Delegation and Leverage in the Age of AI

Why Effective Delegation Will Still Be a Killer Skill When Workers are Artificial

When I was 19, a friend told me about a Yale study (which, as it turned out, was never actually conducted) claiming that goal-setting drives success. Little did I know: I took it seriously, and it became one of my core habits.¹

Last year, I had the usual goals: exercise, eat better, publish better, teach better, etc. But like many of my mid-career peers, I’ve been drowning in obligations, many of which are the result of what I’d call work gluttony.

Work gluttony happens when your plate is full, but the extra work seems manageable…until it isn’t. You take on a little more, then a little more, and then suddenly realize: Oh, I don’t have enough time for this, no matter how efficient I am. That’s when the panic sets in.

You start breaking promises to your friends and colleagues. The work you do manage to complete suffers because your attention is stretched thin. And all the unfinished tasks lingering in your mind make it even harder to focus. Then, your sanity takes a nosedive.

And it’s not like I could stop adding more work to my plate. Part of becoming more senior in an organization is, in fact, being responsible for getting more done.

What I’ve come to realize is that my real problem wasn’t too much work, it was a lack of leverage—the ability to scale output without a proportional increase in my effort or time.²

I can’t say I figured out the leverage problem, but I did realize one thing I wasn’t doing effectively: handing off work to others. So, I’ve added a new goal that had never made my list before: learning how to delegate better.

Then, something else happened: ChatGPT.

Delegation Makes People Productive and Happy

Being the academic that I am, the first thing I did was search Google Scholar for research on delegation. I found plenty of work by economists on agency theory, and leadership in organizational behavior, but surprisingly, there is little empirical work on delegation itself³ (if someone knows of great studies with high-quality evidence on this topic, please share).

Then, I ran into an incredible study by Stephen J. Anderson (UT Austin) and David McKenzie (World Bank). Researchers and policymakers around the world have been trying to figure out how to make entrepreneurs more successful. For many, the working theory is that entrepreneurs lack critical skills and knowledge, and that training them will unlock better performance for their companies. This seems reasonable.

However, the premise of this study was different. Maybe the issue isn’t the skills entrepreneurs have, but the time they have. So the researchers conducted a randomized controlled trial to see whether outsourcing—delegating tasks to an external person—or insourcing—hiring someone to take on specific tasks—improved entrepreneurial performance more than training entrepreneurs to do the work themselves.

What they found was amazing. Here’s a direct quote from their abstract: “Insourcing and outsourcing both dominate business training and do at least as well as business consulting at half the cost.”

In fact, a few years ago my co-authors, Rembrand Koning, Solene Delecourt, Ronnie Chatterji and I published a study titled “When Does Advice Impact Startup Performance?” (here is the pdf) We conducted a randomized controlled trial with 100 growth-stage entrepreneurs in the Indian software product industry (along with iSPIRT). We found that entrepreneurs who received mentoring around effective delegation practices from their peers—e.g., structured management approaches including regular meetings, goal-setting, and feedback—significantly impacted startup success. Entrepreneurs who received advice from peers advocating these formal management practices experienced 28% greater growth and were 10 percentage points less likely to fail than those who followed a more informal management approach.

From Chatterji, Delecourt, Hasan and Koning (2019).

While these studies highlight the benefits of delegation for firm performance, delegation likely also has significant effects on the well-being of the delegator.

A fascinating study by Whillans et al., titled "Buying Time Promotes Happiness," used a field experiment to show that when people spend money on time-saving purchases like hiring someone to clean their house or babysit their child, they are significantly happier than when they spend that same money on physical goods.

Delegation is not just cost-effective and beneficial for business performance. It also makes people happier.

The surprising thing, I suppose, is why we are not more explicit at teaching delegation to our MBA students or, frankly, to our kids.⁴

Delegation Types: Structuring vs. Task Involvement

In my exploration of delegation, it became clear that (setting aside incentives for a moment) the delegator faces two primary choices.

The first is the extent to which the delegator is involved in the tasks they delegate. This refers to how much they engage in the micro-details of how the work gets done.

The second is the level of structure the delegator imposes. This includes defining the broad parameters of the task, outlining how it might be approached at a high level, and specifying the expected outcomes and how they should look.

One could imagine a 2x2 framework (yes, I teach MBAs) where one axis represents the level of structure imposed, ranging from low to high, and the other represents the level of direct involvement in the task, also ranging from low to high.

The ideal delegator sets clear expectations but stays out of execution. They know how to structure work and hand it off to someone who can complete it at a reasonable level of quality, let’s say at least 80% as well as they would have done it themselves.

Most of us, however, are far from this ideal, often stuck in one of the other three quadrants:

• Absent: Some managers provide neither structure nor involvement. They offer no clear guidance or check-ins, leading to confusion and misalignment. In the end, they are disappointed with the results but never set their team up for success in the first place.

• Micromanager: Others swing to the opposite extreme, setting rigid structures while staying deeply involved in every detail. They waste time overseeing tasks that should be delegated, creating frustration for their team and themselves.

• Collaborator: The most common mistake is being a collaborator rather than a true delegator. These managers stay heavily involved in tasks but don’t provide clear structure or expectations upfront. As a result, the work doesn’t come back in a way that meets their needs, forcing them to step in and fix or redo it.

What’s interesting about these three delegation styles—the absent manager, the micromanager, and the collaborator—is that none of them actually solve the problem of too much work and not enough time. They create little to no leverage for the delegator. Output still depends on the delegator’s time.

The absent manager doesn’t take advantage of having people work for them, so nothing extra gets done. The micromanager gets things done but wastes their own time doing work they shouldn’t be involved in. The collaborator brings in help, but they’re still too involved, using their mental bandwidth on tasks instead of freeing themselves up to focus on bigger problems.

Delegation isn’t just about offloading tasks. It’s about creating leverage, making sure things get done without you having to be in the weeds.

For most of history, delegation was limited by resources, if you couldn’t afford employees or assistants, you had to do the work yourself. But what if delegation no longer required other people?

Delegation in the Age of AI

Delegation used to depend on affordability, but many of us lacked employees to offload work to. That was my situation for a long time (and still is to some extent), and I overcame it by hiring people on Upwork.

Upwork was good for certain tasks, though I probably didn’t use it as effectively as I could have, but it did allow me to get work done that I couldn’t do myself (e.g., remake my slides, collect citations, etc.).

But today, delegation is within everyone's reach. If you consider what generative AI, especially ChatGPT and Claude, can do, these technologies function much like fast, reasonably skilled workers.

A key question, one studied by researchers including Enrique Ide and Eduard Talamas in their papers, is how AI models fit into organizational hierarchies. Their work builds on the research by Luis Garicano on hierarchy and delegation, modeling workers as problem-solvers within an organizational structure. Depending on the nature of the task and the cost of generative AI, Ide and Talamas explore how organizational hierarchies consisting of humans and AI might evolve.

The key insight for me? AI could enable a one-person organization with near-infinite agents handling delegated work—offering leverage at an unprecedented scale.

I don’t think we’re there yet. But my delegation framework above, adapted for the AI era, can help us see when real leverage might happen.

Right now, most of us who use AI regularly are still collaborators—what we might now call prompt engineers. We use AI to refine our work, but we’re mostly not leveraging it to scale our impact. I’m sure there are also tinkerers and programmers—either not using AI at all or approaching it like traditional computing, where every instruction has to be explicitly programmed. We’re mostly gaining efficiency but not leverage.

The real shift will come when AI enables us to scale impact without our constant input.

So, what does delegation in the lower-left quadrant look like? If you’re a financial analyst at an investment firm, it’s not just using AI to quickly scan financial statements or produce rough valuation models at a fraction of your usual accuracy. It’s building a comprehensive AI-driven workflow that can autonomously evaluate investment opportunities, from gathering financial data and industry research to generating detailed investment recommendations, with quality approaching 80% of a seasoned human analyst.⁵

That means integrating multiple AI tools, each specialized for different stages of the investment evaluation process, effectively mirroring the roles and functions of an entire analyst team rather than merely supporting an individual contributor.

In fact, a senior analyst would follow similar principles when assembling a human investment team. A critical question is whether these same organizational design strategies remain valid in this emerging AI agent-driven world.

My co-authors, Prasanna Tambe, Dokyun Lee, David Hsu, and Yuan Gao, and I have a new project titled Managing the Machine: Does Organization Theory Matter When Organizing AI? In it, we explore whether traditional models of organization design still hold when applied to organizations composed of AI agents rather than humans. So far, our findings suggest that core principles of organization design, such as hierarchy and diversity, remain surprisingly portable to the AI world. These concepts function in agentic organizations much as they do in human ones. This suggests that delegation, too, may operate similarly in this new landscape as it did in the old. More to come as the project progresses.

Getting Leverage in the World of Generative AI

I still haven’t figured out how to delegate well. But my guess is that the people who do—especially those who learn to delegate to AI through systemic structuring of AI agents, not just collaborate with it—will have armies of agents scaling their impact. The real shift won’t be just working better, but leverage; it’ll be doing more than we ever could alone.

1. Yes, there is actual research that supports this.

2. You can also decide not to scale and stay small. Many individuals and firms decide to stay small. But I imagine most of us are in the group where we’d like to scale, but might not have the tools or frameworks to help us do so.

3. There are thousands of papers on delegation in economics, organization theory, and even in organizational behavior. This work is excellent and very insightful. Surprisingly however, there are far fewer larger scale empirical studies on delegation. I think this is a fantastic opportunity for students who want to study a first-order organizational process.

4. You might be thinking: My kids already delegate cleaning their rooms, doing their laundry, and driving them everywhere to me!

5. The 80% number comes from a friend who started a chain of healthcare companies. His advice is that when you delegate, you shouldn’t expect 100% of what you do; 80% is good enough. There should be allowable slippage.