Forum Discussion
Microsoft's Copilot: A Frustrating Flop in AI-Powered Productivity
I agree that expectations for Copilot were very high, and many users have been disappointed when it doesn't perform certain tasks as expected.
At the same time, I think part of the challenge is choosing the right AI tool for the right workflow. Copilot works best inside the Microsoft 365 ecosystem, while other AI assistants may be stronger for brainstorming, coding, or longer-form content.
I'd be interested to hear from others:
Which Copilot features have actually improved your productivity?
What tasks do you still prefer to do with ChatGPT or another AI assistant?
Real-world experiences from different users would make this discussion even more valuable.
Two use cases: code and information mining.
CODE - Copilot is fine if the code is 100% in the Microsoft universe ~and~ it's M code, SQL, stuff like that. I can generally get what I need fairly easily. If it's not all Microsoft, I just don't bother; Copilot will just run me in circles. I use Claude. Claude actually works. No circles.
INFO MINING - I can't believe how ungodly limited M365 Copilot's paid subscription is at finding information in my enterprise Microsoft apps. If I invoke Copilot from within Outlook, and ask it to find messages, it routinely misses half of them. I point this out, and it responds with a long-winded version of, "oops". Claude doesn't have access to my enterprise environment, so I can't make a comparison there.
OVERALL PARTLY OUT-OF-SCOPE IMPRESSIONS
If these tools are writing code, doing math, or some other logic-based work, they are varying degrees of effective. But asking any of them to summarize, say, the status of a current federal government policy using gold-standard news outlets - ABC, CBS, FOX, NBC, the New York Times, the Wall Street Journal, the Washington Post, Time, Newsweek, etc.? Every AI tool I've tried is shockingly inadequate. They all try so hard to find information and present it with so much confidence.
Then, I look under the hood. They're not citing gold standard news sources. They present what people write in their personal blogs ABOUT what they THINK the news source says. They use sources like Al Jazeera (!) They provide speculation from university PhD candidates as fact. They ignore the difference between the blog and the university and the gold standard source, even though I explicitly prompt for only specific named sources: ABC, CBS, FOX, NBC, the New York Times, the Wall Street Journal, the Washington Post, Time, Newsweek, etc.
Stick to code. Leave actual information gathering to people.
If Copilot is not going to provide accurate replies to search queries then perhaps it should not intercept searches from Edge and respond with apparently authoratative, but factually incorrect responses.
100% agree. See my response to hafizasadseo above.
