How do you train an AI service for collaboration?
Organizations looking to improve communication efficiency can train AI to automate collaboration workflows, launch meetings and transcribe conversations.
The idea of training AI is central to its value. An AI service needs to engage with employees and customers in a personable way. In essence, AI enables machines -- such as computers and software -- to handle increasingly complex functions, so workflows can be automated and communications can be more effective.
By their nature, AI applications are not off-the-shelf. They learn and get better over time using machine learning, a core technology that falls under the AI umbrella. To train AI, you need to think of an AI service as a blank slate, where machine learning models build on data sets compiled from ongoing usage.
For example, if you say, "Start my meeting," an AI-driven collaboration application would dial you into a conferencing bridge. Over time, the AI service would learn how to handle other tasks, such as dialing in other team members for a scheduled weekly meeting. As the AI service learns each person's style of speech and tone, it could determine the name of each speaker. From there, it could be trained to transcribe conversations, both in terms of speech accuracy and attributing comments to the right speaker.
The key to training an AI service is to understand that AI is an iterative set of technologies, and its value increases as it learns. You must be prepared to start small and build success with simple tasks.
AI may never be good enough to replace all forms of person-to-person communication. But employees will trust it more as accuracy improves, and that's the starting point for managing more complex tasks.
Don't expect a fast ROI with an AI service. But, if employees increasingly engage with AI, then machine learning could perform some additional fine-tuning -- and it won't take long to realize the benefits.
Do you have a question for Jon Arnold or any other experts? Ask your enterprise-specific questions today! (All questions are treated anonymously.)