Reading Time: 6 minutes

Librarians are knowledge workers. We work on things that require attention to detail and investigation, whether we’re cataloging books or answering obscure reference questions. I am always interested in new research on interruptions and thinking around how to improve the work environment so that librarians can be as effective as possible.

Microsoft recently surveyed 31,000 people and used data from its other platforms like LinkedIn as part of an AI advocacy report. 2025: The Year the Frontier Firm is Born focuses on how AI agents will “gain increasing levels of capability over time that humans will need to harness….” You can get a sense of their forecasted progression with this graphic:

A dark blue graphic with white text. The top is entitled "Journey to the Frontier Firm" and "every organization's AI transformation will look different, but here's how we see it playing out over time". Below this text are three graphics. The first is labeled phase 1 and has a graphic that goes with the title "human with assistant".  The text below suggests that every employee has an AI assistant. The center graphic says phase 2 and "human agent teams", with the text below specifying these agents join teams as "digital colleagues" performing tasks under human direction.  The last graphic, on the right, is labeled phase 3 and is called "human-led, agent-operated". The text below says "humans set direction and agents execute business processes and workflows, checking in (with the humans?) as needed".
A graphic entitled Journey to the Frontier Firm with a three-phase approach from a “human with an AI assistant” to a “human-led, agent-operated” environment.

I think we probably all have some concern about artificial intelligence autonomously executing anything, let alone business processes and workflows. With the generative component peeled back, AI doesn’t sound very modern in that we are already using machine learning to automate business processes and workflows. It’s when AI starts creating that things seem to go haywire and the point of processes and workflows are that they are not intended to be creativity generating so much as fulfillment.

But I digress.

Microsoft doesn’t really explain why the firm is on the frontier. The description of this “entirely new organization” is one that is “[s]tructured around on-demand intelligence and powered by ‘hybrid’ teams of of humans + agents….” Given the definition of “frontier“, this sounds like these organizations are operating some distance behind the frontier, although perhaps on another planet.

All Day, All Interrupted

At least they’re finding some general agreement that employees and leaders see eye to eye on one thing: workers lack enough time or energy to do their work (which I suppose is defined as ‘as much work as can be monetized’). The theory sounds great, especially for knowledge workers like law librarians. We are not all commercially motivated or have profit mandates but there is always a hope that we can serve more people, broaden access to libraries and information, no matter what context we’re in. These “phase 2 digital colleagues” (AI agents) will equip the humans with “new skills that will scale their impact—freeing them to do new and more valuable work.” As a professional and as an organization manager, this is appealing if possible. While law librarians have ceilings on how far up the value chain we can go, reaching that ceiling remains a goal.

Already, 82% of leaders say they’re confident that they’ll use digital labor to expand workforce capacity in the next 12–18 months….
[But there is] a capacity gap: 53% of leaders say productivity must increase, but 80% of the global workforce—both employees and leaders—say they’re lacking enough time or energy to do their work

2025: The Year the Frontier Firm is Born, Microsoft Work Trend Index Annual Report, p. 2 (2025)

All of which is a long tangent to the data point that jumped out at me:

During the 9-5, employees are interrupted every 2 minutes by meetings, emails, or pings. Factor in activity outside of core work hours, and it adds up to 275 interruptions a day. 60% of meetings are ad hoc versus scheduled. Edits in PowerPoint spike 122% in the final 10 minutes before a meeting. Chats outside the 9-to-5 workday are up 15% YOY, with 58 messages now arriving before or after work hours. And meetings after 8 p.m. are also up 16% YOY, driven by an increase in cross–time zone work. It’s no surprise that nearly half of employees (48%)—and more than half of leaders (52%)—say their work feels chaotic and fragmented.

2025: The Year the Frontier Firm is Born, Microsoft Work Trend Index Annual Report (online text, data on p. 6)

Surely nobody actually calls it “the 9 to 5“, do they? I thought we were working 110%, 24/7/365 anyway.

275 is a lot. It is based on how people interact with Microsoft systems (“Microsoft 365 productivity signals”). I suppose it distinguishes between mail volume or interactions and notifications. For example, I may get 30 emails a day but, since I don’t keep my inbox open or notification sound on, I only see those emails when I check my mailbox a couple times a day. In my case, it’s not 30 interruptions if the interruptions are based on objects.

No Good Interruptions

There’s been evidence for some time that productivity can be derailed by distractions and interruptions. Each notification on your phone, each chime of your Teams or Slack chat, each new email can jar you out of productive work. Interruptions impact how people experience work and, as knowledge workers like librarians, require that we compensate in order to achieve our service goals.

Our data suggests that people compensate for interruptions by working faster, but this comes at a price: experiencing more stress, higher frustration, time pressure and effort. Individual differences exist in the management of interruptions: personality measures of openness to experience and need for personal structure predict disruption costs of interruption

The Cost of Interrupted Work: More Speed and Stress, Gloria Mark, Daniela Gudith & Ulrich Klocke, p. 1.

One of the things I found interesting in their research was whether there was such a thing as a good interruption. Every interruption has a disruption cost but if the interruption is related to the work you’re engaged in (drafting a budget and then interrupted by an email from your CFO on the budget, for example), maybe the cost isn’t entirely negative.

For example, one might be working on a paper and be interrupted by a completely different topic, such as a question about a budget. If an interruption has a different context than the current task at-hand, this could introduce a disruption cost as it involves a cognitive shift of context to attend to the interruption, and then one must reorient back to attend to the interrupted task. On the other hand, one might be interrupted by a question that concerns the same context as the paper one is working on. This might be beneficial but if the context of the interruption and primary task are similar, this could lead to interference with the primary task and in this way may introduce a disruption cost. A third possibility is that the interruption context may not matter.

The Cost of Interrupted Work: More Speed and Stress, Gloria Mark, Daniela Gudith & Ulrich Klocke, p. 1.

They found that the disruption cost was the same for every interruption, whether it was related to the interrupted context or not.

Together, our study … show[s] that interruptions that share a context with the main task may be perceived as being beneficial but the actual disruption cost is the same as with a different context.

The Cost of Interrupted Work: More Speed and Stress, Gloria Mark, Daniela Gudith & Ulrich Klocke, p. 4.

Whether the number of interruptions is truly 275 or not, then, the interruption is the thing to minimize. If they’re all going to negatively impact your work or impede your ability to achieve or remain in a flow state, then they need to be minimized or sequestered.

Interruptions v. Intelligence

This leaves me wondering why AI will be impactful unless we offload our communications and other interruptions to focus on the work. But the things that generate interruptions, whether they’re ad hoc meetings or a chat ping, are still integral to the work. The human agent shepherd (or “agent boss” as Microsoft terms it) seems like they’d be experiencing a greater number of interruptions managing the AI electric sheep than if they just created better boundaries against interruptions.

And that returns me to the leaders and employees issue. Leaders that expect 24/7/365 access to employees are the problem that needs to be solved. Employees should be able to experience “the 9-5” and then step away from the work. That would seem to cut a lot of those 275 interruptions out. They should also be able to do “the 9-5” with as much flexibility, both in time and space, as possible, if we’re really trying to maximize how individual knowledge workers actually work. I see you my fellow larks.

But it’s not just the balance of work within the realm of your broader life. Knowledge workers are working 168 hours a week (take THAT “the 9-5”). Our brains are processing our tasks and projects while we sleep. We get insights and ideas when we do a craft or play a video game or walk the dog. There’s a reason we “sleep on it.”

Ideally, then, leaders would eliminate interruptions for their employees. Fewer meetings would be one place to start. For staff who have required interrupted time—I’m thinking of reference librarians who have to be on a desk for hours and available to whatever issue comes in—a leader can ensure they have time that’s ring fenced for projects. This can be expanded by creating a clear culture around communications: it’s okay to only check your email periodically through the day, for example. Or to use asynchronous tools so that knowledge workers who have achieved a flow state can remain in it and check in on communications (that seems to be the biggest interrupter) once a milestone is reached.

Do I need to be an AI agent boss? Or even a shepherd? I think we can achieve many of the same outcomes without AI agents and the complexity and cost they create, and the skills diminishment they’re likely to cause. If we want all-day access to information, can’t we do that with text or video prepared by the knowledge worker that is then web accessible? If the issue is findability of that information, will findability be better if there is an AI agent available instead of a text file?

I think the reality is that the high value work cannot be done around the clock. We may be able to deploy digital colleagues after hours—while we do human things—but we can’t delegate the high value work to them. I don’t mean technically: we’re seeing plenty of lawyers—and even the State Bar of California for the recent bar exam—delegate the high value work to AI and the results are unsurprising.

It reminds me of the 24-hour access to libraries crowd. It sounds great but most people sleep, even the owls and the larks. A 24 hour library is an often empty library. Most people (and if we’re being profit-motivated, we are trying to reach the broadest audience in the cheapest way) who are not dealing with an emergency will be open to daylight hours. In fact, if we allowed employees to work hours that better suit their personal work style, I expect a law library could offer reference from 5am to 8pm with staff who would probably embrace the chance to start earlier or later. And, really, who cares what time technical services people work?

I don’t work in a “Frontier Firm” as envisioned by Microsoft, not now nor in the past. I understand their interest in finding a place for AI that doesn’t frighten business and supports the hope for expanding capacity and profits. But I think the answer, to the extent our productivity is impacted by interruptions, is already within the control of those leaders. We don’t need AI to achieve greater productivity and knowledge workers who are able to, regularly, be in a flow state and maximize their expertise.