AI is Not a Product

Starting in 1888, and for 100 years thereafter, Sears-Roebuck was the original virtual mega retailer. You got a catalog to buy anything that could be bought and shipped, sent away your order and money, and, some weeks later, were unboxing the results. In 1918 you could buy everything from a new pair of shoes to heroin (for health).

Amongst this collection of wares, you could also purchase a relatively novel Electric Motor with Attachments. It was advertised as a stand alone electric motor with attachments for everything from a hand mixer to a vibrator; "Aids that every woman appreciates."

The electric motor was the main item. It was a powerhouse that could be modified for your needs. You did not just buy a mixer or a vibrator, you bought the motor and then the specialized attachments. This is a stark contrast to today's shopping where buying an electric motor requires going to specialized suppliers and, generally, a knowledge of what you need (torque, voltage, temperature rating, size, etc.). I'm guessing that, these days, the vast majority of vibrator or hand mixer consumers do not think about the electric motor specs of their device at all.

AI is currently being marketed along this same model. Buy (or more accurately, subscribe to) an AI, and then customize it to get whatever you need! This is generally how big new technologies enter the marketplace, because new technologies are aimed at engineers, designers, and builders who will then make things. Novel technologies are not just aimed at early adopters, they are aimed at builders. And the builders tend to be the ones that think about what consumers actually want and use.

When lasers entered popular consciousness. The technology was amazing and opened up whole new worlds from long range communication to holograms. And amongst those whole new worlds, maybe there was a pitch about encoding data (CDs, DVDs, digital laser discs), but I do not recall ever seeing a pitch about creating the best handheld cat toy or a device that would ensure your pictures are hanging straight. The original inventors were thinking about high power devices to send signals long distances or for drilling. They were not thinking about the low power, ease of use, or household tasks. I don't have sales numbers, but I'm guessing the number (if not the revenue) of laser pointers sold dwarfs the number of satellite optics alignment kits.

Even in the case of CDs, the vast majority of people don't care about the lasers, they care about the music or the videos or the games that are contained on those shiny little discs. And while I'm confident Sony cared about selling DVD burners and blank discs, I'm guessing they made a lot more money selling the albums, videos, and games. What's more, they were able to optimize their burn, duplicate, and package pipeline once they knew their endpoint was albums, movies, and games. Is there a PlayStation without the laser-powered CD business? Would we have a PlayStation if Sony was solely focused on making better lasers?

The quotidian needs tend to come last, but also tend to have the greatest breadth and most ingrained usage. This is because those quotidian products are the most user-centric. Unless you're an electrical engineer, there is no reason for you to think about what kind of motor is in your vibrator. And as far as I know, a fascination with electrical engineering is not why most people buy vibrators.

The vast majority of humans do not think about security, databases, efficient packet transfer, compression, matrix math, etc. except when required by their profession. What most humans care about is their access to SpongeBob videos and comparing couches and seeing what their friends are up to online.

If GE—or whomever—understood that their real product was vibrators and mixers earlier, they would have sold a lot more motors a lot earlier. By focusing on what users want and how they want to do things, you end up creating better technologies at lower labor and production costs for products that consumers actually want and use.

So it's weird how many conversations I've had or heard about with Directors, VPs, and CEOs at tech companies who are addressing AI with a monolithic focus. "We must AI!" From what I've seen, it's not even a question, it comes out as a statement, a vague and emphatic command on the promise that AI will increase productivity, lower labor, and create greater benefits for customers.

But AI is not a product unto itself. AI is not a product unto itself any more than concentrated light is a product unto itself. It's not an endpoint, it's technology that can power endpoints. What's more, treating it as an endpoint ensures no one is set up for success, except for the AI companies selling access to their models. Which is great for them, but bad for everyone else, including both businesses who treat it as such, their employees and their customers.

From a business perspective, a more effective tack is looking at how AI might be applied along different business and production vectors to the benefit of internal and external customers. Can we automate translation from one language to another or one form to another or one API to another? Would summarizing content be helpful for our customers or would it create a worse result than the human refined analysis and copy? How effectively do we currently prototype, and how could we do that better? What is our capacity for prototypes? Can we eliminate processes that create a better user experience? How is any of this evaluated?

From a customer perspective, a better tack is thinking about what the dumbest possible application of this technology would be that solves human needs. We use the culmination of decades of optics and power research to drive our pets crazy. We use the most advanced communication technology in the history of the world to trade bumper sticker-level arguments back and forth in public view. We obviously use them for other things too.

Bill Gates is almost certainly right when he says the future of work is AI-driven automation of just about everything. But we're not there yet. And, what's more, if you're in a leadership position, wanting things to be done cheaper, faster, and (theoretically) better, you are unlikely to get there until you start having real discussions about what your goals are, who they serve, how to measure them, and why anyone should care.

Once we re-center the user, whether internal or external, our tools will get better. Systems will become optimized and more reliable. And we will see greater user adoption, because internal and external customers will know why they are buying something and what they can expect from it. After all, you're probably a lot closer to selling vibrators than you are to selling electric motors.