This post was originally published on this site.
In their latest Work Trend Index Annual Report released this week, Microsoft researchers address some of the biggest unresolved questions prompted by businesses’ use of AI, including what it means for the quality and quantity of jobs and how the organization and management of companies need to evolve. Some takeaways from the report:
- Some 33% of senior executives globally say they will consider using AI to reduce headcount in the next 12 to 18 months. Some 45% say they will consider maintaining headcount but using AI as “digital labor.”
- Nearly half of executives say their organizations are using AI agents to fully automate processes. Leading-edge businesses are more likely than their peers to use AI for marketing, customer success, internal communications, and data science.
- As AI agents come into increasing use, Microsoft posits that the organizational chart—also known as the “org chart” or organigram— could be replaced by a “work chart.” This is “a dynamic, outcome-driven model where teams form around goals, not functions, powered by agents that expand employee scope,” write Microsoft’s researchers.
- AI will allow workers to do “more complex, strategic work,” earlier in their careers, according to 83% of executives surveyed.
- Around-the-clock availability is the number one reason workers say they use AI over turning to a colleague or a manager, with 42% saying that.
- AI could help manage the intense pace of white-collar work. Work-related messages have increased 15% outside of the 9am-5pm workday, compared to a year earlier, and meetings after 8pm are up 16%. Edits in PowerPoint documents surge 122% in the final 10 minutes before a meeting, suggesting workers are pushed right up against deadlines as much as, or more than, ever. Microsoft found that workers at firms with more pervasive AI usage don’t seem to experience such disrupted, long workdays as much.
To further explore these issues, we spoke with Microsoft senior research director Alexia Cambon, who led the research in the latest report. Here are excerpts from that conversation, edited for space and clarity:
You write in the report about the shift from an org chart to a ‘work chart.’ If I wanted to create a work chart, are there any models?
We explored several, and it’s telling that we decided not to put any in the report. Because, for one, it’s very hard to imagine what work looks like. The second part is that work as we know it today will change. So it’s hard to create an org chart for work we don’t know exists yet. But we saw some principles that we felt could help shape what the work chart looks like. The first being obviously that it’s shaped around jobs to be done, not around siloed areas of expertise, which is how the org chart works today. We specialize very early in the human life cycle, and then we build companies around that specialized human expertise. When expertise doesn’t live in humans anymore, but lives in agents in addition to humans, you can start to question whether or not the org can be structured differently.
So we call out the Hollywood model as one example of a different type of org chart where on movie sets right now, they draw in the best experts, the best cinematographer, the actors, the crew, and then they all work on a movie for nine months and then they disband and never come together again unless they’re making the sequel. That’s one scenario. Another one that we raised in the report is the idea of agent-first functions and human-first functions. The way you structure the org chart is we are going to divide work up based on what is the best type of work for an agent to do, and what is the best type of work for a human to do? That is a big question every leader should be asking themselves right now.
The Hollywood model is a successful model, but it’s different from most people’s models for employment. What are the implications for the quantity of jobs and the permanence of jobs? People today have jobs indefinitely in organizations because they have expertise that sits in an org chart and what you’re describing is a different model….
For one, I don’t think we’re going to run out of work, and I don’t think digital labor will substitute human labor. The Hollywood model, the reason that model is hard to replicate in current corporate organizations, it’s because beyond the monetary costs associated with talent—which again are justifiable; real talent, real expertise costs money because it’s not abundant, it’s scarce—it also carries a lot of transactional costs. If I want to bring together a team of experts to work on a specific project, there’s an onboarding cost, there’s an upskilling cost, there’s an assimilation cost, and the luxury of time is not something organizations often have.
But what happens in a world in which those transactional costs don’t exist when you’re bringing together agents? That then begs the question again of, what are the types of work that we want to outsource to agents so that we can onboard that expertise much more quickly where it’s needed? Then all the things that humans bring to the table—whether that’s value or judgment or high-stakes decision making or friction and collaboration, which is an incredibly important aspect of human work—where are those needed and where do we then want to pay more of those transactional costs because it’s worth it?
What would you say are the enduring skills for humans to focus on, to train for in this scenario that you’re foreseeing?
You’re essentially asking what are the skills that AI will rarify? It will be all the things that AI and agents can’t do. The ability to feel connected to another human is not something an agent can provide. Daniel Susskind is a great AI economist and he released a paper that I loved a few weeks ago that literally asked the question, what will remain for humans to do? He comes up with three categories, the first of which is the types of work where it’s just more efficient for a human to partner with AI, not to outsource it fully to AI. The second is where we have a human preference for humans to do the work. For example, if I was to go get a medical diagnosis, maybe I don’t want to get that from an agent. Maybe I prefer to get that from my human doctor. The third one is moral imperatives. Where have we as a society decided we are holding ourselves accountable for this decision being made by a human?
All of those types of work, my hypothesis would be that they involve fundamentally very human traits. The ability for us to feel connected, the ability for us to feel seen. Those are not things agents can provide.
There’s a recent study of material scientists that showed that their happiness went down even as AI made them much more productive. And there’s the study of coders using GitHub Copilot that showed that their collaboration and interaction with colleagues went down when they’re using AI tools. This research suggests that when you use AI tools, your job satisfaction and your collaboration could both drop. Do you have any analysis there?
Have you read the ‘cybernetic teammate’ paper? It’s an experiment they ran with Proctor & Gamble, about 700 employees. That found the opposite to what you said. They saw outcomes across quality, speed, but also satisfaction go up for individuals and teams equipped with AI. There are a lot of variable factors that we need to take into consideration when we are looking at the outcomes of these studies.
But I will say what we found in the data for this research, one of the questions we ask people is, why do you turn to AI over a colleague specifically? Not just why do you turn to AI in general—why do you turn to AI over a colleague? My hypothesis going into that question was that the answers would be all related to things to do with frustrations about your human colleagues. So Alexia is really slow, or I don’t find Alexia to be particularly fun to work with. All those things ranked bottom. The least selected one was, I can take the full credit when I work with AI as opposed to working with a human colleague.
All the things that ranked top were to do with the net new unique qualities of AI that I don’t think people are talking enough about. The top most selected answer was 24-7 availability. That’s something we’ve never really experienced before, having a 24-7 on-demand resource that we can turn to. The second selected answer was machine speed and quality. The fact that AI can compute a thousand data points in a second can reign over large amounts of data. That’s not something a human can do, quite rightfully. So the fact that our attention is finite is why it’s so valuable. Then the third one was ideas on tap. The fact that if I were to ask you now, Kevin, give me a great idea in the next three seconds, you’d probably struggle—with good reason. You’re a human. You need the creative environment to stimulate you to provide that idea, but you can do that with AI. That told me that AI is not substituting humans. In collaboration, it’s an additive force.
Read the new Microsoft Work Trend Index Annual Report.