Predictathon is BACK, BABY!

Let’s get right to this one, since the question is already in the form we prefer:

Will AI Lead to a Productivity Explosion?

Short Answer: Not for most organizations.

Why: Well, there’s two really good reasons for why it won’t lead to a productivity explosion (for most organizations). One is pretty well understood, the other is more speculative but I feel quite confident in it.

Reason 1: Because similar advancements have not produced productivity explosions.

As much as Sam Altman would like you to believe LLMs are a technological development unlike anything the world has ever seen they are, in fact, not. He (and many others) argue that it is more impactful than electricity. This is absurd. How can it be more impactful than the technology required to create and run it? That would be like saying fuel injection is more important than the invention of the car. It’s crazy.

I think you can make a huge list of technology that has had the potential to be as impactful as LLMs. These would probably include:

• Electricity
• Writing
• Agriculture
• The printing press
• The transistor
• The personal computer
• Calculus
• Indoor lighting
• Relational Databases
• Object oriented programming
• Networks
• The world wide web

You get the idea. Of these technologies I think you can make a compelling case that only “writing” and “agriculture” led to any kind of “productivity explosion.” The rest, even though they were incredible inventions, tended to lead to gradual improvement over time — not explosions.

Now part of that is because that’ simply how society works — it takes a long time to fully integrate a new technology into how we live our lives.

Another part is that most technologies are double edged swords, leading to decreases in productivity as often as increases (more on that in a second). After personal computers and spreadsheets were introduced to offices managers waited with bated breath to fire the teams of accountants they felt certain were now useless. Instead, most organizations hired more.

The whole point of this section is to simply point out this: LLMs are not special. They are just the latest in a long line of technological innovations, and what preceded them had similar expectations, and similar disappointments.

Reason 2: LLMs are a particularly sharp double-edged sword

OpenAI published an article recently about how awful AI will be when it comes to productivity.

I mean, they didn’t realize that’s what they were pointing out. They titled their paper “How AI is expanding what people do at work” and seemed quite pleased with how AI is letting people do things that they couldn’t previously do.

They call this phenomenon “task crossover.” They mention that:

A small-business owner can independently draft copy, review a contract, or perform basic financial analysis. A salesperson can use AI to explore a customer dataset that might once have gone to an analyst. A marketer can troubleshoot a website without waiting for a developer. In each case, AI changes not just how work gets done, but who does what.

Now, there are several important questions that they don’t even bother to ask but I would like to take the time to interrogate just two of them.

Are these people doing these new tasks well?

If we go back to how an LLM functions, we remember that it produces statistically average responses to prompts. SO! Are small-business owner’s reviewing contracts particularly well? No, they are doing so at a statistically average level. That doesn’t seem so bad until you remember that it’s statistically average according to the internet, not according to all lawyers (who don’t frequently post on the internet about their contract reviews).

So we can say that they are likely doing it average for a lay-person, and below average for a lawyer. Odds are also good that, no matter the contract they review, they’re getting similar advice.

This is because of what HBR has called “Trendslop.” Advice on the internet tends to look pretty homogeneous, so when asked for advice you usually get relatively similar answers. In some situations this isn’t bad, for example, most LLMs will tell you to diet and exercise if you ask it how to lose weight.

But success in business depends on finding and exploiting an advantage over your competitors, but when asked for strategies LLMs would return roughly the same advice, no matter the situation the organization was in. Essentially their advice was “Just do what everyone else is doing.”

So are they doing these tasks well? Not really, no. And because of that, producing an above average outcome, if possible at all, will probably take longer than if you’d just found someone who is good at that skill and asked (or paid) them to do it.

Are these people more productive?

This is much harder to answer! And that’s because we are AWFUL at measuring productivity. In fact, let’s define productivity to make sure we’re on the same page.

Most people believe productivity and business are the same thing. They would define productivity as “checking off to-do list items.” But this makes it easy for people to perform productivity without getting anything done — in fact, I guarantee you’re thinking about some people who do just that right now.

In reality, someone is productive when they are moving important goals closer to completion.

Even this simple definition is tricky. What are important goals? Who decides on them? What if management chooses a goal that isn’t actually important? Is moving a goal assigned to you by management closer to completion productive if you know accomplish the goal won’t have the outcome management believes it will?

In “How to measure anything,” Hubbard tells the story of an analytics team that he worked with who created dozens of daily, weekly, and monthly reports for a large organization — these reports and dashboard had all, at one time or another, been requested by leaders. When he asked them to dig into who reads the reports what they found is that they weren’t read. They sat un-perused in inboxes and on internal websites.

Were they busy? Yes! Were they productive? Well, by most measurements, yes. They were doing exactly what they were asked to do. But were they truly productive? Not really. They weren’t moving anyone closer to anything — they were doing something that had no impact.

It is a basic tenant of management to have people do things they are uniquely suited to. But what happens when everyone is suddenly capable of doing much, much more?

Well, as OpenAI pointed out, people start doing those things! In the article they brag that roughly half of occupation-related chatGPT messages are about some other occupation — they are “task crossover” events.

But let’s put it another way. Instead of spending time on things they are ostensibly good at, people are now spending roughly half their ChatGPT time on tasks that they are explicitly NOT hired to do, and thus will produce average or below average outcomes, with no way of double checking to know if that’s what they are producing, since it’s outside their domain of expertise.

In other, other words, people are spending significant time doing stuff they’re not good at poorly, instead of focusing on tasks associated with their unique skillsets.

Are these people more productive?

They are busier, that’s for sure. But are they moving an important goal towards completion?

So what organizations WILL benefit from AI?

Yes, yes. In the beginning I said the productivity explosion would pass by most organizations, but not all. So who are the few, the proud, the organizations that will actually see productivity gains from AI?

The organizations are those that have exceptional management, that can create a … let’s call this a “chain of true productivity.” To do this management must:

1. Know what their customers want and why
2. Have a unique strategy that outlines how their organizations provides for their customer’s needs
3. Knows how each role within the organization contributes to that overall strategy
4. Communicates this knowledge to all employees
5. Assigns them, or empowers them to self assign, tasks that directly contribute to the overall company strategy
6. Helps them stay focused on those tasks

This is way harder than it looks! Ask yourself these questions:

1. Do I know exactly how we fulfill our customers’ needs?
2. Can I explain exactly how the tasks I do on a day-to-day basis contribute to that?
3. Does management help me stay focused, or do they frequently give me directives that split my focus?

I think if you’re in sales you probably can say yes to numbers one and two, but if you’re in a support role, or it’s a particularly large organization, number one is probably pretty hazy, and number two might be a straight up guess.

But no matter what, I’m going to guess that the answer to number three is “they frequently give me directives that split my focus.” Almost every manager does this! And not because they’re typical, movie-style bad managers, but because they want to manage. And when you’re a manager it’s hard to go “Here’s this thing, work on it for a month” and then just … let people work.

Because AI leads people to be ostensibly capable of doing a wider variety of tasks, management (and even people themselves) will try to do a wider variety of tasks. But that is fools gold!

Those that use AI to do more of what they are already good at will see productivity benefits. Those that use it for things they haven’t typically done with see busy-ness benefits, but not productivity benefits. And most companies won’t even realize that’s a problem — OpenAI didn’t.