Most organisations have AI somewhere in their stack. Budgets are growing. Boards are asking questions. And yet the productivity gains remain stubbornly out of reach for the majority.
McKinsey’s 2025 State of AI report puts the numbers in context. 88% of organisations now use AI in at least one business function. But only 23% are actively scaling it across their enterprise. The rest are stuck in pilot mode.
Victoria Chin, Senior Director of Product Strategy for AI at Asana, has watched the pattern play out since the early days of the LLM boom.
“There was this big promise, all this hope to dramatically change things,” Chin told UC Today. “What we’ve seen is that many AI tools have been really powerful for individual use cases. What we haven’t seen is truly scaling AI across multiple teams or entire organisations.”
The Coordination Tax Nobody Is Solving
Asana’s Anatomy of Work research surveyed over 10,000 knowledge workers and found that 60% of working time goes on what Asana calls work about work. Status updates, chasing approvals, following up on tasks that should already be moving.
“When you think of the most strategic work that happens in an organisation, it’s typically the work that requires an entire team or multiple teams to come together to execute on something that’s actually going to move the needle,” Chin said.
Copilots and personal assistants have not touched that problem. The overhead still belongs to humans.
Why Most AI Tools Are Built for the Wrong Unit
Most AI tools serve one user at a time. They do not see the team, understand the workflow, or retain context across multiple people and multiple projects.
Asana’s own research studied 3,182 knowledge workers and 560 IT professionals. 67% of organisations have not scaled AI beyond isolated experiments. Only 29% say they are beyond the pilot phase.
The organisations that do scale successfully treat AI as infrastructure rather than a collection of separate tools. They measure adoption and productivity, not just cost savings. That is the problem Asana AI Teammates are built to solve.
Built for the Whole Team
Where most agents answer to one person, AI Teammates answer to everyone on a project.
“Anyone on an entire team or multiple teams can interact with them. You can redirect them. You can give them feedback right in the flow of work where your team is already working,” Chin said.
Context makes that useful. Asana has built its work data model over 15 years, and AI Teammates inherit that foundation.
“They know your goals, they understand your timelines and dependencies, they are not starting from scratch every time,” Chin said. “They have shared memory where an entire team can benefit, not just the single person who prompted them.”
From Hallucinations to Reliable Results
Trust has been hard to build. For years, hallucinations were the primary concern customers raised. Chin traces that problem back to a lack of context.
“Asana provides context on who is doing what, by when, how, and why within your organisation. It’s that layer that gives LLMs more predictable, reliable and accurate results,” she said.
The memory also stays current. “If something changes, you can remove it too,” Chin added.
Where It Is Already Working
Asana ran a beta programme with over 200 customers before going live. The results span industries and company sizes, not just early-adopter tech firms.
Morningstar, the publicly traded investment research firm, deployed multiple AI Teammates across IT and research functions. They cut their ticket intake process by two weeks and saved approximately 15,000 person-hours in a single year. KW Automotive, a German vehicle suspension manufacturer, uses an analyst teammate to correlate data across multiple projects simultaneously, saving several hours per report. Human IT, a nonprofit refurbishing devices for low-income households, uses a teammate to catch data errors before they affect other teams.
For more on where AI is delivering real returns right now, see our guide to the best AI productivity use cases in 2026.
Governance That IT Leaders Can Work With
Control is the question IT leaders ask first. Asana’s answer is built on the role-based access controls already in the platform.
“Our AI Teammates respect the controls we have already been building for years,” Chin said. “Admins have central control over who can create teammates within their organisation.” Permissions cover who can view, use, or manage a teammate’s memory at both the organisation and individual user level.
CIOs and senior IT leaders were involved as design partners before the product reached beta.
The Road from Pilot to Scale
AI Scalers are 43% more likely to report revenue growth than organisations stuck in experiment mode. The gap between the two groups is widening.
“There aren’t many tools that actually make entire teams or organisations more effective,” Chin said. “That is what we are here to do.”
The organisations that close that gap first will not do it by giving individuals better tools. They will do it by making AI work for the whole team.
Watch the full UC Today interview with Victoria Chin, Senior Director of Product Strategy for AI at Asana, here.
