Clippy: The Unauthorized Biography

Before Cortana, before Siri, before Alexa, even before smart interactive voice response (IVR) systems, there was Microsoft’s Office Assistant, one of the first widely deployed digital assistants that shipped as part of Microsoft Office in 1997. You might remember this feature better as the animated paper clip Clippy, even though Microsoft insists its (his? her?) official name is ClippIt. Doesn’t ring a bell? How about this: “It looks like you’re writing a letter”? Even though Microsoft officially retired Clippy in April 2011, the character continuously resurfaces as a meme in the zeitgeist. Try https://www.reddit.com/r/Clippy/ to see an assortment. Or check out Microsoft Cloud Developer Advocate Chloe Condon’s business card in her tweet (one of our inspirations for making this history): https://twitter.com/chloecondon/status/1098732324622020609?lang=en In this episode, a16z Board Partner and longtime Microsoft executive Steven Sinofsky shares his personal remembrances of Clippy with a16z Operating Partner Frank Chen. They discuss these and many more vignettes from the history of Clippy: - What problem the newly formed Microsoft Office team was trying to solve: https://youtu.be/8bhjNvSSuLM?t=129 - Which Stanford University and Microsoft Research ideas informed the design of the feature: https://youtu.be/8bhjNvSSuLM?t=396 - How the project was approved, what size team worked on it, and what BillG thought of it (warning: f-bomb ahead!): https://youtu.be/8bhjNvSSuLM?t=1121 - What user testing the team did on both the look-and-feel of the feature, as well as the cast of characters that ended up in the gallery (dolphins! geniuses not named Einstein! leprechauns!): https://youtu.be/8bhjNvSSuLM?t=2155 - How a pair of legendary Disney animators inspired the "stage" the characters perform on: https://youtu.be/8bhjNvSSuLM?t=1944 - How the feature got launched and marketed, and early reactions from users and reviewers: https://youtu.be/8bhjNvSSuLM?t=2851 - How Clippy got an elaborate real-world retirement party in San Francisco, complete with a blues song: https://youtu.be/8bhjNvSSuLM?t=3427 - Watch Clippy featured on Conan O'Brien's late night TV show: https://youtu.be/8bhjNvSSuLM?t=3927 - What Clippy can tell us about the future of digital assistants: https://youtu.be/8bhjNvSSuLM?t=4030 - The nature of risk-taking in software product development: https://youtu.be/8bhjNvSSuLM?t=4424 Enjoy watching the full arc of Clippy's life from prototype to retirement to cultural meme. *** The views expressed here are those of the individual AH Capital Management, L. L. C. (“a16z”) personnel quoted and are not the views of a16z or its affiliates. Certain information contained in here has been obtained from third-party sources, including from portfolio companies of funds managed by a16z. While taken from sources believed to be reliable, a16z has not independently verified such information and makes no representations about the enduring accuracy of the information or its appropriateness for a given situation. This content is provided for informational purposes only, and should not be relied upon as legal, business, investment, or tax advice. You should consult your own advisers as to those matters. References to any securities or digital assets are for illustrative purposes only, and do not constitute an investment recommendation or offer to provide investment advisory services. Furthermore, this content is not directed at nor intended for use by any investors or prospective investors, and may not under any circumstances be relied upon when making a decision to invest in any fund managed by a16z. (An offering to invest in an a16z fund will be made only by the private placement memorandum, subscription agreement, and other relevant documentation of any such fund and should be read in their entirety.) Any investments or portfolio companies mentioned, referred to, or described are not representative of all investments in vehicles managed by a16z, and there can be no assurance that the investments will be profitable or that other investments made in the future will have similar characteristics or results. A list of investments made by funds managed by Andreessen Horowitz (excluding investments and certain publicly traded cryptocurrencies/ digital assets for which the issuer has not provided permission for a16z to disclose publicly) is available at https://a16z.com/investments/. Charts and graphs provided within are for informational purposes solely and should not be relied upon when making any investment decision. Past performance is not indicative of future results. The content speaks only as of the date indicated. Any projections, estimates, forecasts, targets, prospects, and/or opinions expressed in these materials are subject to change without notice and may differ or be contrary to opinions expressed by others. Please see https://a16z.com/disclosures for additional important information.

• 0:13Speaker 1

  Welcome to the A16Z YouTube channel. I'm Frank Chen. Today I am here with the legendary Steven Sonofsky. Stevens is a venture and board partner with us before he came to us at A16Z, he had a storied career at Microsoft. And unless you're like three years old, you've used his products, Windows and Office. And today, the origin for the video that we're going to film now, we were inspired by a Twitter user named Chloe Condon. So hello, Chloe. Chloe posted a tweet of her brand new business card which had Clippy on it.

• 0:53Speaker 0

  Clip it.

• 0:54Speaker 1

  Clip it. Flip it. Nobody calls them Clipping. No, Mike. Yeah. Clippy. And Clippy was a feature inside Microsoft Word for a couple releases. And based on Chloe's tweet, we wanted to do the definitive history of Clippy, because I know a lot of you have seen Clippy, use Clippy, and have a love-hate relationship with Clippy. So let's talk all about Clippy. So welcome, Steven.

• 1:17Speaker 0

  Well, thank you, and thank you for that overly kind introduction. And Chloe, thanks for the inspiration for the feature. Although I will say that many, many people have inspired us in in this regard. But really the bit making the business card and the jewelry and the pins, I think it's time. So I'll be doing the best I can to share my perspective on the story. Of course, you know, there were a good dozen people working on this, and I was just one. So it's

• 1:47Speaker 1

  this is a group effort for sure. Yeah. Awesome. Well, part of the reason we want to do this is Clippy keeps on showing up in the cultural zeitgeist. And you know, we'll sort of weave that into our conversation in Chloe's business card and jewelry is sort of only the latest manifestation. It's crazy. Yeah. It's awesome. So maybe let's go back in time so that people understand why you built Clippy. So maybe let's start with what were PCs like back then and what year are we talking about?

• 2:13Speaker 0

  Sure. So we're actually talking about the genesis of the feature being in the early to mid-1990s. The big thing is it's before Windows 95. And so the you know, the graphical interface that we know as the desktop and stuff is is only on the Mac.

• 2:33Speaker 1

  Yeah.

• 2:33Speaker 0

  And it so therefore it's only on like five or 10% of people using a computer. Right. And so like a computer back then is something with two to four megabytes, megabytes of RAM. Yeah, not

• 2:48Speaker 1

  megabytes kids.

• 2:49Speaker 0

  Yeah, a 20 to 40 megabyte hard drive maybe. If you

• 2:54Speaker 1

  were rich. If you were rich. Otherwise, you got the floppy version.

• 2:58Speaker 0

  And uh the screen was, you know, uh a VGA 640 by 480 screen with 16 or maybe 256 colors.

• 3:07Speaker 1

  Yeah. If you had the fast graphics card, you got 256 colors.

• 3:12Speaker 0

  So that's that's basically the hardware at the time. Now the the Office products were relatively new. The the first successful version of Office on Windows was only two years old. And so what we were doing as a team was actually building the next version of Office, uh, which would become Office 95. Actually, just in Microsoft Lingo, it would become Microsoft Office for Windows 95 version 7.0. Uh but of course. Um and we were actually building two releases of Office in parallel. So we were all agile. We were actually building Office 95, and then what would become Office 97, um, which would be ship in the end of 1997 in November, and was the debut of the Office assistant. But it actually goes back further because the the problem we were facing was that the computers were just super hard to use.

• 4:02Speaker 0

  They they were they were filled with thousands of features. You know, for all the ease of use over the the windows, icons, menus and stuff. The the problem was you you would just stare at the computer, and there would be like 250 words staring back at you in all the menus. And and to use the product, people would cruise through like one at a time trying to understand it, and we would do these tests like let's just give Frank a copy of of Office and change all the menus to German. And they it would be just as easy to use. Right. Meaning it was equally impossible to do. Yeah. And and assuming you don't speak German. Yes. And so good assumption. And so lots of things were being tried. And one of the ones that we had tried in Office was the creation of this notion of a of a wizard.

• 4:51Speaker 0

  Not it's not an animated thing. A wizard is just literally a sequence of steps. Like you could think of it today, like when you check out of a store buying stuff on the web. It's next next next, and then you finish, and you could go backwards and things like that. But that was an invention. In fact, it it was invented for Excel and on making charts. Because making charts was this like ridiculously complicated thing where you had to like basically use essentially like a Photoshop level debugger to make your chart, and Chart Wizard all of a sudden was like, whoa, we can make charts. That was in the business world. And all of a sudden, you know, there's you know, 40, 50, 60 million computers like entering the home market worldwide, and people are like, well, in the home, whoa, it's gonna be totally different because computers at work, you have time and you're paid.

• 5:45Speaker 0

  And at home, we're gonna be doing like recipes and PTA meeting notes or whatever. Yeah. And so the whole idea was that it needed to have a uh an easy-to-use interface. Whatever that meant. Like as if the people on the business world were not sitting there the whole time trying to make it easy.

• 6:04Speaker 1

  Right.

• 6:04Speaker 0

  The problem was that like it was not easy. I mean, you have to imagine. Imagine the leading application like Facebook today. And to use Facebook, you would have to go to get to a course. Like you would go to a strip mall and spend like $200 and go to Sears and spend like $200 to learn how to use Facebook. But you would go spend like $500 to buy a word processor on your $4,000 computer and then spend $250 at Sears to go learn how to use it. So that's what the world was like. And so in the consumer world, they were really trying to figure out how do you do this. And so this is like educational software and all that. And so they actually came up with this idea of like how can we like make this product super easy to use. And so they were pioneers in in the whole motion movement to assistant.

• 6:53Speaker 0

  So for example, they went out and found some research at Stanford, uh called uh called the Media Equation by two Stanford professors, uh Nass and Reeves.

• 7:03Speaker 1

  Yep.

• 7:03Speaker 0

  Clifford. And Brian Reeves, and they were doing this amazing research that basically was pointing out that how people use technology was very much a social interaction, and that people assigned personalities and interacted with technology in ways that were surprising, but in ways that um that you could uh ultimately predict. And so for example, like when people use a computer, there's this kind of implicit social thing. Well, the computer must be smarter than me. And so then when like they they go off and try to do something and they they fail, which was most of the time, they thought it was their fault. Right. And so they had sort of sort of pioneered this notion, well, if you computer could interact with you as though it were like a smart person, but in a polite way, it would be but also people are very polite with computers.

• 7:55Speaker 0

  So like today we all say thank you to Siri and Alexa without even thinking about it. Well, they showed that they demonstrated that 25 years ago. Yeah. That it was that was what was gonna happen. And it was just sort of the way that people did that. And so they they were pioneering this whole notion of social interaction with a computer. There was a research group at in Microsoft Research that was working on uh an actual social interface, uh a project that would be like an animated character. But of course, the hardware and stuff was all like it was all like super high end and you couldn't really do it. Yeah. Um, 640 by 480. It's like you could fit only two paragraphs of Word on the screen. Right. You're not gonna take up the screen and put a face or a person.

• 8:41Speaker 0

  And so, but this uh brilliant insight in uh in the consumer software world, this uh woman named Karen Freeze, uh, who coincidentally's brother was the development manager for Word, um, but that was Microsoft back then. Like, you know the um Karen was really pushing this idea of building easy-to-use software for the home. And so she had created two products, uh Microsoft Creative Writer and Microsoft Fine Artist. And so it was basically a home kids school word processor and a home kids school paint program. Yeah. And uh introduced in that was, and it was very it was like it looked like a world. Like you would enter the world of writing, and it had like world-like toolbar buttons and the like, but it was um it was really about uh being easy to use, and one of the things that they added was a derivative of this research, and it was a character called McZee.

• 9:35Speaker 1

  And

• 9:35Speaker 0

  McZ was down in the corner helping you while you were trying to word process or paint. And what did McZee look like? Uh well, we McZee looked like a crazy weird purple people eater-ish kind of character,

• 9:49Speaker 1

  as you can see in a picture.

• 9:51Speaker 0

  Yeah. And but it was fun. I mean, it was well, and of course they had tested this and animated it, and it was it was really uh amazing in an engineer. Engineering sense to get all of this to work. Yeah. Like it was essentially like game programmers. Right. Like trying to build productivity software that was easier to use than anything. And it tested super well. And it was actually quite well for the home. And it was in that era when we thought what you used at home would just be so dramatically different than what you used at work. And so the thing was is that that McZee was just a little helper. And Karen had a broader vision for how all of oh, by the way, Karen's manager was uh Melinda French Gates, who ended up marrying Bill later on.

• 10:40Speaker 0

  That's funny. Uh Melinda was also part of this innovation. And so then the idea was like, how do we elevate this interaction? So and that like how does it how does this interaction take over the entire um the entire uh model of the UI? And so that was really the origin of the next level true assistant with Microsoft Bob. And so Microsoft Bob ended up being a whole new product based on those ideas. But Bob was where the assistant was a replacement for all of the UI. So you you don't you didn't have like the toolbars and the menus and things up up at the top like you would have thought with McZee helping, instead you had Bob like always doing everything. Yeah. So the

• 11:24Speaker 1

  idea that when you booted your PC, you'd see Bob.

• 11:27Speaker 0

  Right, Bob was a replacement for the shell basically.

• 11:32Speaker 1

  Yeah, it was

• 11:32Speaker 0

  still a program, but there wasn't really Windows 95 yet, so like you had to sort of imagine all of this stuff. And so that was really taking all this research to the next level. And and we're over in Office planning like what to do for these two releases of Office. And we have a slightly different set of problems, but we all share this like how the heck do we use stuff. Instead of having a hundred things and create a right, we have thousands. So in Office, we had solved the problem of how to use the product by having thousands of pages of documentation. So like how to remember those. They were thick documents, how to use Microsoft. Yeah, and really what they were were like actually the most inexpensive monitor stands you

• 12:12Speaker 1

  could buy. They were thick and powerful. They could support a monitor. These things were like 40 pounds of the time. Yeah,

• 12:17Speaker 0

  and so you know, uh a box of a box, because the software came in a box of Office, not only did it come with 27 floppies, yeah, but it came with you know printed documentation, like four or five inches thick. Yeah. And we had hundreds of people that would slave over this, and they were constantly like, what's the best way? Task-oriented or alphabetical or by category, trying to figure out how to map like what human beings thought their problem was, like something really simple like I want to like change the spacing between paragraphs. Where is that in the index? Right. Like, and then really sophisticated stuff, like I would like to send the same letter to a bunch of people on a list. So we call that obviously the obvious name for that is mail merge. Right, of course. Like, why wouldn't it be a good idea? Normal people would think

• 13:02Speaker 1

  about it, mail merge.

• 13:03Speaker 0

  Right. Or I want to analyze my sales data by region. Duh pivot table. Right. And so like this whole idea that you had to learn a foreign language to use the computer was really, really hard. Yeah. And we were just really struggling with that. So that solving that problem was all of a sudden a uh l uh an idea came, which was let's use natural language processing and some sort of way, magic, to go from like things people would type, like send Christmas cards to my family to letter wizard.

• 13:38Speaker 1

  Yeah.

• 13:38Speaker 0

  And that was where we tapped into another group at Microsoft Research Stanford-based, um, doing uh research on using Bayes theorem and a little bit of NLP to map like human queries to like actual topics that we had. And so it was very interesting.

• 14:01Speaker 1

  What these days we would call intent classification.

• 14:03Speaker 0

  Absolutely. And in fact, the classification was how we thought about it, and we had this all these fancy new words people were making up, like, oh, it we're gonna have metonyms and a metonym might mean um might be something like format, which then fans out to 20 different words, like bold and italic or look pretty or emphasize or things like that. And then that would go to the format font dialogue that would pop up. And so we took all of our content, and then we took a core set of people that make content for a living, and they started crafting, hand crafting this whole elaborate graph of all of the topics, and possi and and we would go through all of our product support calls and questions to try to figure out how to map that how questions that people would ask, how to map those to topics.

• 14:50Speaker 0

  That was all that resulted in this feature that we ended up staging by doing first in in Office 95 called Answer Wizard. And so Answer Wizard was this breakthrough AI, it was the first time we shipped any kind of AI thing. Um believe me, in the press, we talked all about like fancy Stanford researchers and all of this stuff. And it was very helpful. Like you would go and you would go, instead of typing in like instead of basically getting an index, you could type in like how do I format my paragraph?

• 15:21Speaker 1

  And then natural language query.

• 15:23Speaker 0

  Right. And it was the first time like anything had done that, and then we're doing it for a scale of hundreds of millions of people. So it was kind of a big, and it had to work on one of these tiny computers.

• 15:31Speaker 1

  Yep.

• 15:31Speaker 0

  Like and that was also not connected to the internet or anything.

• 15:35Speaker 1

  There's no connection. And let me remind the viewers two to four megabytes of RAM.

• 15:39Speaker 0

  Right.

• 15:39Speaker 1

  Right, two to four megabytes.

• 15:41Speaker 0

  And so actually it was really quite uh breakthrough in engineering, and it was a great collaboration with Microsoft Research and things. But we were also working on this other release in parallel to do the assistant. And so we had seen Bob. And like, but it's still early. Like Bob has has not become like the bad Bob yet in perception. Bob is still, ooh, that's kind of cool and neat. And so what we what we started doing in Office was like, well, what could we do to have an assistant in office that could help you navigate these thousands of commands? And we decided that we we could have take the NASA and Reeves point of view, which is we need to have some friendly guru to ask questions. And

• 16:23Speaker 1

  that's social.

• 16:24Speaker 0

  And that's the social part with the linguistic part of Answer Wizard.

• 16:28Speaker 1

  And

• 16:28Speaker 0

  then we added one more special ingredient, which is well, how does this guru help you to use your product? Well, what happens is you ask the guru, like, how do I reduce the space between paragraphs? And the guru is kind of leaning over your shoulder. And the guru is the a human. It's like the person down the hall who knows how to use Office. Right. And the guru is like looking over your shoulder and saying, no, not that. What you really want to do is use the format paragraph dialogue. And so we developed all of this technology, again using the Bayesian engine that we had built, which is really just like a couple pages of code now, but was huge back then. And to build out like this monitoring of what you were

• 17:08Speaker 1

  trying to do. So you were watching them type and mouse and that type of thing in Word. Right. So you had that data should be.

• 17:14Speaker 0

  Right, or in Excel or in PowerPoint. And so we were like watching the sequence of events. And so, in particular, things that would catch our attention, like try something, undo it. Try something, undo it, try something else, undo it. And we'd be like, ooh, light bulb, you know, and like let's let's figure out like what could really happen that would what we could really offer to a person as like a um an example of how to do accomplish a task. And so this monitoring became the third leg of what was going on in Clippy. Like was just and again, just to be clear, even though it nobody would care. Uh like there's no internet, we're not sending this all back to Microsoft or anything. Like this is the old days. Like this stuff, we can't store it anywhere. We have no disk space, we have no memory. Like we're we're building, and that's actually part of the challenge is we're building this statistical model offline and just applying it to all the people who use Office the same way.

• 18:07Speaker 0

  Right. That we'll come back to that in a minute. Yeah. And at the same time, we're building more and more wizards throughout the product to help you and things like that. And so that was our idea.

• 18:18Speaker 1

  Wow. That is a rich origin story. So uh it surprises me how many strands of both the university research and sort of the goals that you had and the things that you were trying to do, which is we were trying to teach people how to use the word processor kind of for the first time. Yeah, yeah. Um, and how all that led to to Clippy. So that's uh that's amazing. So maybe um talk a little bit about what was your role at Microsoft at the time, and we'll start talking about the team that sort of builds this. So

• 18:48Speaker 0

  this is like 1994. And so I basically had just done a couple of years in a staff role at the company uh after working on development tools. And so I had moved over to this newly formed group called Office. And so again, like, wow, there's a new group called Office. Well, there we had group called Word and a group called Excel and a group across the street in Silicon Valley called PowerPoint. Um and but and we were selling Office on the Mac and just started selling it really well on Windows that was really new, not Windows 95 yet. But we weren't building it. So we built a team that was called the Office team. And the Office team was chartered to do features across all of the Word Excel and PowerPoint products, and also um build like sort of shared code, because that memory constraint, like if you wanted to run Word and Excel at the same time, and they both had their own toolbar code, that was just a disaster.

• 19:43Speaker 0

  Yep. Like I'm gonna run out of memory in a second here. Right. And so my job then was I I was I was what we called the group program manager for Office. And so we had a team of about 20 program managers, which in in Valley Lingo is product manager, um, not marketing, we were the writing the spec. And the features. But the Office team overall, meaning all of Word, all of Excel, all of PowerPoint, and this Office feature-ish team were we were about, I think about 600 people in engineering overall. Big team. It was a big team, but it was a few billion dollars in revenue too, which you could divide by three to make it SaaS ARR if you wanted. But we were a big, and you know, but just for the you know, about half of our revenue was on the Mac. Oh, at that at that time,

• 20:30Speaker 1

  it

• 20:30Speaker 0

  was that much. Yeah, the Mac was an immensely successful business for Microsoft at the time.

• 20:36Speaker 1

  The first GUI, right? Right.

• 20:38Speaker 0

  And we had, you know, we were the leading products and everything. Windows, we were Windows hadn't cemented itself yet, and we were still coming from behind relative to Lotus and WordPerfect and other products. Yeah.

• 20:49Speaker 1

  That's amazing. And then uh so did the Clippy team leave live as part of the Office team separate from the Word team PowerPoint. Right. And so

• 20:59Speaker 0

  what we had, so we had about you know these uh 15 or so program managers in office, and uh guy named Sam Hobson was in charge of making uh who who actually like created the idea of like, hey, let's do this. Let's ship a suite. Let's ship. Well, no, not a suite, sorry, that say of let's do a clippy. Oh, okay. And so Sam, uh the suite, actually, I know they were listening. That's that's Mike Slade and a bunch of other people. Um but the uh the uh Sam was like, hey, let's we should do that. And he sort of led the charge on this. And we had two other program managers, one basically doing the assistant side of it, and one doing the answer wizard side of it.

• 21:37Speaker 1

  Yeah.

• 21:37Speaker 0

  And um and so along and then we had an engineering team that was about six people that were writing all the code that mapped to those those features, led by a guy named Ben Waldman. And so it was a it was a pretty exciting time, and they basically that was what they said, hey, let's just go do this. And he partnered with Karen Freeze and with Microsoft Research and both of those groups, came down, worked with NASA and Reeves, and he was sort of gluing it all together to create the the what we call the Office Assistant.

• 22:09Speaker 1

  Got it. So it's a it's a fair size team, right? So if you flash forward to today, you've got you know six engineers, you've got a product manager, you've got uh sorts of researchers behind the scenes, that's like a Series A company, right? It's

• 22:21Speaker 0

  pretty it's it's a good size for sure. For us, it was it was the we we would replicate that across the product, and that was equivalent to about you know 30 or 40 teams like that that built Office, if you think about it without the MSR people.

• 22:34Speaker 1

  Yeah. So how did it get green lit? How did you make a decision? Yes, we're gonna staff it, like Clippy is a good idea, let's go for it. Well, like was Bill G involved at all.

• 22:44Speaker 0

  Yeah, so Office is a pretty bottom-up organization at the time, and so like we're just like, sure, let's do it. And we were very we weren't we were very focused on delivering on time what we said we would deliver. But our risk profile was like this was a galactic problem to solve. This making our product because you know, imagine you shell out like $800 for Office and you can't use it. Yeah.

• 23:08Speaker 1

  Like it's kind of existential problem. After the 100 pound documentation shows up. Right. And $800 later.

• 23:16Speaker 0

  Our competition on you know, on the character mode apps, people know how to use them.

• 23:21Speaker 1

  Right.

• 23:22Speaker 0

  Like they're they're weird and archaic and arcane. Right. But

• 23:25Speaker 1

  you mastered word the word perfect function key codes, and like there's a

• 23:29Speaker 0

  template you figured out, and then so change becomes hard. Right. Now what helped us, of course, was the number of people who knew those was far smaller than the number of people getting a computer for the first time. Right. And that was and that computer was a Windows computer, and that that's how that that that sort of shift happened. And so we didn't we don't have like a an approval process per se. We have a consensus-based approach to to building the product. And and so we just went forward. Now that doesn't preclude you from having to like show up at Bill's office and tell them about what you're doing. And so very early on. And Bill loves love the answer wizard and natural language and the work we did with Microsoft research. And so we thought thought, well, this will be great. Yeah. We'll just show him this next level of that.

• 24:16Speaker 0

  And but Bob has starting to become more Bob, like the punchline. The perception of Bob was like, oh, this is not how I want to interact with my computer. This is great for the home. And so the worry, and so Bill's immediate worry was like, how is this going to apply to the office? Now it's not like we weren't thinking about that. Like we that was our lives. And so we were very focused on our thing was the assistant was a helper. And it wasn't gonna be like a replacement for the user interface of the product, it was gonna be a helper. And it would be inobtrusive and also fun. That was sort of part of the goal. And so we uh we show up in Bill's office and we're gonna show him a demo of it, and uh a prototype of it. And uh Bill, you know, this is like a bill review, like these things that you used to hear about, like, oh, that's brain dead, or like that's the stupidest thing I ever heard, or I could do that in a weekend in basic and 4K, or you know, that and like they really weren't like that.

• 25:12Speaker 0

  I literally just spent two years sitting in bill reviews, and so they really weren't like that. There were moments. So we were about to have our moment. And so we're showing them a bunch of plans. These are all plans. There's no code or anything like that. And Bill sort of does this the best way to describe it is like a routine where he's he's like, so say I want to make a chart. So then like the the clown pops up. The clown and the clown. He literally

• 25:37Speaker 1

  says the clown.

• 25:38Speaker 0

  Well, we didn't have a name or anything, and and and McZ kind of clown like. Right, right. You know, so the clown pops up, and like I'm like, I want to make a chart. Next, the next, the next, and make like and then pretty soon you just want to kill the freaking clown. And

• 25:55Speaker 1

  it's a profit, because uh ultimately. And

• 25:58Speaker 0

  each step we're like, no, it's not like that, or we have a different idea. But you know, you're you're sort of having this discussion. And and but we walked away, like, well, we're still gonna do it. And but we did have the uh official Hungarian coding name for the assistant became TFC for the fricking clown.

• 26:16Speaker 1

  Uh that was the data type. Now, for the record, Bill did not say fricking. Uh I'll just leave the rest to make sure we don't get censored on YouTube.

• 26:24Speaker 0

  Yeah,

• 26:25Speaker 1

  awesome.

• 26:25Speaker 0

  And so um so we actually just start building it.

• 26:29Speaker 1

  Huh. And even though Bill G kind of made fun of it. Um the team just thought, like, look, this is a good idea. We're just gonna keep on doing it.

• 26:37Speaker 0

  Well, it it was easy to make fun of. Like there's no doubt. Like you're imagine this world. Like the our competitors are all like, they look like airline reservation systems today. Like they're character mode, 3270 terminals. We are swimming in a sea of like menu and toolbar buttons and things. Something had to give. Like we we were building software that people couldn't use. Right. Like there was just no, we had to try something.

• 27:03Speaker 1

  Yeah. And the did the team think uh maybe our documentation can be less complex or like how did you decide?

• 27:11Speaker 0

  That it turns out that's a really when you make a bold bet, like there's always someone who was gonna say, I could have fixed it, just give me more time. And and in particular, like, you know, there were people who thought, like, if we just had more buttons, there were people who thought if we just did more wizards, there were people who thought if we used more color or made the buttons bigger. I actually have a whole series of screenshots that I used to do at college recruiting, like answers to ease of use were all way more words. Right. And and like it turns out like you're just the it's the interface from the human brain that is basically sitting at PowerPoint thinking, I need to get a promotion.

• 27:48Speaker 1

  Right. Right. Something abstract, high-level.

• 27:51Speaker 0

  Which doesn't really get to like squeeze more bullets on the slide without screwing around with the font size too much. Like the gap is infinite. Yes. And it gets worse like in Excel, like make me rich. Right. Or Word or Word, where you're just staring at a blinking cursor. Right. I don't know how to start.

• 28:10Speaker 1

  Right. Great American novel is on the other side of this, but like how do I start? Yeah.

• 28:15Speaker 0

  And so we, you know, our and Office's risk profile for our time there was always fairly high, even uh given our our position in the market, because like there's so much people want to get done. Yeah.

• 28:30Speaker 1

  That's awesome. Um so now let's talk about how it got built, right? So you had these research strands, you definitely wanted to make it social, you had the Bayesian uh sort of classifier to try to figure out what were you trying to do in the first place. So now we're starting to write code, we're starting to write demos. So yeah,

• 28:48Speaker 0

  so very early on, uh Sam and team had built this uh this prototype uh uh called Pogo. And so Pogo is like a it's a VB program in Excel 5, uh which is the 16-bit version of Excel that existed at the time they were shipping. And and we would go do usability tests, like behind one-way glass, and we'd give a person a task, and and and so it was a fascinating, like we have thousands of hours of these tests. And so like I one I used to show on college recruiting is a person trying to do like grid formatting in Excel. So long before today you look at templates that say I want color bands or vertical bands and click a few things. Like you just basically stared at the screen trying to figure out when I print it, the boxes don't show up. Right. But but I want the boxes to show up.

• 29:35Speaker 0

  Or when I print it, the boxes do show up, and I don't want them to show up. And then like the only way to do it was to bring up like this low level like format debugger for for edges.

• 29:46Speaker 1

  Yeah.

• 29:47Speaker 0

  And uh so we would so we actually had like many people go and use Pogo to try to figure out if it worked. And like the response was crazy positive. People would type in like, how do I fix the borders? Yeah. And then Like watch it, and like we had animations for thinking. Like one test, like the guy's like, ooh,

• 30:04Speaker 6

  he's thinking, he's thinking, sweet, you

• 30:07Speaker 0

  know. And like, and then we after the test, we would ask people, like, do you like getting help this way? Because remember, help other ways was just an index, a very long or a full text search on all those German words you don't understand. Right. And like people would resoundingly say, like, oh, I like getting help this way. It's much more friendly and it's much better. And that would constantly happen to us. And even Pogo

• 30:32Speaker 1

  the prototype was animated, like that seems like a pretty

• 30:35Speaker 0

  It was well, yeah, it was

• 30:37Speaker 1

  hard to implement. I mean, we're in 640 by 480. Yeah, yeah.

• 30:42Speaker 0

  But we could see it. And you can see for yourself in this footage.

• 30:51Speaker 5

  I want to hear.

• 31:08Speaker 5

  He's thinking. Sweet.

• 31:17Speaker 5

  Well, go. Alright, based on the request. I've seen that before. And again, slight the Mies. They've come up with a lot of the same things. But that's definitely what I want. So like the sales you want formatted. I want to do the whole smear. Well. So Pogo here is leading me right to what I want. This is pretty quick.

• 31:49Speaker 7

  How likeable is he? Well, he's cute without being stupid. So I mean, there's a fine line. I like the way his eyes looked. I think he's fine. How much did you like getting help from him? I liked getting help from him much better than online help.

• 32:12Speaker 0

  I mean, it was a VPA prototype on a platform that barely worked. So we we had a lot of really positive early signs for how to go build it. So one of the more interesting things was just literally the animation and the stage and where we would be. So today, when you see uh clip it, meaning then, um, you know,

• 32:36Speaker 1

  he's in this box. Sort of lower right corner of this like there's a paper clip sitting on top of what looks like a stage. Right. Yeah.

• 32:43Speaker 0

  And so what the problems we had was like on these computers, you couldn't do animations with no frame. Like it had to be a win, it had to be a Windows H-wind to really do this, especially because we wanted to share the code across the products and all this other confusing stuff. And so we were constrained by this box. It was it was really painful because we really wanted to do more with it. Plus, we we could only have like you know 30 different little animations, like types of things, like here's waving or smiling or looking thinking or whatever. And each one can only have a few frames because we we had to fit on the floppies.

• 33:24Speaker 1

  Oh, that's right. And you couldn't buy it on a hard drive, you're not downloading it from the internet. Right, and so

• 33:29Speaker 0

  we had a lot of engineering constraints. This the stage one really bothered us. So one amazing, amazing opportunity that uh between uh the consumer bob team and and Sam that we had was to speak with um the these animators from Walt Disney named Frank and Ollie.

• 33:47Speaker 1

  And

• 33:47Speaker 0

  so it's two guys that were friends from like 1906 and lived to be like 95 years old. Yeah. And uh we got to speak with it was like this incredible moment. Like I grew up in Orlando, so seeing the Disney animators was just And these are the legendary. Yeah, these are not like animators.

• 34:03Speaker 1

  Two dudes that painted.

• 34:04Speaker 0

  These

• 34:04Speaker 1

  are like Fantasia and Pinocchio

• 34:06Speaker 0

  and

• 34:07Speaker 1

  Snow White, like

• 34:08Speaker 0

  these guys were the creators, yeah, and they have they actually had this very famous um essay about the sort of these rules of animation

• 34:17Speaker 1

  about

• 34:17Speaker 0

  exiting and entering a scene and how bouncing should work and all this. And they just loved the assistant. And when and I remember talking to them, because our we were rooted in knee deep in this box, and we were like, how are we gonna get out of the box and all this stuff? And we just Ben said we just couldn't do it. And the problem was Bob wasn't in a box, and so we felt like wow, we gotta be out of the box. And Frank and Ollie were like, you want the box. And we're like, we do. Yeah. They're like, because every character needs a stage. And and this allows the character to exit and enter with grace, because otherwise it would be like an alien and it would just disappear. Right. Right, right. The window's gone. What happened? Right. And and then, but then they said the thing that I'll never forget.

• 35:02Speaker 0

  He he's like, what you really want though, is like it needs a helper. And we're like, what? The helper needs a helper? And there's gonna be an infinite loop here. Right, but then think about Jiminy Cricket and think about all the animations. And every of every character in Disney has like a little helper. So we're like, you should have like a little thumbtack that helps the paper clip out. And like is like a foil for and we're like, wow. And we're like, how are we gonna fit that? You're right. Right.

• 35:31Speaker 1

  640 by 480, and then we're in this tiny little corner.

• 35:34Speaker 0

  And so like that, it was very, but it was just like it was such a relief in a sense to hear that like that constraint will actually help animations and animators to uh to get their job done. And so we're just busy building this thing. Yeah. And making it work and tying things together.

• 35:54Speaker 1

  And then get from pogo to the paper clip. Like, how did you guys decide the paper clip? And could you I don't remember actually, could you choose multiple assistants? Well, that was one of the big advances that Sam came up

• 36:08Speaker 0

  with was this ability to pick your assistant. So we all remember Clip It. Yeah. Um but we it was one of the more elaborate parts of it. And as Sam's manager, it was one of the more parts where we're like, wow, that's a lot of money we were spending. And so Sam basically um we went out all around the world to test different characters. Like, what should be the character for Microsoft Office? What would represent Office the most? And of course the answer is you can't have one. Right. But Nass and Reeves had had really told us, like, yeah, you know, it we'll be able to predict, but it's not it's gonna be seem sort of random to you, and it's just gonna be an emotional response. And when you think about it, of course, like you know, Hello Kitty was like 25 years old then, and like universal appeal. You know, I remember I was in France, and the bank, uh, Credit Leon had had like a tiger-ish kind of the Tour de France

• 36:59Speaker 1

  character. And

• 37:00Speaker 0

  like nobody cared. Like it was their mascot, their logo. And people have this, they have they develop an affinity for a wide variety of ways. But Sam went all around the world, thousands of renderings of assistance, and literally you get like shopping malls and and uh usability tests and surveys and all sorts of stuff. Like the things you can't you could easily do on the internet that you had to do in person in multiple countries, you know, all sorts of stuff. And it just came back with like we could do whatever we thought was cool. We tested also we tested like lifelike or purely uh mythical or real life objects and all sorts of stuff. And they all had their followers. Like there was no obvious answer. It

• 37:44Speaker 1

  had to be a paper clip, or it had

• 37:45Speaker 0

  to be a

• 37:45Speaker 1

  tiger, or

• 37:46Speaker 0

  and so one time we actually showed up uh with uh Mike Maples, uh Mike Maples Sr. For those of you that are Mike Maples Jr. So Mike's father was the sort of the godfather of Office, and most of the success of Office I think is attributed to the structure and the processes and the bets that Mike made uh as a leader is just phenomenal person that I was so lucky to work with. Um so Mike is this native Oklahoma, worked for IBM for many years before he came to Microsoft, and we're all like we're gonna show this to Mike, and we wonder what he's gonna pick. You know, Mike, he's an

• 38:18Speaker 1

  outdoorsman, he has a big ranch in Texas. So maybe the the horse. Yeah, like the

• 38:22Speaker 0

  something like

• 38:23Speaker 1

  if we had a Jeep or whatever we thought Mike

• 38:26Speaker 0

  would pick. And

• 38:26Speaker 1

  so

• 38:26Speaker 0

  we show him this whole wall of pictures. Yeah. And he takes like five seconds and picks like this fluffy bunny. This like fluffy pink bunny thing. And we're like huh. Why do you like that, Mike? And we're all like sociologist, anthropologists. Like, we're trying to understand what is Mike doing. Is he playing with us or what? And and he's like, well, I see a lot of rabbits on my ranch. Reminds me of my ranch. Oh, nice. And that actually was like how people picked them. So we we came up with a cast. Now we were constrained in how many we could ship. Because um, like you we floppy space and stuff and the animations, and so we had uh a core set of characters. Um we had clip it, we had this dot that was sort of uh derivative of some of the work in Bob. Uh we had uh a butler named Bosgrove. Um we had a puppy and a kitten and all and they all had names that I'm spacing on right now.

• 39:16Speaker 0

  Yeah. We could click through and I could show you the gallery that you could pick. And it became a thing. We were certain that this level, like this is a kind of customization. Like, it's this is like dark mode level customization. Like people were gonna get excited about this. Yeah. Um and uh we had a a smart uh genius like character that kind of looked a lot like Albert Einstein. Um and then we got a C Son's assist letter from the estate of Albert Einstein. So we had to looks too much like Albert.

• 39:48Speaker 0

  Um but it became very interesting as we got closer to shipping. Uh we got a very long uh email from our friends in the Japan subsidiary. So Japan, it was more. More than half of the office business. Like it was an incredible part. It's a whole story about the business model and how office rolled out in Japan, some really smart decisions from people. But they were like, yeah, this is this is somewhat disagreeable to us. In that very

• 40:17Speaker 1

  polite

• 40:17Speaker 0

  Japanese. It was one of these official, it was translated from English English and Japanese at both. You got the memo. We got a memo. You got the memo. It was it was tough. And these are all good friends of ours. This is a very tight I you know, been to Japan many times. Like I like I'm still even to this day, Facebook friends with this whole group. We talk about stuff all the time. And they went on to explain that the problem is that the characters have four fingers.

• 40:40Speaker 1

  Yeah, Clippy had all the characters. All the characters.

• 40:43Speaker 0

  Of the characters the fingers, they had four. And the dot, you know, everything. Like Bosgrove. And like that's just not really not the best approach for us.

• 40:55Speaker 1

  And

• 40:55Speaker 0

  and why? Well, there's a lot of lore about this. In fact, there's like a whole Wikipedia article these days about like why it's four fingers. I all of them it it's a Japan. It's a Japan cultural thing. And it's a cultural thing, and you just respect it. Um it turns out it's the Simpsons had this issue, like tons of things have these issues.

• 41:15Speaker 1

  The four versus five. Five. Because

• 41:16Speaker 0

  animators, five fingers is it's a lot of drawing. And so they they actually don't like five. Right. And three looks like an alien.

• 41:24Speaker 1

  Right.

• 41:24Speaker 0

  So they don't like three, so they've historically animators had used four. Mickey has four.

• 41:29Speaker 1

  And

• 41:29Speaker 0

  that was sort of our defense. And they're like, but you're not Mickey.

• 41:33Speaker 1

  Once you're famous, you can have as many fingers as you want.

• 41:36Speaker 0

  So we negotiated a compromise, a win-win sort of situation where we would offer remove a couple of assistants and then offer some ones we specially made for Japan. So we made this lovely dolphin that swam. And then we made one that made us uncomfortable at the time, which was office sensi, which was it translated loosely as as office lady.

• 41:59Speaker 5

  And

• 41:59Speaker 0

  if you know, when you visited an office building in the 90s in Japan, you were greeted by an office lady.

• 42:05Speaker 5

  Yep.

• 42:06Speaker 0

  And that's just how everything was. That was business. And we were like, wow, this is not what we would like as a brand. Yeah. And so you get this complexity of developing a worldwide product. I mean, we were shipping in 40 odd languages, 150 countries. You know, how do you balance all of this? You we certainly don't want to offend. Yeah. But is it offensive if the assistant is available? And so we were like, oh, maybe we should write code that the assistant won't work in the English and Western versions of the code.

• 42:34Speaker 1

  Interesting, right? Where it would be perceived as offensive.

• 42:36Speaker 0

  Right. And and the thing was it's like they're like, it's not offensive.

• 42:39Speaker 1

  And we're like,

• 42:39Speaker 0

  but it's offensive to us. And we're like, where in Japan? And we're we're in this loop of like how to- and so we ended up shipping it. And the truth is uh it sort of gets into the after the after assistant but like it wasn't offensive, and it's it's kind of you really have to go, well, is it really offensive ever? Like it it was their culture, and you have to respect it. Yep.

• 43:04Speaker 1

  Well, this is one of the sort of deep design questions there is, right? Which is empathy for the user versus I know what's best for the user, right? And this is the age-old tension.

• 43:12Speaker 0

  Right. And you you overlay like this is a really important business decision. Right. Like, and if the subsidiary is not behind the feature or the product, a whole other set of things happen. You know, it's it's very complex. It was fine. Like it rolled out fine, it was all good.

• 43:29Speaker 1

  Yeah. So it's super interesting. You have this cast of characters, dolphins and office ladies and clippy, and it's clippy that sort of achieved the cultural resonance, right? Yeah, clippy keyboard. Yeah, I'm

• 43:39Speaker 0

  surprised because we did like a special one for for reviewers. We did a DOS prompt, um like where um it was just like a C with a poll and a backstory. And that's funny. And the

• 43:50Speaker 1

  reviewers love that. Yeah, you know that's funny.

• 43:52Speaker 0

  I I actually did one custom, we had a kit you could build your own, which all the corporate customers wanted. We had a kit, and I built um one. I how to give a talk at the board of directors at Big G serial, like Lucky Charms and stuff. And I uh I built an assistant made out of Lucky Charms guy, which was a huge copyright violation. Right, right. But but I just the leprechaun? Yeah, the leprechaun. And I just built like a really quick static assistant. And I showed it, and basically I didn't s make a big deal. I just gave my I just opened PowerPoint, started doing my slides, and the leprechaun pops up. And the and the left car pops up, and they were like, what is that? And they were so excited and wait until they saw it, because I actually animated in my slides. Oh, not like I had all of the honeycomb carrio thing animated, but but we had it was it was there was it was it's fun like we had intended it to be fun.

• 44:40Speaker 1

  Yeah. Maybe this is the best time to segue to talking about sort of the launch and the sort of, you know, you you mentioned a reviewer's kit, which some of the viewers might not even remember or know what a reviewer's kit is. So uh before we get there, anything else you want to tell about the story of building it, uh otherwise we'll segue to the launch and the marketing.

• 45:00Speaker 0

  The thing that's interesting about building a feature like this is when you're when you're taking an uncomfortable risk, it's uncomfortable. And what'll happen is, you know, someone is gonna write a long piece of mail that just says this is the dumbest feature I've ever seen, and here's 87 reasons why. Yeah. And it's very, very hard because like one of my favorite sayings, sort of riffing off Buckaro Bonzai is like no, no matter what happens, someone always said it would. And so but you have to have a risk profile and a set of data and uh and a belief that what you're doing makes some sense. And that was really what we faced, because look, this was a crazy idea when you think about it. Like, let's I know, let's make formatting a paragraph have like a dancing paper clip.

• 45:45Speaker 1

  Right.

• 45:46Speaker 0

  And so we went with it even after we had all sorts of meetings and people would concern. The one thing we did, which was actually really for performance and disk space and stuff, was we added a switch so that you could always turn it off. And that turned out obviously to be a savior. Um but it was very, very hard on the team. Right. I've slaved on this thing, and you're gonna hide Clippy forever. Right. And and I mean, not just hide them, like just delete them. Yeah. You know, like and it became a thing. You could install Office without Clippy even being installed, and admins would do that. And so but this notion that like someone always said it would, is very interesting because that describes the 10 years after Clippy from like when he shipped until he became a cult figure.

• 46:35Speaker 1

  Was

• 46:35Speaker 0

  like all of a sudden everybody who didn't like it was like, I told you all along we weren't gonna like it. And that's like, boy, it it didn't hurt. Actually, I kept giving talks for years about and Clippy was always like first people would laugh nervously, a few years later, they're like, well, that's kind of interesting, and then it became a cult thing. But but like people like, oh, I said it wasn't gonna work in in Europe, or it wasn't gonna work for women, or it wasn't gonna work. And it worked is the same everywhere. We just made a few mistakes, which we'll talk about in a second. But like the reviews, you know, Steve Wallstrom from Business Week was like, I was skeptical, but I really liked it a lot. And uh Peter Coffey at PC Week. Those would be the two mainstream and tech reviews. Peter Coffee was like, this is a great idea. Like this makes it easier to use.

• 47:21Speaker 0

  Like so the reviews were like that was good. And there were plenty of reviews that were like, that was the stupidest thing I ever saw. But we got that for everything. Yeah.

• 47:30Speaker 1

  All right. So let's transition now and start talking about the launch and the public reception and sort of the iterations of it. So maybe let's talk about the launch. How did you guys decide to launch? Was this going to be a big part of the office launch, a small part?

• 47:44Speaker 0

  Well, there it was fascinating. First, you know, when you launch back then when you would launch a product, it was this elaborate process because it was all you know, the way that you think about launching movies in a cinema. Like you have to line up distribution, you have to line up the press, you know, you're dealing with print magazines that have three, six-month lead times. And so you you start the communication and everything way, way earlier. Like there's no surprising people. Um that's also what made this one a little bit more difficult because any skepticism, you now had six months of sitting there simmering. Right. Of like, oh God, they were skeptical. Which is not we were all used to because you know, you could show up and do a demo and the demos would all crash, or it would take more memory than people thought, or they just weren't impressed.

• 48:37Speaker 0

  And then you had six months of like, no, don't worry, we'll fix it. Right. Um but this one, because it was already, you know, this semi-emotional feature across the team. Yeah. Highly risky. Highly risky. It was already you know, we were already set for like, you know, how is this stuff gonna how is this stuff gonna work? And so what we did is we we put together this reviewers kit, which is like basically a bunch of documents that you send to the reviewers to read before they meet with you for the you know, six times and and talk about it. And we uh we had lots of discussions about about the technology, about Microsoft research, about NASA and Reeves, about all of these parts that that contributed m artificial intelligence and all that.

• 49:22Speaker 0

  Um and then we went, it was always about the cast, like of the number of they always wanted to know about that. They always want to know if you could make custom ones. It just it it was magical that way. So there was this engagement. And then the problem we had that really started it all was uh we went too far. And what we went too far was on what became the meme, which is it looks like you're trying to write a letter. And so it I did this demo 50 times for reporters, and the problem we had was the words. Team was busy building more wizards, just like the Excel team was and the PowerPoint team. That was what the teams were doing. And what we thought we would do with the assistant is why don't we just wire up the activities you're doing to the right wizard?

• 50:09Speaker 0

  Like it makes total sense. Because otherwise, how would you find the wizard? It's just more of the text in the menus. And so one of the wizards we made in Word was called the letter wizard. Now, most of you kids out there are thinking, why do you need a wizard to help you write a letter? Well, because back then a letter would be like a cover letter for a job or a cover letter for you know why your rent is late or whatever. And it involved you writing this letter in a format that you learned in grade school and then printing an envelope with the return address in your laser printer, which was like you like a

• 50:47Speaker 1

  to

• 50:47Speaker 0

  this

• 50:48Speaker 1

  day that remains a pretty hard task.

• 50:50Speaker 0

  It's like a

• 50:50Speaker 1

  fields metal level task. Like you, you know, you try to get that centered on the envelope. Well, you have

• 50:56Speaker 0

  to get it centered, you have to get it, um, you have to get it oriented right. Like so you would usually print it on the back or upside down. It was a nightmare. So we had and like if you look at the letter, it's like really awesome. Like it had like pictures of the envelopes, it would know about your printer, which was a giant database because there's no internet and all this stuff. And so we thought this is such a good wizard that when a person starts to write a letter, we should help them write a letter. But but it turns out most people weren't using gonna use the wizard because most people weren't gonna do all those fancy things. They were just gonna print out the letter and write the address and stuff it in ML. Can't write that and be done. And most letters were for one person, they weren't merged to 50 different recipients and all this other stuff.

• 51:43Speaker 0

  And so the little engine we had that was monitoring your activities, looked for a lot of things. Like, oh, if you typed an asterisk at the beginning of a line and space and typed and pressed return, it would turn into a bullets, like we see all the time in every product today. But we had the assistant, hey, we just did that for you, because it would look like a devil had taken over your computer and stuff. And so then you typed in dear mom and press return. Hey, looks like you're trying to write a letter. Because you're trying to write a letter. Like it's not ambiguous. This is not even like some weird AI guess. Yeah. Like if you type dear Mr. Smith, you're like God's are you're writing a letter. Writing a letter. Our problem was that we offered Clippy, or whoever you chose as your assistant or Bosgor of whatever, were like offering the wrong level of assistance.

• 52:30Speaker 0

  We were caught up in our own features at the sort of the debugger level. And what people really wanted, like had we thought about it this way was if they if we just had given them the letter.

• 52:41Speaker 5

  Like

• 52:42Speaker 0

  if we would have just had all of the contents of the letter to choose from. Like imagine a gallery that's like this is a job application, this is a cover letter, this is your rent is overdue, this is I owe you money. Like then, and it turns out all of that was where we were heading. Right. Like that was our. You would

• 52:59Speaker 1

  eventually add templates and

• 53:00Speaker 0

  but like not with 27 floppies and not with no internet and so it it the demo itself became the meme. Like we were we were responsible for that. Like that was the demo I did for like Byte magazine and PC magazine and all of that. Um it was cute. Yeah. It and and at the time, letter wizard was like super helpful.

• 53:24Speaker 1

  Yeah.

• 53:25Speaker 0

  But like to a normal person in an office,

• 53:28Speaker 1

  like

• 53:28Speaker 0

  that probably had like a Dymo label maker. Yeah that was gonna print out the address that way, it was it was kind of bizarre.

• 53:36Speaker 1

  So so interesting. Like what you thought was the feature turned out to be a bug, right? Because this is where all the sort of memes came making fun of um clippy. But you were also saying, look, the the initial reviewers, and at the time they're basically like two or three viewers that mattered, right? Because everybody read about their software in print magazines, they were actually pretty positive.

• 53:56Speaker 0

  I would say, I mean, I wouldn't go overboard. Like they were all intrigued. Yeah. And because it was, you know, frankly, a big important product, they weren't the the broader the reach the writers were, the the less likely they were to just go, it's annoying and useless. They would all say, well, I'm intrigued. And there was a little warning in there, which is in any feature that you do, when someone says, I don't like it, but it's good for like less skilled people.

• 54:23Speaker 1

  Yeah,

• 54:23Speaker 0

  there's a warning in that. And the warning it could go either way, which makes it even more complex. Because the warning could be, okay, well, it's not good for those people. This person is just doesn't like it and they're trying to be polite. Right. Or the the warning could be the tech enthusiast wants more tech enthusiast stuff and doesn't care. Yeah. And like you might be on to something, but they're just gonna they're sort of distancing themselves from it. And that's sort of this core problem in that the reviewers that we're talking to are all for the most part, not all of them or some, you know, amazing people that obviously aren't um are very tech enthusiasts, and that was the focus of our own efforts at the time.

• 55:00Speaker 1

  This is like uh for the car reviewers, automatic transmissions. Exactly. For the people who can't row their own, yes, there's this automatic transmission thing. But like why would you want that?

• 55:09Speaker 0

  Yeah, or even like today, like something like dark mode, which is like the tech enthusiasts love it, and everybody else is like, I can't read it. Right? Too

• 55:16Speaker 1

  dark. Um

• 55:17Speaker 0

  so the we the reception ended up spiraling and like not getting better, which was a bummer. But you know, like we lived, and we so what we ended up doing was the next release of Office, which became Office 2000, we sort of demoted Clippy and and reduced his interactions or its interactions or her interactions um some.

• 55:41Speaker 1

  And then what was the decision making that led to that? You could have just pulled it. Yeah. So how did them

• 55:48Speaker 0

  the truth is it's just it wasn't it wasn't, it was only negative in a certain way with a certain set of people. And like I remember I was on the press tour and I um for Office 2000, and I was at um at uh National Public Radio in Manhattan. I remember it vividly, and I walked in, and like the receptionist has Office 97 and is using it. And like I watched the receptionist type in a help question and kind of giggle.

• 56:15Speaker 1

  That's awesome.

• 56:16Speaker 0

  And you know, and here I

• 56:17Speaker 1

  was, I was about to walk. We're on the press tour, right? To go tell

• 56:20Speaker 0

  everybody how we toned it down. Right.

• 56:23Speaker 1

  And there was a victory in the world. Yeah, it was

• 56:25Speaker 0

  it was it was super interesting. I you know, but they were, you know, and the weird part is of course there were dozens of features where we had to backtrack or redo. I mean, we wrote mail merge four different times in the history of office. Like people couldn't use pivot tables, then they became essential, and you know, it all these things all have cycles. And so part of me was like, well, maybe this is a cycle. But we backed up a little bit, and then in Office XP, we were like, this we we did a bunch of UI work uh with it with a thing like more like Photoshop, these sort of task pane interface, um, which was probably not a good idea. Um but we did it, and so that gave us an opportunity to say, look, we did all this other ease of use work. Right. So we're gonna retire Clippy.

• 57:07Speaker 1

  Got it. And then it turns out that you guys threw Clippy an actual retirement party. Yeah, we

• 57:13Speaker 0

  uh like Lisa Gurry on the office marketing team, like had this like, let's really go crazy. Yeah. And like actually like have a retirement party. And so we went and we flew down to San Francisco, we met with the press to tell them this, and then at one point we just had somebody uh dress up as Clippy, and we wrote around on the on the train and

• 57:36Speaker 5

  he

• 57:36Speaker 0

  held signs up, like we'll work for food. And we we then did a bunch of animated shorts because the web was new. Right. And like we do a bunch of flash animation shorts about Clippy's retirement.

• 57:57Speaker 3

  Why hello there? I'm leveraging real-time legacy compliant collaborative microbranding to complete the most important project in my company's history. As soon as I finish this proposal, I

• 58:08Speaker 2

  it looks like you're writing a letter. Do you like help? You. You little metallic. You little metallic hmm. Try rephrasing a query. For example, print multiple copies of a file will lead to

• 58:24Speaker 3

  Stop. Listen. Next to Microsoft Bob, you are the most annoying thing in computer history. You know, Bob, he's a friend of mine. Get lost. Didn't you hear? Office XP has put your skinny metal butt in the digital dustbin. Just look.

• 58:53Speaker 2

  X XP.

• 58:57Speaker 3

  Isn't there any way to

• 58:58Speaker 2

  turn that neuronic paper clip off?

• 59:01Speaker 3

  Memo to build games. If that paper clip pops up one more time, I

• 59:06Speaker 2

  can't

• 59:07Speaker 3

  take that thing anymore. I quit.

• 59:14Speaker 3

  Your toast pal. I'm gonna need you to back out of the cubicle.

• 59:18Speaker 2

  Oh, really? Office XP, eh? Well, we'll just see about this. Hey! You

• 59:25Speaker 3

  it looks like you're writing a letter! When'd you like

• 59:28Speaker 2

  beat it?

• 1:00:01Speaker 2

  Hello. Tech test. Yes! It looks like you're using Office XP. What? Who is this? Don't you miss having a talking paper clip to guide you through common tasks? No! Ow!

• 1:00:15Speaker 3

  So did they upgrade your computer to Office XP? No, not yet. Huh?

• 1:00:22Speaker 2

  I didn't say that. Hey, nice hat.

• 1:00:30Speaker 3

  Hey, I've got that Office XP now. Oh, yeah.

• 1:00:33Speaker 2

  What do you think? Because I hate

• 1:00:36Speaker 3

  Say what? Hey, when did you get glasses? Oh, hold still. Hold very still.

• 1:00:45Speaker 2

  Ow! Here

• 1:00:48Speaker 3

  we're

• 1:00:48Speaker 2

  upgrading

• 1:00:49Speaker 3

  to that new Office XP. Yes. I'm setting you up next.

• 1:00:52Speaker 2

  Office XP stinks. But

• 1:00:55Speaker 3

  I hear Office XP stinks.

• 1:00:57Speaker 2

  Really? Who

• 1:00:57Speaker 6

  says that? Listen, I'm a tech guy. I make a lot more than this plant.

• 1:01:03Speaker 2

  Office XP

• 1:01:04Speaker 6

  is a great upgrade. And it doesn't have that paper clip.

• 1:01:07Speaker 2

  I kind of like that paper clip.

• 1:01:10Speaker 4

  But I kind of like that. No, I don't. I I hate that thing. Plant, you're fired.

• 1:01:17Speaker 0

  We recorded a song, like the Clippy Blues song. That is awesome. We're

• 1:01:22Speaker 1

  gonna r feed it right into this video so everybody can listen to the blue song. You

• 1:01:26Speaker 0

  don't have any need. There's no work for me because you have Office XP. I mean, like all sorts. Gilbert Gottfried, the comedian was the voice.

• 1:01:34Speaker 2

  I will now perform a traditional blues lament.

• 1:01:44Speaker 4

  It looks like you're writing a letter. Would you like some help along the way? Yeah, it looks like you're opening a spread. She's now mom. That's the kind of thing I used to say. I told you how to tempt the cops where she dated to a database. You just yelled at me to go away. It looks like you're writing a letter. I do believe I could format it better, but since you've upgraded 2XP, looks like you got no use from me. I beating my case to Microsoft. I told him I make PowerPoint

• 1:02:24Speaker 2

  hip.

• 1:02:27Speaker 4

  Yeah, I believe it's my case to Microsoft. Told him they use this thing, I'm a trick. But then told me a little company secret. XP stands for X Paperclip. That ain't right.

• 1:02:50Speaker 0

  And when we did a press release, like announcing his retirement. Yeah. And then Amazon was kind of new to a lot of people still. So I got dozens of copies of gift books from Amazon of this like thousand and one uses for a paper clip. Oh,

• 1:03:04Speaker 1

  that's hilarious.

• 1:03:04Speaker 0

  And it was like paperclip bending art or something, and and like, you know, it was basically art with office supplies.

• 1:03:11Speaker 1

  Yeah.

• 1:03:11Speaker 0

  But there was a big paper clip on the cover, and it was clearly the search results for paperclip at the time. And so I have like a whole shelf of this book.

• 1:03:18Speaker 1

  Yeah. That

• 1:03:19Speaker 0

  is awesome.

• 1:03:20Speaker 1

  Now, at the time was this was Clippy already a cultural meme, or did the retirement party become part of that? And like I actually,

• 1:03:29Speaker 0

  I think it was sort of a punchline for like four years or five years from like 1997 to 2002 when we retired him. Um and I think I honestly believe in hindsight that the sort of the boldness of out front retiring him as opposed to like just secretly cutting him and pretending like he never existed. Yeah. Yeah. Or it never existed. I keep saying he, but there's no gender or no gender for clipping. Um clip it. Uh I think that contributed to the sort of the the burgeoning cult status. Yeah. And it was really funny because you know, I you know, a few years later, after Office 2007, I would move to Windows. I kept every year I kept doing college recruiting, I'd come down to Stanford, I'd go to the East Coast, Southwest, all over and I'd all I had I either get asked about Clippy,

• 1:04:18Speaker 1

  yeah,

• 1:04:18Speaker 0

  or um, or I would just bring them up. Right. And and like it was constantly a uh uh evolution from like joke, and then like what would happen to like like feeling bad for me to joke. And then what would happen would be like students would bring up Clippy, and now we're like it's you know, 2010, and I'm like, how do you know about Clippy? They

• 1:04:44Speaker 1

  probably never used it or seen it. And

• 1:04:46Speaker 0

  then they would sort of well, my dad had Windows 95 when I was like 15 years old. I'm like, no, but you were like eight years old. Like, I'm so confused. Right. And and that's when I sort of started to realize that like it it had become like some weird cultural icon of sorts. And and I remember I one time I was in a hotel room during the evolution of the downsizing of Clippy, and like a incident, I mean uh um a moment in the Microsoft versus the United States trial happened. And like I'm watching Conan like everybody my age did in a hotel room, and like I hear Conan, like, well, Bill, you should really relax.

• 1:05:27Speaker 6

  Now come on, Bill, you really gotta buck up. You're actually getting off easy compared to what the government did to Clippy, that annoying paperclip icon that pops up on your Microsoft Word software all the time. This is true. Their government's been pretty tough on Clippy. Uh check out this uh chilling footage of what the government did to Clippy earlier today.

• 1:05:57Speaker 6

  That's not right.

• 1:06:00Speaker 1

  So he appears on the late night.

• 1:06:02Speaker 0

  It's on like late night, and I'm like mortified. And so I immediately and we quickly go and and like get see if we can show this in team meetings, and because it was

• 1:06:12Speaker 1

  wow, it was weird. And but it kept showing up there.

• 1:06:15Speaker 0

  You're on prime time television. I remember when um I can has Cheeseburger, the first meme website launched, and then very early on, like one of the early memes done with that impact font. Yeah, you know, like was a clippy one, and I I had it, and I saved it, and I was like, I think this might be like the earliest clippy meme,

• 1:06:32Speaker 1

  you know? Yeah. And it was it was always clippy. It wasn't Einstein or it wasn't the butler. It was Yeah,

• 1:06:37Speaker 0

  well, Clippy ended up being the default. Yeah. And so that was probably why. And also the absurdity of a paper clip telling you to write a letter.

• 1:06:43Speaker 1

  Yeah. Yeah. The animated paper clip with expressive eyes and eyebrows. Yeah,

• 1:06:47Speaker 0

  it was it was super well done. It

• 1:06:48Speaker 1

  was it was very well done. Like, you know, look. Animation was uh risk then, right? Could you actually make it look and then uh animating a paper clip? Like talk about degree of difficulty. This is an inanimate object that's all irregular shapes. Yeah.

• 1:07:04Speaker 0

  I mean, we created a character.

• 1:07:05Speaker 1

  So that's awesome. So you got the launch, we got the retirement party. Um let's now talk a little bit about sort of the the future. Uh because uh digital assistants are back uh with vengeance, right? So you either are talking to your speaker or there's features embedded in uh all of your apps, like auto reply is in Gmail, right? So like AI is definitely resurfacing. Do you think there is room for uh a social personified agent? Like are we gonna see an enterprising company sort of uh be inspired by Clippy?

• 1:07:39Speaker 0

  You know, I I feel like that that's sort of happening in a way. Like if you go way way back, um the the origin of of the actual animated assistant helping you goes back to sort of whether you want to say knowledge navigator or memex, this whole MIT Media Lab era of like intelligent agents, which were a big thing. That was like a thing that got written about in Wired magazine, uh General Magic talked about intelligent agents. Yep. And the idea was that there would be this like small piece of code that you would send off into the ether, so to speak, to solve a problem for you. Uh the first mainstream discussions of it were like imagine an agent that would always keep an eye out for low price plane tickets, or an agent that would understand where to go buy a car, or like these these small kind of things.

• 1:08:30Speaker 0

  And so our use of the assistant was rooted in those ideas, but like so often, like the ability to draw something in Photoshop or director or or whatever your modern prototyping tool is and make it in code that can ship to hundreds of millions of people, that's a big gap. And so that's really what happened with us with with Clippy is we were just early. And as you know, my favorite, like being early and wrong are kind of the same thing. And so we were just too early. So now you have the internet, you have languages, you you know, you have serverless apps, you have all these stu things that can solve it. And so and like we have tons of agents. Like if you want alerts on eBay for a certain kind of item, it's always running, and you'll get an alert.

• 1:09:17Speaker 0

  That's an intelligent agent, exactly like Knowledge Navigator, presumed you'd have. You can have travel tickets you want on kayak, and they'll alert you when there's a price alert. Like all of that stuff happens now. Yeah. And so we have those agents all the time. We what we don't have yet is is the right way to use sophisticated products that you're in over your head using. Like if you sit down with uh Salesforce and you don't know what you're doing. Like it's the same problem we had in 1995. It does more than you have a mental model for what it can do. So there's absolutely room for this all of our enterprise startups that have a product for the enterprise that's used at uh a seat have this. Problem. Like if you know whether you're a lawyer or an HR person or a talent person or in finance.

• 1:10:07Speaker 0

  Like you have a product lands on your machine or your phone that does thousands of things that you don't know how to make it do. And so there's it has to happen. Because the need to do all of those isn't going away. Plus, the competitive market is not going to allow products that don't do enough to win. So every product that's successful is going to add more stuff, making it harder to use until there's a solution for this sort of flat list of commands.

• 1:10:37Speaker 1

  And then the pressure has gotten even higher because you know consumerization of enterprise IT, which is you want it to act like your iPhone and be that easy to use, right? So the pressure is on to actually make the first 30 seconds, the first two minutes.

• 1:10:52Speaker 0

  Yeah, and like these overlays, like all that kind of stuff, like these solutions, these the overlays and solutions like that are really just modern versions of oh, let's do a wizard and let's do a task pane. Like these modalities and this sort of interaction style, it isn't right. Like you you need to get from these abstract thoughts in your head, like I would like to, you know, I would like to know all the customers to call today from that to like a feature in the product. And that is gonna happen. Of course, the the area that this has been bounced around quite a bit for years and years and years has been on just querying databases. And a lot of products have made a lot of progress on that, but that's a very, very small sliver of the kind of help. I think the kind of help that matters the most are like literally I have this is a tool that does 3,000 things at varying levels of abstraction.

• 1:11:45Speaker 0

  How do I use it? And like that's the thing that I think is the most important problem. And whether you do that with a social interface or just natural language, but there something has to bridge that abstraction

• 1:11:58Speaker 1

  gap. Knowledge Navigator is this Apple proof of concept of intelligent insistence, right? So you can actually find it here on YouTube a video of sort of, I think they had this sort of bow tied butler instead of the animated paper clip, but this was sort of the the idea of the time, right? Which is you're gonna have a social agent helping you with complex tasks on the computer and representing you as an agent in the outside world.

• 1:12:22Speaker 0

  Well, in fact, the butler character was derivative of that. And like for that very reason. It was this notion of like disciplined helper. Also, there's the psychology in there. That but using a butler is implied that they kind of know how to solve the problem that you were asking. Like this a butler is not just like a literal translation, like put a rectangle around something. It's the, oh, you want to emphasize that. Let me show you the correct way to emphasize that. And it's very interesting. And believe me, that at the time, that videotape all over the team. We knew all about it.

• 1:12:55Speaker 1

  So thank you for sharing all the old war stories. And now they are old. This is 20 years ago, and Clippy keeps on coming up.

• 1:13:03Speaker 0

  You know, like I said in my quote tweet of Chloe, like it's it's sort of amazing that your biggest failure can be something that somebody grabs onto and and thinks is super cool. Um I I genuinely think that we're at um what some friends and I referred to as the happy days moment. And the happy days moment is like all of a sudden that people think the 50s and poodle skirts and leather jackets are cool, and my parents were like, yeah, that wasn't the cool thing back then. Uh we're confused. And nostalgia has this sort of 20, 25 year window, and that's where we are right now. And so it's it's fascinating and it's great. Like it it shows a renewed interest in this field, and it shows a renewed interest, frankly, in risk taking.

• 1:13:48Speaker 0

  Like I if there's one thing I would love is for people to realize that you could have a multi-billion dollar business and do a crazy feature like that and become a

• 1:13:58Speaker 1

  multi-hundred billion dollar business. Well, that's the perfect way to head out on this episode, which is look, the nature of a startup is about taking risk. And so the courage that Steven described to sort of persevere, like on paper, this is the weirdest idea ever, uh animated paper clip guiding you through letter writing and helping you do pivot tables, like it's a crazy, crazy idea, and took a lot of courage for the team to persevere. So thank you, Steven, for coming to share all those stories. If you liked what you saw on our episode, go ahead and subscribe and then share your clippery memories in the comments. We'll see you next episode.