn, can yield keen insight into how social networks are structured in the workplace and how they can be optimized to enhance an organization’s productivity.

What data can you collect? Consider what happens each time an employee writes text and sends it through a web application, whether it’s a direct message to a single coworker or a public post to an entire cohort. The cloud service hosting this application will store in a database the exact time this post occurred; the identity of the person who posted it; where exactly it was posted; the text comprising the message; and any responses the text receives, whether it’s a direct message in a private chat, an indented comment on the post, up-votes and down-votes, shares, or endorsements of the author.

What analytics can you derive from this data? Artificial intelligence (AI) can reveal different aspects of the social learning networks (SLNs) that summarize how enterprise communication occurs. Consider even a simple visual that connects employees based on how often they communicate: The more frequently they speak, the stronger the link that is drawn between them. Comparing this network graph with the organizational structure of the enterprise – who reports to whom, who works with whom and so on – can reveal unintended collaborations that the organization may want to factor into how work groups are defined (or, alternatively, how they may be disbanded). This understanding can be particularly helpful while onboarding new employees; organizations can use connections formed during training to define cohorts from the outset.

The topics of employee discussions are another important dimension of SLNs, one that the textual component of posts can reveal. Natural language processing (NLP) algorithms can be applied to this data to extract the key recurring themes of communication across an organization. Knowledge of these topics, and how they evolve over time, would be useful in their own right; they would indicate materials and/or skills that could be introduced to assimilate employees during training, for example. Combining this information with the connections formed between coworkers would show which topics are discussed in which groups and vice versa.

Further, consider an employee who has a question on a specific work-related topic. They search through help support and colleagues in their personal SLN but are unable to find a thread or person to adequately address their inquiry. An AI system that has discovered the full SLN would be able to recommend a strong candidate to whom this employee could reach out.

What factors of the SLN must be considered here? A “good” person with whom to collaborate on this topic would not only have a job title covering the area but also a history of discussing this topic with others. There is another subtlety here, too: Research suggests that the person chosen should be a demonstrated disseminator of knowledge on this topic, as opposed to a seeker. How can you make this distinction within the SLN? Some propose using question/answer detection algorithms to measure when someone is asking (seeking) or answering (disseminating) a question in any particular post. You can also combine linguistic cues with votes, shares, endorsements and an employee’s inferred skill levels.

These are just a few of the many ways that uncovering SLNs can benefit an organization. The data brought about by online communication allow us to define these networks more concretely than ever before. It will be interesting to see over the next few years how organizations leverage SLN analytics to enhance the workplace.