The Promise of AI Productivity

Executives across the performing arts sector are asking the same question right now: how will AI make us more efficient? The expectation is reasonable. Free up staff time. Redirect capacity. Do more without adding headcount. For those of us in technology leadership, the question isn’t whether to do this. It’s how to do it in a way that delivers on the promise.

We’re piloting an AI assistant project with my team at the Boston Symphony Orchestra, so I’m engaging with these questions daily. The pilot is in early days — around a dozen users to date, with plans to enroll twice as many in the coming weeks. As we add people, I’m thinking a lot about boundaries. Too tight and there’s less room for creative experimentation. Too broad and you open up risks.

Two articles I read recently are sharpening my thinking on those boundaries.

What research in HBR found

Aruna Ranganathan and Xingqi Maggie Ye at UC Berkeley’s Haas School of Business spent eight months embedded in a U.S. technology company of about 200 employees. The company offered enterprise AI tool subscriptions but didn’t mandate their use. The researchers observed in person twice a week, tracked internal communications, and conducted more than 40 in-depth interviews across engineering, product, design, research, and operations. They published their findings in Harvard Business Review in February.

What they found was not a clear story of freeing up time.

Workers didn’t use the reclaimed time to slow down or focus on higher-value work. They expanded. Product managers started writing code. Researchers took on engineering tasks. People attempted work they would have previously outsourced, deferred, or avoided. The researchers describe these as “experimental interactions” that “accumulated into meaningful job scope expansion.” One engineer put it simply: “You had thought that maybe, oh, because you could be more productive with AI, then you save some time, you can work less. But then really, you don’t work less. You just work the same amount or even more.”

The boundaries between work and rest eroded too. Because AI reduced the friction of starting a task, employees began prompting during lunch, during meetings, during idle moments. Work became what the researchers called “less bounded and more ambient — something that could always be advanced a little further.”

The result was a self-reinforcing cycle: faster task completion raised speed expectations, which increased AI reliance, which broadened scope, which expanded the workload. An intensification rather than a reduction.

A critical caveat: this was a tech company, so one might assume a culture that rewards taking on more, moving fast, and breaking things. A performing arts nonprofit is a different environment. Product managers writing code is not a risk at most arts organizations.

But the underlying mechanisms are more universal. The way AI reduces friction until work becomes ambient. The way speed expectations normalize without anyone explicitly raising them. The quiet expansion of what feels like “my job” when the tools make more things possible.

What this means for nonprofits

Rasheeda Childress at the Chronicle of Philanthropy applied these findings to nonprofits. Burnout is already a persistent structural problem in our sector, and the temptation to use AI as a way to add capacity without adding staff is real. Nathan Chappell, founder of fundraising.AI, called the productivity paradox “one of the things I worry about the most with AI in our sector.” His answer: rather than letting the gains disappear into more tasks, “safeguard that 20 percent time to do things that are going to be really additive. And that means things that are inherently human.”

What this means for the BSO

Chappell or Childress nailed it. That idea of safeguarding time is key. Or the way we framed it: let the machines do the machine work to give humans time to think and work like humans.

The approach we’re taking is workstream by workstream: where are skilled people spending time on mechanical tasks like matching, verifying, and cross-referencing? Would a machine be better at that specific work without losing anything in the handoff?

Where the answer is yes, reclaimed time should do one of three things: accelerate work already planned, open up space to think strategically and creatively about what’s next (including rapid experiments), or give an opportunity to draw a line and be done. What it shouldn’t do is open a cascade of new tasks with creeping scope and no clear goal.

What I’m watching for

Our pilot is weeks old, so we’re still proceeding carefully through this theory of change. We have a solid concept of boundaries, but I continue to look for research like this or peer case studies to help fine-tune.

Whether the boundaries hold when more people have access, whether the reclaimed time goes where we intended, and whether the structure leaves enough room for the unexpected wins. Those remain open questions.

Ranganathan and Ye put it well: “Without intention, AI makes it easier to do more — but harder to stop.” The promise of AI productivity is real, but it requires intention and planning.