The AI Productivity Trap 

AI can make almost any process faster. But what does that speed actually accomplish? 

Measure Value, Not Just Productivity 

I used to start every Saturday morning the same way, planning what to feed my family of five for the upcoming week. I’d pace between the fridge and the pantry, phone in hand, scrolling through recipes and putting together a meal plan and grocery list, all while taking inbounds of “I don’t like blueberries anymore” and “Can we have taco night again?” 

This would generally take an hour and a half to two hours. Thanks to AI, my wife and I handle all of this asynchronously now, with an AI agent doing most of it in 20 to 30 minutes. 

I’ll be the first to admit it: there’s nothing groundbreaking here. A lot of parents I know are doing something similar with AI. 

But what I’ve noticed is that the hour or two it hands back doesn’t automatically become valuable just because I got it back. Some Saturday mornings it means more time with my kids. Other Saturdays it means a morning workout, which is nice when you’re trying to beat the summer heat. Same time saved, completely different outcome, depending on where it goes. It’s a part of AI productivity that I think gets missed. 

The first piece in this series argued that AI should earn trust through evaluation sets and evidence. This one is about a different kind of confusion: mistaking AI-enhanced outputs for AI-enhanced outcomes.  

The Principle 

AI can make productivity cheap, but it doesn’t automatically create value. Measure the value, not just the productivity. 

Efficiency measures like hours saved or documents produced have always been easier to track than outcomes like revenue recovered or problems prevented. It’s not new and it’s not really about AI. What has changed, I think, is how easy AI makes it to stop at the efficiency metric. 

When a task that used to take an hour now takes five minutes with AI, it’s tempting to declare victory. But five minutes is only useful if the work was worth doing in the first place, or if the time saved gets put toward something that matters. AI can make almost any process faster. That only matters if the thing you’re doing faster is actually worth doing. 

Sometimes productivity really is the outcome. If you’re trying to reduce the cost of processing a transaction or shorten a cycle time, efficiency is exactly what you should measure. The mistake comes earlier in the process: skipping the step where you decide what outcome you actually want before you start measuring anything. 

Peter Drucker made a similar distinction decades ago: “Efficiency is concerned with doing things right. Effectiveness is doing the right things.” AI makes the distinction especially relevant because it can make almost any process faster. Before celebrating the efficiency gain, make sure you know what the gain is supposed to accomplish. 

You 

Before you measure how much time AI saved you on something, consider what you were actually hoping to get out of it. 

Sometimes the answer really is time. Meal planning is a good example. Real time back, real value, assuming equal quality. 

Sometimes what you’re gaining isn’t time at all. I use an AI agent to track and suggest home maintenance, HVAC filter schedules, smoke detector testing, appliance ages and when they’ll likely need service, things I’d otherwise have to remember myself. It doesn’t save me any time. I think of it as insurance-policy AI. Its value is in helping me keep track of things and potentially catch small problems before they become expensive ones. 

And sometimes I’m just experimenting. A new tool comes out, or there’s a better image generator, and I can lose an afternoon setting it up and poking around just to see what it can do. That’s not necessarily wasted time; learning how to use AI is a legitimate outcome. But if I can’t finish the sentence “I want this so I can _____,” I should probably stop and figure out what I’m actually trying to accomplish. 

AI doesn’t have to save you time to create value. It can help you remember something, catch a problem early, make a better decision, or learn something new. What it can’t do is create value on its own just because you spent time with it. 

Your Team 

The same gap shows up at the team level, usually when output gets mistaken for outcome. 

Picture a hospital revenue cycle team that starts drafting twice as many appeal letters an hour with AI’s help. Great. But if the goal is recovering more revenue, what matters is what happens downstream. How many of those appeals actually result in overturned denials and payment? If the number of letters doubles but recovered revenue doesn’t move, the team has become more productive without becoming more effective. 

And sometimes the work you’re making more efficient exists because something upstream is broken. 

Paul LePage, via Becker’s, recently captured this problem particularly well: 

“The most dangerous trend in revenue cycle today is the normalization of workarounds instead of fixing the root cause of revenue leakage. Organizations continue to add staff, vendors, technology, and work queues to manage denials and other problems rather than addressing why those problems are occurring in the first place. AI and automation can make this even more dangerous by allowing us to become incredibly efficient at processing bad processes. The best revenue cycle isn’t the one that works the most problems; it’s the one that creates the fewest problems to work.” 

If a hospital has thousands of preventable denials, building an AI system that can process those denials twice as fast might look like a productivity win. But if the underlying goal is to collect the revenue you’re entitled to, preventing the denials may be far more valuable than processing them faster. 

Consider a banking team that is using AI to turn credit memos around 40% faster – this is clearly improved productivity. But what happens because they’re faster? More deals evaluated? Faster decisions? Lower cost? More time for the cases that actually require judgment? 

Those are the outcomes. If none of them change, the organization may simply have become faster at producing credit memos without improving any other areas of the business.  

Your Systems 

At the organizational level, measurements come a key focus. Decide what outcome you’re building toward before you build the dashboard, not after. 

The closer your measurement gets to business value, the more useful it becomes. AI adoption tells you whether people are using the technology. Productivity tells you whether they’re doing more. Outcomes tell you whether any of it matters. 

A dashboard that counts users, documents, or hours saved tells you something is happening. It doesn’t tell you whether the thing happening is the thing that matters. Work backwards instead. If the goal is faster claims recovery, track recovery. If the goal is better customer service, track the customer outcome, not the number of responses generated. 

None of this is new. Organizations have wrestled with the gap between output and outcome for decades, long before anyone was measuring what AI could do. AI didn’t create the need for this discipline. It has simply made it cheaper and easier to produce more output, which means there is more opportunity to mistake productivity for value. And when something becomes dramatically cheaper to produce, we tend to produce more of it. 

What This Means for Leaders 

Before starting an AI initiative, or even an individual experiment, ask: 

• What outcome are we actually trying to move? Separate that from the task AI is speeding up. 

• What would success look like? Define it before rollout, not after. 

• What are we measuring because it’s easier? If hours saved is the easiest number to produce, ask whether it’s the number that matters. 

• Who owns the outcome? That may be different from whoever owns the tool. 

Try this with your leadership team: pick the AI initiative you’re proudest of. Write down the output metric everyone keeps citing. Then write down the outcome it was supposed to serve. If you can’t draw a clean line between the two, maybe that’s your next conversation. 

The Takeaway 

AI can make almost anything faster, and sometimes fast is exactly the goal. 

But speed is a means, not a guarantee of value. The fact that AI can produce more work, faster and cheaper, doesn’t tell you whether that work was worth doing, or what you gained by doing it. 

Getting more done isn’t the same thing as getting more out of it. 

Don’t measure AI by how much it outputs. Measure what you get from it. 

About the Author

Matthew Hisscock is a Senior Consultant at Thrivence specializing in AI strategy, digital transformation, and enterprise technology. Prior to Thrivence, he served as a Vice President and AI Product Manager at Goldman Sachs, where he led enterprise AI initiatives, product strategy, and large-scale transformation programs. Today, he helps organizations turn emerging AI capabilities into practical, scalable solutions that drive measurable business value.