Even though I'm retired, I still like to watch the livestreams from conferences. Last week I was viewing the Ubuntu Summit 26.10 (named after their last release) and this week it has been Microsoft Ignite.
Microsoft Ignite is the developer conference and, as such, there was a lot of product announcements, mostly around AI. As expected, Microsoft was touting its comprehensive AI strategy, especially the integrated knowledge platform they call Microsoft IQ. There was the usual attempt to talk about their other products - the stuff that actually makes money - by connecting it to AI. "Host AI on Azure because it's built for it" or "Look at what AI can do in Microsoft Office 365 (plus look at all these other features in Office 365)" sort of thing.
What did get my attention was the subtle shift in how they talked about the value of AI. Up until now, AI vendors have had some version of "AI can do what humans can do" messaging with the unspoken ending of "so you don't need so many people." That was always a bit softer at Microsoft who had other products to sell and could leverage those products to get you to use AI. Still, the idea that a business might not need so many people because of AI was the subtext that sold AI to CEOs whether they admitted it or not. It became clear that CEOs hoped to reduce headcount, or at least growth in headcount, by using AI
That has changed. First off, many CEOs have noticed that AI can't do what humans do. AI is still a machine and is not aware or sentient. They just do rinse and repeat better than regular software does especially with unknown inputs. Second, they saw how much it cost, making the ROI on AI much harder to realize and, hence, more difficult to justify to their boards.
Microsoft under Satya Nadella has become good at noticing the change in the weather gauge. They saw this shift happening and changed direction. It's subtle but it's there. Instead of "AI can do what humans do" their Ignite messaging was more along the lines of "AI can help humans do what they do." The change is from human replacement to human augmentation. Don't get rid of those junior people; Instead, get the junior people working more like experienced people. Instead of AI taking away your job, let it take away your drudgery.
This should be much more successful messaging. Let's be honest, who wants to help train an AI to take over your job? No one, that's who. In fact, your motivation is to subvert the AI, accentuating its failures and downplaying its successes. Say to companies and their workers "You know that boring, annoying part of your job? Off load that to the machine so you can do more interesting things," and they will say "hell yes!" That's how they always approached computing.
Now, here's the downside - AI has to deliver. In my own AI use, I'm often disappointed by AI in a one step forward, one step back fashion. Here's some examples. Recently, I asked Copilot to merge two spreadsheets including specific information and add some calculated columns. Note that I said information not column names. The goal was to merge these spreadsheets based on what the information said, not the column headings. It worked brilliantly. I could have done this task with a bunch of copy and pastes and a few macros but that was drudgery to me. The analysis was important not the mechanics. AI wins one.
I then asked, in human language, to create a list based on the numbers in two of those columns. It failed spectacularly. Even after I told Copilot it was wrong and why, it could not fix its mistakes. AI loses one.
This sort of thing happens all the time. I asked Copilot to merge two documents from the Chat window. Nope, it just couldn't. I ask it to do the same within Microsoft Word, no problem. In this case, one step back. one step forward. So, I find that the messaging is spot on, but the execution is lacking. AI is, by nature, unpredictable. That's not what average end-users expect from their computer systems. They especially dislike that in mission critical applications or analysis that impacts the business.
In fact, predictability is one of the reasons we use computers at all. Yes, there is the speed of a computer. Fast but wrong, however, won't cut it. If I have to type prompts over and over or learn how to use them in just the right way, then AI offers little over conventional software.
In the end, I like what Microsoft has to say. Unfortunately, I've yet to see it do what it says. That's true for all the AI players by the way. Now ask me what I think about Anthropic getting a valuation of $1 trillion dollars...
