The Asana app already has workflow automation tools, but the addition of LLMs allows for more powerful automations. “What’s really exciting about embedding LLMs into work management is that there’s judgement, there’s nuance,” said Costello. “Instead of it being a black and white rule — ‘if these things are missing, ask this’ — it’s directional, like, ‘here’s what we typically look for; use your judgment.’”
That said, given the unreliability of LLMs, there’s potential for the teammates to introduce errors into the work management process.
Costello said it’s important to have a “human in the loop,” when the AI assistant takes action. “We want to make sure that humans are the ultimate deciders,” she said. “They’re the ones that are accountable, they’re the ones completing the work: AI is just helping. AI shouldn’t be moving work forward in a way that is not permissioned effectively by the team or the person driving that work.”