Joe Fuqua
Intelligent Automation Architecture Strategy & Governance
Algorithm & Blues · Weekly
Charlotte, NC · Est. 1988
Algorithm & Blues

Room for One More

AI can make additional work worth attempting. When useful experiments become regular expectations, the time saved may already be committed to a larger workload.

Issue #79
Published November 15, 2026
Series Weekly publication
Source Original
A person works at a branching wooden desk while AI figures occupy connected workstations throughout a dimly lit office.
Conceptual editorial illustration.

There are plenty of things I’d like to try if I had more time, and AI makes some of them practical. That’s part of what I enjoy about using it. But when an experiment produces something useful, colleagues may start asking for it regularly. After a few of those, saving time becomes a little harder to account for.

UC Berkeley researchers Xingqi Maggie Ye and Aruna Ranganathan studied how employees at a roughly 200-person technology company incorporated generative AI into their work. Over eight months, through workplace observations and more than 40 interviews, they found employees taking on tasks they would previously have left to colleagues or not attempted. Their responsibilities expanded, often without a manager asking them to do more.

Employees also kept several activities going at once and sent prompts during lunch breaks or in the evening. They described enjoying the process of prompting and experimenting, yet feeling busier and more stretched when they reflected on the workday as a whole. The enthusiasm for each task didn’t necessarily carry over to how they felt about the workload.

The study followed one company, so it doesn’t establish that AI generally increases workloads. But it suggests a complication in estimating the time AI will save. As tasks become easier, employees may attempt work they wouldn’t have considered before.

A team could investigate questions it once had to leave unanswered. An analysis that was too expensive to produce regularly might become affordable. Those improvements could justify the investment even if the team spends just as much time working. They can also change what colleagues come to expect.

Consider an analyst using AI to add a regional comparison to a monthly report. It helps explain a result that the overall figures obscured, and colleagues ask to see it again. Over the next few months, they begin using it in their planning and requesting more detail when a region’s performance changes.

Preparing the comparison may take relatively little time. Explaining an unexpected result could require conversations with regional managers, additional analysis, and a judgment about whether the difference warrants action. What began as an addition to the report now includes answering questions and following up on concerns.

That changes how much time the analyst has available. To the extent that the saved hours go into preparing and explaining the expanded report, they’re already being used. Counting those hours as available for another assignment would count the benefit twice.

I suspect this will be easy to overlook when the analyst is pleased with the result. The comparison was their idea, and colleagues are making use of it. There’s little reason to discuss the additional work until it begins competing with another deadline.

By the next planning cycle, the expanded report could simply be part of what the analyst is expected to deliver. Continuing it may be worthwhile, but the time required needs to be accounted for. If the next assignment assumes the analyst still has all the hours AI originally saved, they’ll have to fit it around work the business has already come to rely on.

Algorithm & Blues publishes on Sundays. If this was forwarded to you, you can subscribe at joefuqua.blog.

Research reference

UC Berkeley Haas, “AI promised to free up workers’ time. UC Berkeley Haas researchers found the opposite.” February 18, 2026. Research by Xingqi Maggie Ye and Aruna Ranganathan.