https://www.msn.com/en-us/money/other/microsoft-scales-back-ai-goals-because-almost-nobody-is-using-copilot/ar-AA1S5WkO
ChatGPT and Gemini are destroying the competition. Charlie is on the wrong side of this situation due to his financial ties to M$. Time to jettison him and find someone who actually knows what they are doing.
@cp You are correct, Agetic AIs are the future workers. Wells will fall further behind as the d-mb leadership is choosing the wrong tool, therefore will still have to rely on human workers for a long time without real automation, all the white cutting the human work force, and they stresse out with the best talents leaving in disgust.
The funniest thing is Copilot is a big time waster for me because somewhere I read they are scoring us on Copilot usage. So I make sure I spend time asking why oceans are blue or what is gravity. I bet my adoption score is sky high.
How weird that the obsequious slop bot generated slop in defense of slop. Truly an oracle to be consulted for everything.
@cp spot on-- it's mostly performative (copilot, training, hype). The actual agent AI stuff is going to be like cloud-- slow and a failure. Heard that many things won't even get started until end of next year (in other words, get something in 2027). There might be some development next year. But there's so much governance, vying for dollars that nothing is moving. And I get a peek at a set of use cases, they're lame AF but have inflated cost savings associated with them.
What's funny about this whole thing is they've been hyping it up like it's some wonder tool. It's fine for basic things, but unless you have a license you can't use it for internal chat type things. You can cut and paste a document, in the chat I suppose but that's about it. Even with the license it is limited to what it can see and goes after your email for some content as well.
Probably useful for coding related things and git though.
Bhai, this whole thread is focusing on the wrong fight. Arguing whether Copilot or ChatGPT gives more hallucinations is like arguing which leaky tap in a crumbling building is worse. The core issue isn't the chatbot UI; it's the fundamental quality of the underlying model and the strategic shallowness of just buying licenses.
CS can spoof everyone that purchasing tens of thousands of Copilot seats is "doing AI." It's not. It's just giving everyone a fancier, often unreliable autocomplete. The real value—as seen in what firms like JPMorgan Chase are building and what McKinsey advises—isn't in these isolated chatbox experiments. The trap is getting stuck in endless "generative AI" demos that don't move the business needle.
The true, durable value is in Agentic AI. Let me break it down simply:
Chatbox (Copilot, ChatGPT): It's a reactive tool. You ask a question or give a prompt, it gives a single answer or block of code. You do all the thinking, planning, and step-by-step work. It's a powerful assistant, but you are still the driver of every single task.
Agentic AI: It's an autonomous workforce. You define a complex objective (e.g., "Fix all the security vulnerabilities in this codebase," "Reconcile this month's financial transactions across these 3 legacy systems"). The AI agent breaks this goal down itself, plans the steps, uses tools (like running tests, querying databases, executing code), recovers from errors, and loops until the job is done. It's the driver.
Our leadership is making the classic mistake: chasing the visible, shiny vendor purchase (Copilot licenses) instead of building the foundational, intelligent automation (Agentic workflows) that actually transforms cost and capability. They want to say "We use AI" without doing the hard work of integrating true intelligence into our processes.
And about that MS article—low adoption happens when the tool is a mediocre model wrapped in corporate policy. It can't even OCR a simple table from a screenshot reliably (try getting Copilot to read a pasted image of a table; ChatGPT does it, but good luck here). If it fails at basic accuracy, how can we trust it for real work?
The death knell isn't for Copilot; it's for a strategy that thinks AI is a procurement problem, not a fundamental re-engineering one. They're buying spoons when they need to redesign the kitchen.
Copilot gives me wrong answers all the time. I’ll ask it questions against vendor documentation and it gets much wrong.
I do the same with Gemini Pro and I get the right answers.
Hate to inform you, but co-pilot is ChatGpt.
Yeah the big issue with Charlie is that he's not pushing OP's preferred hallucination engine lol
People pushing slop do not know what they're doing, no matter the chatbot.
deranged take, as always