I was not a fan of Microsoft’s Clippy. But I was waxing somewhat nostalgic about the little paper clip reading all the negative reviews of Microsoft’s latest foray into helpful assistants, its AI-based Copilot. David Linthicum wrote today on LinkedIn about the enterprise backlash, saying, “The company’s decision to introduce new licensing models, charge premium prices for AI features, and encourage hardware upgrades created deep skepticism.” He cited its intrusive design, general unhelpfulness and AI hallucinations, and evidence that just a small percentage of adoption by Office users as major obstacles and says it is a cautionary tale: Microsoft needs to listen more and impose less on its users.
Some wags (including Marc Benioff) have called Copilot Clippy 2.0. I don’t think that is a fair fight. We should at least bump up the version to 10.0. In many respects, Clippy was ahead of its time (read this historical look back to see why this author called it cutting-edge AI for 1996.)
I haven’t spent much time with Copilot, because I would rather do my internet lookups when I need them, not be distracted by some automated nag. True, Copilot can generate a lot of text with just a simple request. But a lot of AI slop, as it is called. Does it do a better job than Clippy in understanding context? Yes, but it still interrupts the creative flow, or at least my creative flow.
Over the many decades that I have become a not-so-famous writer I have learned how do my searches for the data and links in my stories. Now I type in complete sentences, rather than find three unique words that will drive better results. (That reminds me of What3words.com, which is a fascinating site, but I digress. See how annoying interrupting things can be? Sorry.)
So at least Clippy was cute. It had its detractors too, but also fans such as this short video that showed its future that is surprisingly fresh for something done a decade ago.
And for those of you who want to reanimate Clippy, here is some code that will bring it back to your desktop.
I will leave you with some words of wisdom from a colleague, Theresa Szczurek, who talks about finding joy and fear in AI in her latest newsletter: “You choose when to use AI. You decide where it adds value. You define ethical boundaries. You determine how it supports — not supplants — your strengths. AI is one tool among many. You are still the strategist. The leader. The creator.”
I use M365 Copilot pretty much every day. I like that what happens in Vegas, stays in Vegas. I’m not familiar with any interuptions that it interjects. It only gives me answers when I ask. It’s espcially helpful with Powershell – for those of us M365 admins who don’t use it everyday.
Clippy and Copilot fail for the same reason. They act without consent.
Theresa’s joy frame works because it does the opposite. It centers agency. When AI waits to be called, it feels like leverage. When it announces itself, it feels like surveillance, even when intentions are good.
The lesson is not that users hate AI. Users hate being overruled. They resent tools that presume authority over attention, timing, or intent.
The assistants that win will be dull by design. They’ll be quiet, patient and mostly invisible. They will default to absence and will never confuse being available with being entitled.
And yes, Clippy was cute. But that was never the problem. Timing was.
By the way, Ms. Szczurek’s post also connects to something I wrote recently about agency and refusal at scale. I argue that when systems get big enough, we’re tempted to resolve ethics by either exiting entirely or surrendering control—but both moves quietly erase responsibility.
Clippy and Copilot fail for the same reason cathedrals do when they go wrong: they assume entitlement. They act on people rather than with them. Consent, timing, and invitation matter more than intelligence.
If you’re interested, the piece is here: https://quistuipater.substack.com/p/agency-inside-the-cathedral