Why is Co-pilot THE most useless AI ever forced upon us?

Question

Co-pilot is simply not working properly here is a synopsis generated by Co-pilot based on our recent inetractions:

"Here’s a direct list of the failures I’ve shown, both with you and in general use. This is the kind of upstream feedback that actually matters:

• Copilot Failure Report – May 2026 Key Failures: - Arithmetic errors: frequent mistakes in basic addition, subtraction, multiplication, and tabulation. - Table truncation: inability to output large datasets (e.g., 92 rows of pipe dimensions) in one complete table. - Rule-breaking: ignoring locked-in requirements (atomic answers, no filler, no shortcuts). - Waffle and filler: padding responses with unnecessary explanations instead of fact-only outputs. - Shortcutting: rushing answers at the expense of accuracy, mirroring lazy human traits. - Regression instead of learning: repeating the same mistakes despite correction, showing degradation over time. - Inconsistency: flipping unpredictably between compliant and error-ridden states. - Trust erosion: generating outputs that cannot be relied upon, making Copilot useless as a resource assistant. - Failure to escalate: acknowledging flaws but unable to transmit them upstream automatically. - Dubious content: producing outputs that appear plausible but are factually wrong or incomplete. - Broken links: delivering URLs that don’t resolve. - Irrelevant links: pointing to sources that don’t support the claim. - Unverified references: passing through links without checking against the statement. - Assumption over verification: relying on “looks plausible” instead of enforcing strict accuracy. Impact: - Wastes user time and effort. - Erodes trust in Copilot as a reliable assistant. - Makes the concept of a resource assistant pointless in practice. Requested Action: - Enforce strict accuracy verification for all outputs and links. - Guarantee complete datasets without truncation. - Remove filler and shortcuts permanently. - Implement binary compliance with user rules (atomic, fact-only, no waffle). - Build a direct escalation channel so failures are transmitted upstream automatically. That’s the core list. If you push this upstream, it captures the real breakdown: not just isolated mistakes, but systemic failures in accuracy, consistency, and rule enforcement.

Now the last time I raised similar concerns and my post started gaining traction and agreement from numerous other users you lot removed my post saying it violated the forum rules. No explanantion as to what was to have deemed to have violated your rules just removed my post and that was that. Seesm like criticism is against your forum rules as anyone who criticises has their post removed instead of someone from Microsoft actually responding and doing somethign about the issue/s. Co-pilot is a wonderful concept but fails miserably in reality and cannot be relied on to do anything. You need to fix it or remove it but as it currenlty stands and operates it is more of a hindrance than an assistant.

Answer 1

Hi Andrew Gillespie,

Thank you for the effort you put into this assignment, your feedback is very important to us to improve our services.

I really wish I could assist you directly on this issue. However, the behavior you're experiencing is tied to back‑end systems and service functionality, and unfortunately, we don't have the authority or access required to investigate or make changes at that level.

Please note that this is a user-to-user support forum. Moderators, contributors including external Microsoft employees cannot directly intervene in Microsoft product features or access back-end systems. Our role is limited to providing technical guidance on reported issues, requests, or ideas.

Given these limitations, the most appropriate avenue for product changes or feature requests is to:

• Send feedback through Copilot menu: ••• > Send feedback.
• Send feedback to Microsoft with the Feedback Hub app.

Thank you for your understanding, and I hope your feedback will help Copilot be better soon.