Microsoft Build 2024: Power Platform’s copilots navigate toward more human workloads

Power Automate, the automation suite within Microsoft Power Platform, is gaining new and updated Copilot capabilities that the company says will allow it to learn faster than ever.

Around Microsoft’s annual developer conference, Build, company officials have explained that, in some cases, users will be able to train their AI “just like a person would have been trained.” And for other needs, developers may be able to specify just the building blocks of a workflow and let the system design and execute the automation when called upon to carry out the work.

Across these updates, Microsoft’s messaging emphasizes advances that allow the AI tools to not just take on repetitive, time-consuming tasks, but to assume more responsibilities in an organization's operations. Power Automate will add new flow authoring tools for human users, new capabilities that will allow the platform to assemble processes on its own, self-monitoring capabilities, and new AI training and co-working techniques that aim to make the Copilot interaction patterns seem more human.

“Low-code automation and AI accelerate development, make automation accessible to everyone regardless of their skill level, and help businesses tap into new value potential,” wrote Sangya Singh, VP for Power Platform Automations. “Developers spend less time building and more time innovating.”

Design an automation plan rather than an automation

AI flows are a new way to automate “that lets generative AI reason over unstructured content and determine what steps and sequence are needed to achieve a process outcome based on instructions and parameters provided by a user in natural language,” according to Microsoft. Instead of complex process designs, “users work with AI to create an automation plan.”

A demo video for this capability, complete with energetic music, shows AI being trained how to reject an auto insurance claim. In the scenario, the AI is instructed to check corporate documents and apply rules based on a range of information sources in making its decision.

Train your AI as if it were a co-worker