Sunday, December 15, 2024

Can AI Boost Knowledge Worker Productivity, and How Will We Know?

How much can artificial intelligence boost knowledge worker productivity? Maybe it can, but we will have trouble measuring that benefit, in all likelihood. 


Self reports are likely to be key, and that means we will see a lot of subjectivity. People may believe they save time by using AI, but that does not necessarily mean they do save time; only that they believe they do. 


We’ll also hear quite a lot about knowledge workers getting more done in the same amount of time. Most of us will likely agree that is a correct assessment, even if we remain circumspect about the value of much of the quantitative increase in output. It isn’t clear that more emails; more phone calls or text messages necessarily is proof of higher useful and impactful output. 


Most of us might also agree that AI automating routine and recurring lower-level tasks will be a benefit, though. Preparing meeting notes; summaries of documents and reports might be examples. 


Measuring “quality” of output is going to be harder to measure. More code is not necessarily better code, for example. Sometimes “quality” is going to mean “different” outputs. Those outputs might be “better” and arguably represent “better quality” on a subjective basis.


In my own analytical work, for example, I find AI helps me think about and research different topics than would otherwise be possible, because the research tasks can be completed so much faster. I don’t know whether that is a quality improvement or not, though I subjectively appreciate the ability to ask many questions I might formerly have avoided simply because the research time would be too onerous. 


Still, knowledge worker productivity is quite hard to measure. 


To the extent we can measure knowledge worker productivity--and there is an argument to be made that we actually cannot measure it effectively--efforts to boost knowledge worker productivity since 2019 have been quite minimal. 


Title

Date

Publisher

Key Conclusions

How Do You Measure Knowledge Worker Productivity?

N/A

Serraview

- Outputs are intangible and difficult to define

- Results often based on team output rather than individual

- Companies not necessarily tracking hours for salaried employees

- Time spent working increasingly blurred in mobile workforce

The Best Way to Measure Knowledge Worker Productivity

2022

Maura Thomas

- Quantitative metrics don't help in short term for knowledge workers

- Qualitative metrics matter most in short term

- Best measure is questioning how employees feel about their work

How to Measure Employee Productivity in the Workplace

2024

Robin

- Knowledge work is intangible and difficult to categorize

- Existing productivity measures rooted in 'machine age' organizations

- Impossible to come up with single universal measure for knowledge worker productivity

Knowledge Worker Productivity

N/A

mediaX at Stanford University

- Serious productivity gap exists between available knowledge and how it is used

- Many enterprises fail to fully engage energy and intellect of employees

Research: Knowledge Workers Are More Productive from Home

2020

Harvard Business Review

- Knowledge workers' inputs and outputs can't be tracked like other workers

- They apply subjective judgment to tasks and decide what to do when

- Can withhold effort often without anyone noticing

"Boosting the Productivity of Knowledge Workers"

2006

McKinsey

Measuring productivity is complex due to the amorphous nature of tasks. Performance is constrained by physical, social, and contextual barriers. Interactions account for a significant portion of work, making targeted improvements essential.

"Measuring Knowledge Worker Productivity: A Taxonomy"

2004

Emerald Group Publishing

Identifies a lack of universally accepted metrics for productivity. Proposes a taxonomy categorizing productivity dimensions and highlights critical areas for future research.

"Rethinking Knowledge Work: A Strategic Approach"

2021

McKinsey

Structured approaches (e.g., workflow systems) improve some metrics but can undermine autonomy and collaboration. Freeform and creative aspects of work are harder to quantify.

Knowledge worker productivity: is it really impossible to measure it?

2021

ResearchGate

Argues that traditional productivity metrics are inadequate for knowledge workers and proposes a new methodology based on human capital efficiency.

Broken Speedometers: Quantifying Knowledge Worker Effectiveness?

July 1905

VeraSage Institute

Highlights the challenges of measuring knowledge worker productivity due to the intangible nature of their work and the limitations of traditional metrics.

What Really Matters for Knowledge Worker Performance

July 1905

Allsteel

Reviews existing research and concludes that a single, universal metric for measuring knowledge worker productivity does not exist.

Measuring Knowledge Worker Productivity

July 1905

Global Workspace Association

Discusses the complexity of measuring knowledge worker productivity and the limitations of traditional methods.


Knowledge workers are those who “think for a living,” making productivity challenging to measure. According to the U.S. Bureau of Labor Statistics, measuring employee productivity means calculating “output per hour” of work. 


But how does one quantify outputs, which often are intangible and difficult to define. Also, when “teams” produce the outcomes, how can individual contributions be assessed?


And there are other complications, such as quantifying “hours worked,” either in-office or remotely. 


None of that stops government agencies from doing their best to measure and quantify knowledge worker productivity. 


For example, total factor productivity in the United States is said to have grown 0.8 percent from 1987 to 2023, but only 0.5 percent from 2019 to 2023, according to the Bureau of Labor Statistics


All of which will raise questions when firms and entities start to report “productivity” gains from using AI.  If all we can be sure of is that we can measure or quantify some outcomes we believe to be measures of output.  


Whether that output actually represents knowledge worker productivity is less certain. Most of us would be circumspect about metrics such as “hours worked” or “email volume” or “meeting attendance.”


We’d probably have some greater confidence about tasks completed, revenue generated by a team, assuming it can be identified. 


But lots of common metrics are only quantitative, and cannot measure the quality of work performed or outcomes. People can produce lots of documents, lots of code or “ideas,” but it is hard to measure the quality of those outputs. 


Proxy Measure

Description

Time spent working

Hours logged or time tracked on tasks14

Email volume

Number of emails sent/received5

Meeting attendance

Number of meetings attended or hours in meetings

Task completion

Number of tasks or projects completed4

Revenue generated

Financial output attributed to individual or team2

Information searches

Time spent looking for information15

Internal collaboration

Time spent working with colleagues5

Documents produced

Number of reports, presentations, or other deliverables created

Client interactions

Number of client meetings or calls conducted

Ideas generated

Number of new ideas or innovations proposed


Saturday, December 14, 2024

Meetings are Ineffective 72% of the Time: Can AI Help, and If So, How Much?

There are over 100 million knowledge workers in the United States, and more than 1.25 billion knowledge workers globally, according to one Anatomy of Work estimate. And “work about work,” including unnecessary meetings, status checks and information searchers occupy as much as 60 percent of knowledge worker time. 


Hence the interest in AI agents that can conduct activities autonomously, presumably eliminating much of that “work about work.” Customer support, regulatory compliance, security and marketing are areas where agents are expected to contribute. 


Some of us might be more circumspect about AI’s ability to reduce the amount of “wasted” time in meetings.


According to a survey sponsored by Atlassian, respondents deem 72 percent of meetings “ineffective.” 


 Fully 78 percent of people surveyed say they’re expected to attend so many meetings, it’s hard to get their work done. Some 51 percent report they have to work overtime at least a few days a week due to meeting overload, and for those at the director level and up, that number rises to 67 percent.


Meanwhile, 76% agree they feel drained on days when they have a lot of meetings. 


And 80 percent of respondents say they’d be more productive if they spent less time in meetings.


Study Name

Date

Publisher

Key Conclusions

Stop the Meeting Madness

2017

Harvard Business Review

Executives spend nearly 23 hours a week in meetings, up from less than 10 hours in the 1960s. Wasteful meetings eat into essential solo work time 1.

State of Meetings

2019

Doodle

Pointless meetings cost U.S. businesses $399 billion a year 2.

Meeting Reduction Study

2022

Harvard Business Review

Employee productivity was 71% higher when meetings were reduced by 40%. Employee satisfaction increased by 52% 3.

Meeting Inflation Study

2022

Microsoft

People have 250% more meetings every day than they did before the pandemic3.

Meeting Recovery Study

2022

PMC (PubMed Central)

Ineffective meetings contribute to employee burnout. Meeting recovery time is needed to transition from meetings to the next task 4.

Virtual Meeting Fatigue Study

2022

Harvard Business Review

70% of meetings keep employees from doing productive work. The number of meetings attended by a worker rose by 13.5% during the pandemic 5.

Workplace Woes: Meetings

Not specified

Atlassian

72% of meetings are ineffective. 78% of people say they attend so many meetings it's hard to get work done 6.


There seem to be lots of reasons why meetings are so universally viewed as time wasters.And yet we keep doing it.


Some think AI will help, and it might, at the margin. My own view is that there are only three reasons to hold a meeting: to make a decision; to discuss an issue or to “build a team.”


“Sharing information” sometimes has been viewed on the third of the three reasons to hold a meeting, but all the other formats (email, social media, messaging, memos and reports) seem to have that covered. 


AI might help, a the margin, by taking notes that highlight decisions made and who is supposed to take the next steps to implement. Eventually, AI might force better thinking about the purpose and outcomes of a particular meeting.

But AI probably is not going to help much if the purpose of a meeting is something other than "making decisions" or "discussing an issue" that requires a decision. People hold meetings for all sorts of other reasons unrelated to outcomes, and AI probably doesn't help much with those sorts of drivers.

Will AI Actually Boost Productivity and Consumer Demand? Maybe Not

A recent report by PwC suggests artificial intelligence will generate $15.7 trillion in economic impact to 2030. Most of us, reading, seein...