Ask a COO how AI adoption is going and you'll often hear two answers at once. The reports say time is being saved, drafts come back faster, meeting notes write themselves, status updates go out on schedule. The team says they're busier than they've ever been. Both can be true, and the gap between them is where this series ends up.

Part 1, The Accidental Manager, closed on three commitments, judge AI by its benefit to people and customers, prepare people to manage agents, and set a tempo people can sustain. Part 2 covered the data foundation that makes the first one measurable, and Part 3 worked through the second by redesigning one role. This last piece is about the third, and it starts with a simple question, where did the hours go?

Faster and Lighter Aren't the Same Thing

The assumption behind most AI business cases is that time saved becomes time freed. Researchers at UC Berkeley Haas spent eight months inside a technology company of about 200 people to see whether that holds, and found something closer to the opposite. People worked at a faster pace, took on a broader range of tasks, and stretched work into more hours of the day, often without being asked. Work crept into moments that used to be pauses, and many people kept several AI-assisted threads running at once.

What makes the finding useful is how it felt to the people inside it. In the moment, the work felt energizing, with more momentum and a sense of expanded capability. Looking back over weeks, the same people described feeling stretched and less able to switch off. Expectations had quietly reset, so what started as extra effort became the new standard.

Services teams are especially exposed to this pattern, because there's always more client work to absorb whatever time comes free. Unless someone decides otherwise, recovered time tends to go wherever the pressure is highest.

Where the Hours Go Is a Leadership Decision

Recovered time doesn't allocate itself. It can go into better client work, developing people, taking on work the team couldn't support before, or simply giving people room to recover, and each of those is a reasonable choice. The risk is when no one makes the choice and the hours get absorbed by more review, more requests, and a higher baseline of expected output.

The research supports treating this as a design decision rather than a side effect. A study in the International Journal of Information Management found that AI can raise productivity and engagement while also creating technostress that wears on well-being, with the balance depending on whether people feel equipped to use it. The same tool can be a resource or a drain, and the conditions around it decide which.

That's why Part 1 argued for an honest count of recovered time, comparing the human hours a result needed before AI with all the hours it needs now, including briefing, checking, and correcting. Leaders who have that number can decide where the gain goes, and leaders who don't usually find out later, from the team.

The View From the Passenger Seat

Part 1 described AI adoption from two seats in the same car. Leaders in the driver's seat see the road and know why the car is moving fast, and teams in the passenger seat feel every turn without that context. Tempo is what the passenger feels, the pace of new initiatives, shifting priorities, and how much has to be absorbed at once.

A tempo people can hold doesn't mean slowing down. It means sequencing change so teams can finish, learn, and adjust before the next one lands. The Haas researchers point to practical moves that fit services work well:

• Batching non-urgent updates instead of reacting to every AI output as it arrives

• Protecting focus time

• Building short pauses before major decisions so speed doesn't crowd out judgment. 

In a delivery organization, that might look like:

• Rolling out one AI-supported workflow per team at a time

• Protecting the oversight time Part 3 wrote into the role

• Agreeing up front how much recovered capacity goes back to the team rather than straight into new work.

Seeing Where the Time Actually Goes

Most of this depends on visibility. A leader can't decide where recovered hours go without seeing where hours go now, across billable work, non-billable work, and the admin in between. When time, utilization, and project health live in the same Salesforce org as the rest of delivery, that picture comes from live records rather than a quarterly reconstruction.

Cloud Coach customers capture about 15% more billable hours on average, and tracking that kind of shift over time is one of the clearer signals of whether recovered time is turning into client value or getting absorbed somewhere else.

## How We See It

Faster isn't the same as lighter, and AI won't make the difference on its own. The firms that get real net human gain will be the ones that:

• Decide where the recovered time goes

• Give people the authority and preparation to manage the work agents do

• Set a pace their teams can keep while they deliver. 

That's the work of leadership in the agent era, and it's what turns speed into something people can sustain.

Frequently Asked Questions Related to Professional Services

• Does AI reduce workload in professional services firms?

It can, but it doesn't happen automatically. Research suggests AI often speeds work up while widening scope and filling pauses, so time saved only becomes time freed when leaders decide where it goes.

• What does a sustainable pace of AI adoption look like?

Change is sequenced so teams can finish and adjust before the next rollout, oversight time is protected within roles, and leaders agree in advance how recovered capacity will be used.

• How can leaders tell whether AI is creating real capacity?

By comparing the total human hours a result takes now, including review and correction, with what it took before, and by tracking where recovered time shows up in utilization and billable work.

Sources Cited

Laura Counts, "AI Promised to Free Up Workers' Time. UC Berkeley Haas Researchers Found the Opposite," UC Berkeley Haas News, February 18, 2026 (Q&A on in-progress research by Xingqi Maggie Ye and Aruna Ranganathan, also featured in Harvard Business Review). https://newsroom.haas.berkeley.edu/ai-promised-to-free-up-workers-time-uc-berkeley-haas-researchers-found-the-opposite/

Ya Ting Chuang, Hua Ling Chiang, and An Pan Lin, "Insights from the Job Demands–Resources Model: AI's Dual Impact on Employees' Work and Life Well-Being," International Journal of Information Management, Vol. 83, 2025, Article 102887. https://doi.org/10.1016/j.ijinfomgt.2025.102887