I was at Microsoft Ignite last week for the AI and Copilot rollout on Azure and Windows.<\/p>\n\n\n\n
As I watched the presentation, I was taken back in time to the 1990s launch of Microsoft Bob and that first attempt at creating a digital assistant called Clippy, neither of which met expectations back then. But now, Copilot will do far more than those two earlier offerings could even conceive of, and the power of the result, both inside and outside of Microsoft, is incredible.<\/p>\n\n\n\n\n\n\n\n
Before the event, I met with a company called Reply<\/a>, which specializes in getting companies ready for Copilot and setting up metrics so that they can confirm the benefits of the technology. Reply raved about how much more productive they and their clients were.<\/p>\n\n\n\n What is truly amazing is that the Copilot technology was still in its infancy only eight months ago. Last week, it went to general availability at top enterprises in medicine, farming, services, software, automotive, banking, and petrochemicals. Now, 70% of users indicate they don\u2019t want to work without this capability.<\/p>\n\n\n\n In the 40 years I\u2019ve been in this business, I\u2019ve never seen this kind of technology ramp so fast from a beta product to what it is today. Let\u2019s talk about that this week.<\/p>\n\n\n\n We\u2019ll close with my Product of the Week, an electric SUV that I just ordered that I think is the best you\u2019ll be able to buy before 2025: the Fisker Ocean. There are 62,000 orders right now, but since they only produce 2,000 a month, I\u2019ve got a bit of a wait.<\/p>\n\n\n\n What I find fascinating about Bob and Clippy is that Microsoft was way ahead of its time with both. It correctly determined that people wanted a computing experience where the computer did most of the work. Bob was an avatar-forward design that could, in theory, take commands in English and then do what you asked. Clippy was an assistant that sat on top of Windows to do the same thing without the avatar front-end.<\/p>\n\n\n\n What was both sad and funny about Bob was that, initially, it was meant to be used by people who were older and unable or unwilling to learn Windows. Sure enough, when surveyed, this group loved the product. But someone inside Microsoft tried positioning it as next-generation Windows, and people who used Windows hated it.<\/p>\n\n\n\n We didn\u2019t yet have AI. Clippy and Bob were scripted, command-oriented offerings, which meant they set the expectation that you could chat with your computer. When they couldn\u2019t meet that expectation, they failed.<\/p>\n\n\n\n The very nature of conversational AI is being able to do what Bob and Clippy wanted to do but could not, which is to allow you to direct the system as if you were talking to another person. Not only does this technology enable you to better converse with your computer, but it also learns how you work and can increasingly do your repetitive tasks, like responding to emails, setting up meetings, and even filling in for you during team meetings.<\/p>\n\n\n\n As I watched the presentation, I recalled a meeting with my old friend and former Microsoft CEO Steve Ballmer and how he struggled to prioritize customer requests. His issue was that he had a huge number of core customers, all national or multi-national companies, and he was overwhelmed with data that made him unable to use customer feedback to set priorities effectively.<\/p>\n\n\n\n Ballmer is one of the most brilliant people I\u2019ve known, particularly when it came to statistical data, but the complexity of the problem was still beyond his capabilities.<\/p>\n\n\n\n At Ignite, Microsoft highlighted Copilot for Services, where an even larger amount of feedback from virtually every customer, not just the largest, could be correlated in real time. The result was a prioritized list based on user-defined criteria, resulting in a customer-validated action plan.<\/p>\n\n\n\n If Ballmer had been able to use Copilot, he would have been far more successful personally, and Microsoft\u2019s products during his tenure would have been far more compelling.<\/p>\n\n\n\n Ballmer\u2019s problem fell into a class considered impossible to solve before Copilot because the amount of data and the speed at which that data changed made conventional analysis methods non-viable.<\/p>\n\n\n\n Now, AI works at machine speeds. It can deliver answers in a small fraction of the time those conventional methods require, suggesting that executives who are performing marginally could use this technology to become superstars, and superstar CEOs (of which there are very few) could use this technology to become unbeatable.<\/p>\n\n\n\n It is a true game-changer.<\/p>\n\n\n\n As mentioned in the opening, I met with Reply the day before the conference began. Reply has been working with many customers who put the beta version of this product into production. About three weeks ago, an update broke the product. Deploying a beta product always carries a huge risk, as evidenced by this incident.<\/p>\n\n\n\n Now, you\u2019d think folks would just say, okay, it\u2019s in beta, and a patch will be out shortly. However, more than two-thirds of users indicated they couldn\u2019t live without the product and had already become dependent on it. They wanted and needed a fix ASAP, and they got it. I don\u2019t recall ever seeing this level of anguish for a break in a beta product.<\/p>\n\n\n\n Initial feedback suggests a productivity increase of between 10% and 30%.<\/p>\n\n\n\n What these stats don\u2019t show is that this improvement is the result of Copilot doing repetitive and annoying administrative things that employees don\u2019t enjoy doing, like responding to emails, taking notes at meetings, creating agendas, mucking with Excel, and trying to remember how the heck to do pivot tables. Copilot was doing much of that for them in its beta iteration, and users were just learning the product.<\/p>\n\n\n\nThe Legacy of Bob and Clippy<\/h3>\n\n\n\n
Fixing Steve Ballmer\u2019s Problem<\/h3>\n\n\n\n
Reply\u2019s Copilot Beta Experience<\/h3>\n\n\n\n