(Microsoft is a client of the author.)
Microsoft has unveiled its ‘Copilot’ generative AI for Office and I believe it is as big a game changer as the launch of Windows (and I was the launch analyst for Windows when I started my career).
With any advance like this, there is always an impact on employment, performance, and the trajectory of the related work. In thinking about generative AI, the third segment of the movie Fantasia — “The Sorcerer’s Apprentice” — comes to mind. It dramatizes what can happen when someone gets an incredible power, but doesn’t know how to use it properly and lands in an ocean of trouble. The most important part of using a force-multiplier tool like generative AI is to learn how to use it well and responsibly.
The tool’s strongest power is quantitative, because you can do far more in less time with it. (It has some qualitative capabilities, too, but those are less mature at this stage.) All too often, people who focus on quality do so at the expense of quantity, so boosting output using AI tools might allow them to turn out more top-notch work; conversely, those who work quickly may be sacrificing quality — so increasing their output will only make things worse.
My warning for the “fast” folks: these kinds of tools might be really attractive, but in the end you still need to focus on the quality of your work.
The Tesla example
Tesla is a good example. When it got into the automotive market, it built its business as a technology company should, with a heavy focus on automation and robotics. What it didn’t have is deep knowledge of how to build cars. As a result, it was able to build affordable electric cars before anyone else, but it struggled with unbelievable quality problems that lingered longer than they should have. Tesla focused on speed and cost containment, but the technology it used, robotics and automation, like Generative AI, sped up production without improving the quality.
In contrast, when Jaguar first built its F-Type in new automated factories, the company hired experts from Mercedes to run the facilities. While Jaguar has a poor quality reputation overall, the F-Type ranked far higher in quality because the people running the factory implemented a high-quality process that resulted in far fewer problems.
A technology such as robotics or AI can massively speed up processes, but it will speed up both good and bad practices, because it doesn’t know the difference. You can teach AI the difference, but if you don’t understand how to achieve high quality, you won’t be able to direct the AI to create it. In short, generative AI gives those who know how to create high-quality output a way to significantly increase their high-quality work. But those who trade off quality for volume will just produce more poor-quality work. That typically doesn’t end well.
While Tesla eventually improved the quality of its product, its initial advantage of being the only electric car company is sliding. Early indicators are that Tesla buyers are some of the first to buy electric cars from companies like Porsche, which has a reputation for high-quality automobiles. Tesla’s eroding reputation could have been avoided, had it focused on quality earlier.
Who benefits the most
What I find fascinating about Microsoft putting generative AI into Office is that even though this technology is very new, it is surprisingly capable, and it is advancing at an almost unbelievable rate. ChatGPT, at the base of Copilot (the name of Microsoft’s effort), is already in its fourth version, and we are just now getting it. The application is new, but the core technology is four generations old.
For a just-released product, the quality of the tool is unusually high, which puts the focus that would result from a low-quality result on the user. If users don’t quickly learn the strengths and weaknesses of this tool and focus on a high-quality outcome, they will become a problem for management to solve.
Those most at risk initially are those who use the tools poorly. They’ll create low-quality results at higher volume, making their shortcomings too obvious to ignore. In contrast, those who understand what this tool can do and train it to higher standards will stand out against their more poorly trained peers; they’ll not only survive this technology wave, but flourish because of it.
Copilot for PowerPoint
I’m incredibly excited about Copilot, but, like you, I’ll be focusing on learning how to use this tool properly before I rely heavily on it — and I’ll be constantly focused on quality while I let the tool improve my productivity (because I don’t want to become obsolete). One very interesting aspect in that regard is Copilot for PowerPoint.
This is a bit different than the other implementations that focus on quantity in that it turns text into presentations. Many of us — and I’m guilty of this, as well — use PowerPoint like speaker notes, but don’t use the full power of that tool to convey a visual message. Copilot can help users create amazing presentations because it synthesizes our ability to create a script with a more visceral visual medium. Ironically, some of my initial success in marketing was the ability to create better presentations, an ability that made me far more visible to management than those who were less skilled.
For those of you who have been kind of cheating with PowerPoint but know how to tell good stories, this tool is a godsend for presentations. It will create images in documents and books that visually convey the concept you’re trying to communicate. Granted, if you can’t tell a story to begin with, you’ll still be in trouble, but this one tool is what I’m most excited about.
How this technology will evolve
Generative AI is the first tool to truly learn from us. Over time, it is capable of learning and automating what we do at an increasingly faster rate. Like any automating process, this means if we work to eliminate our defects from the start, we’ll minimize the proliferation of those defects over time. If we don’t, we may eventually have to spend a great deal of time getting our AI helpers to unlearn all our bad practices.
I expect this tool class to be the foundation for the creation of digital twins, meaning the more effort we make assuring our twin is of higher quality than we are, the better it will be when it matures. We are at the very beginning of the evolution of these tools and it’s clear to me that those who embrace this technology — and learn how to use it effectively — will displace those who don’t, just as folks who embraced computers bypassed those who stayed with typewriters or calculators.
This is only the beginning of the AI wave. As with any such advance, it’s better to quickly learn how to swim with it because the alternative is to drown in it. And no one wants that.