The AI Productivity Paradox: Why More Automation Doesn’t Always Mean Better Work

We’re producing more work than ever.

More emails. More content. More reports. More code. More presentations. Artificial intelligence has made it possible to accomplish tasks in minutes that once took hours.

That’s an incredible opportunity. I work in marketing, leadership, and technology, and I’m excited about what AI makes possible. We can solve meaningful problems, serve people, and pursue opportunities that would have been hard to imagine just a few years ago.

But there’s a problem.

Producing more work doesn’t necessarily mean we’re doing more meaningful work.

You can use AI to write ten articles instead of one. But if nobody reads them, what did you accomplish? You can automate hundreds of emails. But if nobody responds, did you actually improve your marketing?

You can generate a comprehensive business strategy in seconds. But if nobody executes it, what changed?

The technology may be working exactly as intended. The problem is that we’re confusing activity with progress.

And I think there’s something deeper happening.

We’re increasingly measuring our days, careers, and even ourselves by how much we accomplish. AI offers an incredible way to increase our output, but it also raises a question we haven’t adequately answered.

What is all this productivity actually for?

As a Christian, I believe work is a good gift from God. We’re made to create, build, cultivate, solve problems, and contribute to the world around us. But I also believe our work was never intended to become the ultimate measure of our worth.

You don’t have to share my faith to recognize the tension. We all have limited time, people who depend on us, and responsibilities that extend beyond our jobs.

AI can help us accomplish more. But unless we understand what matters, we may simply become more efficient at pursuing the wrong things.

What Is the AI Productivity Paradox?

The AI productivity paradox describes the disconnect between AI’s ability to accelerate individual tasks and its less predictable effect on overall productivity. In simple terms, AI can make certain activities faster without necessarily improving the results that matter.

This isn’t an entirely new concern.

In 2017, economists Erik Brynjolfsson, Daniel Rock, and Chad Syverson published research examining the gap between advancing artificial intelligence and disappointing productivity statistics.

They considered several explanations, including unrealistic expectations, measurement problems, and the time organizations need to adapt to new technologies.

Their conclusion emphasized an important point: The full benefits of a major technological innovation often require changes in skills, processes, and organizations. Buying new technology isn’t enough. You have to change how work gets done.

That distinction is especially important today.

AI can generate a first draft almost instantly. But someone still needs to confirm the information is correct, the message makes sense, and the work accomplishes its purpose. In some situations, those activities take longer than producing the draft. We’ve accelerated one part of the process without necessarily improving the whole thing.

But there’s another question worth asking.

Even when AI genuinely saves us time, what are we doing with the time we’ve saved? Are we solving better problems? Serving more people? Investing in relationships? Becoming more thoughtful leaders?

Or are we simply filling the available space with more work?

The distinction matters because efficiency tells us how well we’re doing something. Wisdom helps us determine whether we should be doing it at all.

The Research on AI Productivity Is More Complicated Than You Think

Substantial evidence shows that artificial intelligence can improve productivity. But its benefits vary significantly depending on the work, the person using it, and how productivity is measured.

Consider two studies.

1. AI helped customer support agents become more productive

A study published in The Quarterly Journal of Economics in 2025 examined the introduction of generative AI assistance among 5,172 customer support agents.

Researchers found that AI increased productivity, measured by customer issues resolved per hour, by approximately 15% on average. The benefits were particularly significant for less experienced workers.

More experienced agents experienced smaller gains, along with slight reductions in some measures of quality. The researchers also found evidence that the technology helped less experienced workers learn from practices associated with more effective colleagues.

That’s encouraging. It suggests AI can help people learn, improve their performance, and become productive more quickly.

But notice something important.

The researchers didn’t simply count how many responses the agents produced. They measured issues resolved.

That’s a meaningful distinction. The goal of customer support isn’t to generate responses. It’s to help people solve their problems.

2. AI made some experienced software developers slower

In July 2025, researchers at METR published a surprising study on AI-assisted software development.

They examined 16 experienced open-source developers completing 246 real software-development tasks. Before starting, participants expected AI tools to reduce their completion times.

Instead, the researchers found that allowing AI tools increased completion times by approximately 19%. Even more interesting, the developers later believed AI had made them faster.

The tools felt productive, but the measured results told a different story.

The researchers cautioned that their findings applied to a particular group of experienced developers working in familiar, complex repositories. They did not demonstrate that AI slows down all software development.

In February 2026, METR updated its assessment, explaining that newer AI tools likely offered greater benefits, although limitations in the newer study prevented reliable estimates of those improvements.

The evidence continues to evolve.

These studies aren’t necessarily contradictory. They demonstrate that AI’s value depends on context, expertise, and the nature of the work.

Helping an inexperienced customer support agent resolve a common issue is different from helping a senior developer work inside a complicated codebase. The same technology can produce different results.

The question isn’t simply whether AI makes people faster. It’s whether AI helps them accomplish the right work more effectively.

Why More Automation Doesn’t Always Produce Better Results

I’ve found it useful to distinguish between outputs and outcomes.

An output is something you produce. An outcome is what changes because of what you produced.

A marketing team might produce twenty additional articles. That’s an output. More qualified traffic, better-informed customers, and increased revenue are outcomes.

A nonprofit might send thousands of additional donor messages. That’s an output. Stronger relationships, more meaningful engagement, and increased support for its mission are outcomes.

AI makes it easier than ever to increase outputs. But several reasons may keep those additional outputs from improving outcomes.

1. We’re Automating Work That Shouldn’t Exist

Not every task deserves to be completed.

Think about your average workweek. How many reports do you create that nobody reads? How many meetings exist because they’ve always existed? How many presentations are assembled because someone requested an update that could have been communicated in three sentences?

Now imagine automating all of them. You might save time completing each task, but you’re still spending time on work that contributes little value.

Technology doesn’t automatically distinguish between essential and unnecessary work. That’s your responsibility. Before automating something, ask whether the activity should exist in the first place. If a weekly report doesn’t help anyone make a decision, the answer probably isn’t to generate it faster. It’s to reconsider the report.

This principle is central to what I’ve written about having a productive day. Productivity starts with understanding what deserves your time, not simply finding ways to fill it.

There’s also a biblical principle here that I find particularly helpful. In Psalm 90:12, Moses asks God to teach us to number our days so that we may gain wisdom.

Think about that.

We don’t have unlimited time. Every hour spent on one thing is an hour we can't use for something else. And we can’t recover the days we’ve already lived.

The point isn’t to squeeze every possible accomplishment into our limited lives. It’s to become wiser about how we use the time we’ve been given.

Eliminating unnecessary work is often more valuable than automating it.

2. AI Can Create Additional Work for Someone Else

One of the most overlooked problems with AI is the downstream work it creates.

Consider an employee who uses AI to generate a detailed proposal. The document looks impressive. It’s organized, polished, and comprehensive. But the recommendations are generic. The assumptions haven’t been verified. The examples don’t reflect the organization’s actual circumstances.

Now the employee’s manager has to review the document, identify the problems, and rewrite substantial portions.

The employee saved time. The organization didn’t.

Researchers from BetterUp Labs and Stanford Social Media Lab studied this phenomenon in September 2025.

In their survey of 1,150 full-time U.S. desk workers, 40% reported receiving low-value AI-generated work from someone else during the previous month. Recipients reported spending an average of approximately two hours resolving each incident.

The researchers called this phenomenon workslop.

It describes something many organizations are beginning to recognize. AI makes it easy to produce work that looks finished without doing the thinking required to make it useful. And that work doesn’t disappear. It gets transferred to someone else.

I think this reveals a character issue as much as a productivity problem. We should take responsibility for the work we produce. That’s true whether we write every word ourselves or use AI to help us.

Passing along something that looks impressive but requires someone else to fix it isn’t good stewardship. It’s avoiding responsibility.

The Christian command to love our neighbors has practical implications for how we work. Our colleagues, customers, and employees aren’t simply resources to help us reach our goals. They’re people deserving of honesty, consideration, and respect.

You don’t have to be a Christian to appreciate that principle.

If your productivity improvement creates more work for your colleagues, you haven’t necessarily improved productivity. You’ve relocated the burden.

3. We’re Measuring Speed Instead of Quality

Speed matters, but it isn’t the only thing that matters.

Imagine two employees producing a research report. One uses AI to create the report in thirty minutes. The other spends three hours researching, evaluating sources, and developing recommendations.

If the first report contains inaccurate information and superficial analysis, was it really more productive?

The answer depends on what the report was supposed to accomplish.

If its purpose was to generate a rough outline, the first approach might be appropriate. If its purpose was to inform a major investment, the standard should be different.

The problem isn’t AI. It’s evaluating performance using the wrong measurements.

A faster response doesn’t necessarily mean a better customer experience. More content doesn’t necessarily mean more influence. More code doesn’t necessarily mean better software.

This is one reason I believe character matters as much as competence in leadership.

Integrity means doing the work honestly, even when nobody can see how much effort you invested. Excellence means caring about whether the final result is genuinely useful, not merely impressive.

For Christians, that commitment flows from a belief that our work is ultimately done before God. For everyone, it offers a simple professional standard: Take responsibility for the quality and consequences of what you produce.

4. Automation Can Weaken the Skills We Still Need

There’s another concern worth considering.

What happens when people stop practicing the skills AI performs for them?

Imagine a junior employee who relies on AI to conduct every analysis, write every recommendation, and solve every difficult problem. Initially, their output might improve.

But are they developing the judgment necessary to evaluate that output? Are they learning how to solve unfamiliar problems independently? Or are they becoming increasingly dependent on a system they don’t fully understand?

The answer isn’t the same for everyone or every task.

AI can strengthen learning by explaining concepts, providing feedback, and exposing people to better approaches. It can also allow people to bypass the difficult thinking through which expertise develops.

Ethan Mollick explores these opportunities and challenges in his book Co-Intelligence. He encourages people to experiment seriously with the technology while recognizing its limitations.

That distinction matters for leaders.

A responsible organization should ask not only what work AI can perform today, but what capabilities its employees will need tomorrow. If automation eliminates every opportunity for junior employees to practice foundational skills, who develops into the next generation of experts?

I believe good leadership involves developing people, not merely extracting results.

Sometimes that means allowing someone to struggle through a difficult problem, ask questions, make mistakes, and grow. That process may be less efficient in the moment, but it produces something more valuable over time.

A more capable person. And that’s an outcome worth pursuing.

5. We Haven’t Changed the Way We Work

Sometimes organizations introduce AI without changing anything else.

Employees receive access to new tools. Managers encourage them to experiment. Leadership announces an innovation initiative.

But existing workflows, approval processes, responsibilities, and performance expectations remain unchanged.

Now employees have another tool to learn and another set of expectations to manage.

The organization has adopted AI without redesigning the work.

A March 2026 study involving nearly 750 corporate executives found that perceived AI productivity gains exceeded measured gains, potentially because some business benefits take time to materialize.

This doesn’t mean AI investments are necessarily unsuccessful. It means deploying technology and realizing its value are different accomplishments.

We need more than new tools. We need better judgment about how the tools fit into the work we’re responsible for doing.

What Is Work Actually For?

This is where my Christian faith most directly shapes how I think about productivity.

I believe work is good.

In Genesis 2:15, before sin enters the world, God places Adam in the garden to work and keep it. Work wasn’t originally introduced as a punishment. It was part of humanity’s purpose.

We’re made to cultivate, create, build, and care for the world around us.

That perspective gives work dignity, whether you’re leading an organization, raising children, developing software, teaching students, or providing customer service.

But Scripture also teaches something about human dignity that our productivity-obsessed culture often overlooks.

Genesis 1 describes human beings as made in God's image.

Our value isn’t determined by how much we produce, the title we hold, or the economic value we generate. It is rooted in something more fundamental than our performance.

That has enormous implications for how we think about technology.

If people have inherent dignity, we shouldn’t evaluate them exclusively by productivity metrics. If work is good, technology should help us do it better. And if relationships matter, greater efficiency should create opportunities to serve people, not simply demand more from them.

This isn’t an argument against profit, ambition, or hard work.

I believe in building successful organizations. I believe in taking responsibility, pursuing excellence, and working diligently.

But success needs a definition.

A company can increase its output while weakening employee development. A leader can become remarkably efficient while neglecting their family. An organization can scale its communications while losing the personal relationships that made those communications valuable.

Those are real trade-offs, not merely philosophical concerns.

Jesus once asked a question recorded in Mark 8:36 about the value of gaining the whole world at the cost of one’s soul. In context, He’s addressing the far greater question of following Him and what it means to save or lose one’s life. He’s not giving a business productivity lesson. But His question confronts something deeper than business performance. It challenges our tendency to pursue visible success while overlooking what matters most.

That challenges me.

You can become extraordinarily successful at work while becoming less present as a husband, father, friend, or neighbor. And no productivity tool can make those priorities for you.

AI Isn’t the Enemy of Meaningful Work

It’s easy to take these concerns too far.

Some critics see AI primarily as a threat to employment, expertise, and human judgment. Others believe technological progress will naturally make organizations more productive and society more prosperous.

Both perspectives contain questions worth taking seriously.

Daron Acemoglu and Simon Johnson argue in Power and Progress that technological advancement doesn’t automatically distribute benefits broadly.

Organizational choices, incentives, and institutions influence who benefits from innovation.

Mollick, by contrast, emphasizes the opportunities AI creates to extend human capabilities and help people do work that previously required specialized expertise.

These perspectives aren’t mutually exclusive.

AI can make individuals more capable while creating organizational problems, unequal benefits, and poorly designed jobs. And we shouldn’t overlook its genuine potential to improve people’s lives.

AI can make information more accessible, help people communicate across language barriers, support medical research, and reduce the burden of repetitive administrative work.

A nonprofit may be able to serve more people with limited resources. A small business may gain access to capabilities it couldn’t previously afford. An employee may have more time for creative work and meaningful conversations.

Those are benefits worth pursuing.

In fact, I would argue that using technology to improve people’s lives is consistent with a Christian understanding of our responsibility to cultivate and care for the world.

I discuss some of these broader questions in What Does the Bible Say About AI?.

The question isn’t whether technology is inherently good or bad. It’s how we develop it, use it, and evaluate its consequences.

I don’t think the solution is to reject AI. I think we need to become much more disciplined about how we use it.

The goal shouldn’t be to automate as much as possible. The goal should be to use automation wherever it helps people produce better results, exercise sound judgment, and spend more time on work that matters.

That’s a different standard, and it requires a different approach.

How to Measure Whether AI Is Actually Improving Productivity

If you’re responsible for introducing AI into a business, ministry, nonprofit, or team, I recommend evaluating it through five questions.

These questions aren’t explicitly religious. They’re practical ways to make better decisions.

But my faith shapes the values behind them: responsibility, wisdom, service, integrity, and stewardship.

1. What outcome are we trying to improve?

Start with the result.

Maybe you want to reduce the time customers spend waiting for support. Maybe you want to help employees complete research more accurately. Maybe you want to improve donor retention or reduce administrative work.

Define the outcome before selecting the technology. Otherwise, it’s easy to mistake successful implementation for successful performance.

Then ask a second question: Who benefits from this improvement?

That helps you distinguish genuine progress from efficiency that merely makes one person’s job easier at someone else’s expense.

2. What does the current process cost?

Measure the existing workflow.

How long does it take? How many people participate? What errors occur? How often does work need to be revised?

You need a baseline. Without one, measuring improvement becomes a guessing game.

Include the less visible costs when possible, such as employee frustration, repetitive administrative burdens, and delays that affect customers.

3. What happens after AI completes its portion?

Don’t stop measuring when the AI tool finishes.

Include the time required for verification, revisions, approvals, integration, and corrections.

A task that previously took two hours may now take twenty minutes to generate. But if it requires another two hours of review, you haven’t saved time.

And if AI is used to interact with people, consider how it affects trust.

Are communications accurate? Is the automation appropriately disclosed? Are people’s privacy and preferences respected?

A process isn’t truly improved if you gain efficiency at the expense of trust.

4. Did quality improve, decline, or remain unchanged?

Establish standards that reflect the purpose of the work.

For customer support, evaluate resolution quality and satisfaction. For marketing, consider qualified traffic, engagement, and conversions.

For research, evaluate factual accuracy, source quality, and usefulness. For software, consider reliability, maintainability, and the time required to deliver working improvements.

More output is only valuable when it contributes to something meaningful.

5. Are we making people more capable?

This is the question I don’t want leaders to overlook.

Is AI helping employees become better thinkers, problem-solvers, and decision-makers? Or is it merely helping them produce more work?

A tool that saves time today while weakening your team’s capabilities tomorrow may create a problem you haven’t measured yet.

And consider what you’re doing with the time you’ve saved.

Are you investing it in employee development, stronger relationships, innovation, and better service?

Or have you simply increased expectations so that everyone has more work to do?

Efficiency should serve a purpose beyond itself.

Don’t Forget to Rest

One final part of this conversation deserves attention.

Rest.

If AI helps you finish your work faster, do you automatically need to fill that time with additional work?

I don’t think so.

One of the most striking aspects of the biblical account of creation is that God establishes a pattern of work and rest.

For Christians, rest isn’t merely an opportunity to recover so we can produce more. It includes worship, dependence on God, and the recognition that we aren’t ultimately responsible for keeping everything running.

We have limits, and those limits aren’t defects.

That’s difficult to remember in a culture that constantly encourages us to do more.

There’s always another project to pursue, another goal to achieve, or another opportunity to consider.

AI only increases those possibilities.

But more opportunities don’t eliminate our responsibilities to the people already in our lives.

Sometimes the wisest thing you can do with the time technology saves is to give your full attention to your spouse, play with your children, spend time with a friend, or simply stop working.

I’ve written more about the importance of protecting our attention in What Does the Bible Say About Distractions?.

This isn’t a rejection of ambition. It’s recognizing that ambition needs boundaries.

You can work hard without making work your identity. You can pursue excellence without treating every hour as an opportunity to produce something.

And you can rest without believing you’ve wasted your time.

For me, this is one of the most freeing aspects of Christianity.

The Christian message isn’t that we become valuable by accomplishing enough, being disciplined enough, or building an impressive life.

It’s that our deepest need is not something we can satisfy through our achievements.

At the center of Christianity is Jesus Christ, who came to reconcile sinners to God through His life, death, and resurrection. God’s acceptance isn’t a reward for our performance. Salvation is a gift of grace received through faith.

That changes how I think about work.

I don’t have to look to my accomplishments for ultimate meaning or worth. I can work diligently, take responsibility, make mistakes, learn, and rest because my hope rests in something greater than my performance.

Even if you don’t share that belief, I think it’s worth considering what happens when your worth is no longer tied to how productive you are.

How might that change the way you lead, work, and live?

The Future Belongs to People Who Know What Matters

We’re entering an age when producing information, content, and recommendations will become easier than ever.

That’s exciting. But greater capacity also creates greater responsibility.

Someone still needs to decide what deserves attention. Someone still needs to distinguish accurate information from convincing errors. Someone still needs to exercise judgment, understand people, and make difficult decisions.

Those responsibilities don’t disappear because technology becomes more capable. In many ways, they become more important.

The best leaders won’t necessarily be those who automate the most work.

They’ll be the ones who understand which work matters, how to improve it, and when technology should support rather than replace human judgment.

They’ll also recognize that people are more than the value of their output.

That’s the kind of leadership I want to practice. And it’s the kind of technology I believe we should build.

Technology that helps people do meaningful work, develop their abilities, strengthen relationships, and fulfill their responsibilities.

Technology that serves us rather than becoming another master demanding our time and attention.

Don’t confuse doing more with accomplishing more. And don’t confuse accomplishing more with living well.

Use AI. Experiment with it. Learn its capabilities and incorporate it into your work where it makes sense.

But keep asking the questions that matter most.

Are we actually making things better? Are we serving people well? Are we using our limited time on what matters?

Because in the end, a productive life isn’t simply one filled with accomplishments.

It’s one lived with wisdom, responsibility, love, and purpose.

That’s worth pursuing, with or without AI.