Some 10 years ago, Microsoft tried to introduce a personal assistant for his Office suite, Clippy, in order to help their users discover new features. Even if there seem to be people that like it (MS says 50 %), it was an overall failure.

How would you frame the problem, and which suggestions do you have to improve its user experience? Which lessons can be learnt for similar assistants?

  • 1
    (I will add the take of Clifford Nass later. If nobody else did, that is :-)
    – giraff
    Nov 2, 2011 at 14:56
  • 29
    It looks like you're writing a question. Would you like help? Nov 2, 2011 at 16:11
  • LOL I always preferred the cat personality to the paperclip. The cat was a lot less manic.
    – PhillipW
    Nov 2, 2011 at 16:35
  • 1
    I would add a Sean Connery voice-over. "Ah, sho it looksh like you're writing a letter..." Oct 3, 2012 at 20:43

5 Answers 5


Siri seems to be the spiritual but not functional successor to Clippy. A major difference is that people request Siri's help whereas Clippy imposed help upon you.

Another interesting thing is that Clippy is an Embodied Agent. For decades people have thought "How cool would it be if using your computer was like talking to a person". From that thought they focused on making Embodied Agents, basically visual Avatars for the computer. Hard Research shows people don't particularly like these things nor do they help, generally speaking. Lots of research has tried to improve these agents, but I still believe the idea is largely impractical. Note that Siri has no "face".

Clippy was also very limited technically; Siri allows natural language and allows the user to tell Siri the problem so it can come up with a solution, Clippy simply inferred what you were doing and tried to guide you along; this is also an Adaptive Interface; something else that sounds like a neat idea--change the UI to what the user probably wants; what could be better? In reality these simply cause frustration.


To me, the biggest problem with Clippy was that it was so damn patronising. As if it knew what you were wanting to do, and just had to help you. I did have it running, mainly because the animations kept me amused ( I am a simple person really ), not for the advice or comments, which I turned off.

And, as @Ben says, it interrupted you doing x to tell you it looked like you were doing x and would you like some help. It should be requested, not imposed. It always seemed to behave like a super-geek, who was always hovering around, telling you what you should be doing.

Finally, it was less clever than it seemed to be. So having pretended to know more than you, it often seems to get the wrong type of answers, because it had no more information that the help system Siri is a novelty, and is significantly better than Clippy, I gather, as it should be so many years later. But I think that Siri will also be ignored in a year or so, after the novelty has worn off.

  • 2
    its the same with the 'would you like windows to diagnose the problem' adventures which: never works; is a just an exercise in clicking windows.
    – colmcq
    Nov 2, 2011 at 17:21

99% of the users didn't know how to use the features that were already there in the previous version, and had ZERO desire to learn something new and different much less get bugged constantly about how they were doing things that were wrong to begin with.

They knew what they wanted to do and were doing how they wanted to do it. In their minds, Clippy caused cognitive dissonance, tell EVERYONE they were doing it wrong all the time.


Clifford Nass says that Clippy behaves socially inappropriate, as it suggests the same help over and over again, effectively communicating that he "sees" the user for the first time.

So instead of improving its algorithms to make it more intelligent, he used a different social pattern: whenever a user answered "No" to his question "Was this help useful?", he would provoke the user to write an angry support mail to Microsoft. The user sympathized with Clippy, and suddenly loved it.

In general, Clifford Nass comes to the interesting conclusion that "computer are social actors" even though everybody denies it when asked.

  • Also interesting: Nass didn't improve the usability of Clippy, "only" its UX.
    – giraff
    Nov 3, 2011 at 7:04

Clippy's problems:

1) Invasive

2) Visually unappealing

3) Lacked UI integration with any of the other Office features

To make it better:

1) Have Clippy be an obvious feature that can be enabled or disabled. Preferably the first time you run the program (and not with every new session).

2) Instead of randomly appearing and floating in the document, make Clippy more visually integrated with the rest of the program. Maybe somewhere in the toolbar? As a drop-down feature that appears upon hover?

3) Get rid of the excessive animation (skidding across the page, etc.). It's unnecessary and a little condescending (i.e. "in order to get your attention, let's have a cutesy paper clip zoom around, kids!")

4) Give the user more control. Let them type their own questions. Let them activate or deactivate Clippy as they see fit. Hide Clippy entirely, or have him visually accessible without being visually "in your face".

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.