Software Agency And Trust

Marc Demarest

July 1995


I had occasion, the other day, to reread Guy Kawasaki's Selling The Dream, and found myself skimming, then studying the Macintosh product introduction plan he included in that excellent little book in a new, and what is perhaps more important, an upbeat light.

Having watched Apple, for the last year or so, struggle unsuccessfully to find its place in the broad desktop universe, and waffle about in the face of Microsoft's relentless marketeering, I have more than once given into my fear-of-Bill and hoped against hope that Apple would discover a neglected strategic asset that would allow itself to squirm out between the cracks in Microsoft's chinese wall and break free, out into the undiscovered market Apple deserves as fair recompense for its early (and right-minded) pioneering.

Reading the Macintosh product introduction plan again, I think I've discovered that strategic asset, lying dust-covered behind the wreckage of the PIM market, next to a pile of forlorn hopes marked PowerPC. It's simple -- blindingly so -- and it's been there since the day Regis McKenna was brought into the Apple camp to help craft Apple's image. It's intangible, like most valuable strategic assets these days, and it's incredibly powerful, if seen rightly and used properly.

It's trust.

About a year and a half ago, I made my parents give up their much-loved Macintosh for a Windows PC. We argued for months about it. My parents accepted all the rational arguments in favor of Windows -- its market franchise, Apple's market focus problems, the long odds on PowerPC -- and, at last, relented in the face of my Windows bigotry. We went to the local computer supershoparama and spend about six grand on PC hardware and Windows software, including what was at that time the best antidote to Windows awful Program Manager interface: Central Point Software's PC Tools for Windows. We lugged the stuff home and had it up and running that evening. No worries, mate.

It would be an understatement to say my folks were stumped by Windows. A year and a half later, they're still stymied by the logic of a hierarchical file system and the increasingly Byzantine treaty lines between hardware and software in the Windows world. One of them calls me every couple of weeks with a problem that the people on my support hotline would close in 60 seconds with the cause indicator user error. My father is always angry -- he feels betrayed, I now realize -- and my mother is always apologetic; clearly, her voice says, I am too stupid to operate this wonderful machine.

The last call I got, a few weeks back, was from my father, and this time it was serious. Six weeks of Quicken register entries had disappeared, my father said, and the Quicken support people were encouraging him to run something called chkdsk. What the hell is chkdsk, my father wanted to know, and where is my lost data?

It was after an hour or so of diagnosis, when I had to tell my father that he wasn't getting his Quicken data back, ever, that he said the T word. "How am I supposed to trust this stuff?" he asked.

And I -- a post-graduate master of the Microsoft school of propaganda -- had no answer. I mumbled something about "sophisticated data-driven software," about "non-locality of cause and effect in complex systems," but he clearly wasn't reassured. He reentered his Quicken data, we rigged a backup script to save his Quicken files more often, but, for one Windows user at least, the nascent trust between a person and his computer is gone, perhaps forever.

A week or so after this, Word 6 ate one of my documents during a save operation. I should have known better than to embed graphic objects in the file. The feature didn't work well in Word 2 -- the error message at that time suggested that the file was too big and that I should "delete some text and try again to save the file," a decidedly peculiar message for a word processor to give its user. In Word 6, the programmers fixed the problem by changing the error message and the error handling; Word 6 told me that my disk was full, reported an unrecoverable disk error, and then locked up. When I rebooted, the file -- about 35 pages of detailed goals for the team I manage -- was unreadable (Word barked a general protection fault during the open operation), and the temporary files (usually my savior when Word goes toxic) were useless.

I should have known better, I said to myself as I buckled down to rekey the document -- save often, and never trust Word with anything but text.

The T word.

The next morning, I had a working session with one of my customers: a bleeding-edge financial services firm with aggressive Internet commercialization plans. The conversation turned, after about an hour, to intelligent agency and what effect agents would have on home-based computing and home-based Internet access. We wandered over the agency landscape together, trying to make sense of the various agency technologies and initiatives together, until one of the group -- a fund manager who'd remained quiet for the most part all morning -- said, "The technology doesn't matter. No one will trust a PC-based intelligent agent."

That T word again.

The fundamental premise of software agency -- for those of you who've been living in caves -- is that, in the face of the bewildering complexity of a world made up of international, networked data and services, the PC user turns to a single software element -- an intelligent agent -- on her desktop, to act for her in that bewildering e-universe. She works with the agent, and the agent works with the e-world. She trusts the agent, and the agent makes sure the e-world doesn't get the better of its user.

And that complexity -- the thing that agency makes irrelevant -- isn't strictly a characteristic of the Big Room: the Internet and its look-alikes. Most corporations have networks, data sources and services as complex in their way as the rat's nest that is the Internet, and are searching desperately for some alternative to a gazillion lines of PowerBuilder code on every corporate desktop.

Agency is a new software paradigm, a social paradigm, a psychological model for software design. The user enters into a relationship, in every sense of that word, with the agent, entrusting the software agent not just with her digital signature (and the identity and economic purchasing power associated with it) but more importantly with the fulfillment of her deepest and most personal expectations. Through our agents, we are intimate with our computers.

Agency is coming. We can see them emerging piece-meal on the desktop: data agents, mail agents, Web agents, financial agents of various sorts. Microsoft, by standardizing the desktop object model and API set, is helping agencyinto being. But, as my astute customer pointed out, the agency concept will most probably fail, because -- while people will deploy the proto-agents they can get today to read their mail and schlurp Web pages and pay bills -- Microsoft owns the desktop, and ultimately no one with any experience in the area cares to trust the Microsoft desktop with what really matters: the fulfillment of desire.

We've seen a taste at least of what Redmond thinks agency means. It's called Bob, and if it's not the most condescending piece of software ever written, then it's certainly the most poorly thought-out interface ever conceived by a major software house. Its assistants are neither cool enough for my son, who's mastered Windows '95 already, or discrete enough for my parents, whose idea of a good time does not include inane tips-and-trick deliverd by a digital beaver. Where do we want to go today? Anywhere Bob ain't.

We're also seeing some of the Internet access vendors get ready to enter the agency market. In particular, NetManage's acquisition of Ecco (a cutting-edge, stumblingly difficult-to-use PIM) is nothing if not the shadow of NetManage's intention to rally agency underneath the two dominant visual metaphors on the networked desktop today -- the Web browser and the shared PIM -- and to integrate the two into a reasonably useful software agent. But winning the agency war requires a market position -- a franchise -- that Internet access software companies haven't got and won't ever get. I use NetManage's products, I like them, and I'll dump them as soon as Windows 95 is released for the (superior) TCP/IP implementation and toolset embedded in Windows '95.

Until I picked up Kawasaki's book, I was in the depths of logical despair -- Microsoft can't be trusted, the agency paradigm requires trust, therefore agency will not accrue, cannot happen, does not compute.

Who, I asked myself, has the franchise mentality of Microsoft, and yet understands how ultimately trust is what makes market share happen?

The answer Kawasaki gave me: Apple, and if not Apple, maybe no one.

Wake up, Cupertino. Trust was your stock-in-trade for years. Even when the bomb appeared on the tiny black-and-white screen, we knew it was OK, that you'd take care of us, that everything would be alright once we rebooted. We bought your technology long after it became trailing edge because we didn't care about the technology; we wanted simplicity, safety and pleasure. Late at night, we stroked the cool skins of our Apples like we'd stroke a favored pet, and dove into term papers, letters and business plans, feeling close to our computing devices, believing we had a collaborator on the other end of the keyboard cable.

Opening your hardware architecture won't help. We'll buy proprietary gear in a heartbeat if it's reliable: it its supplier has market longevity, market share and if we can trust what we buy. PowerPC isn't the way out, except possibly the way out of the commercial marketplace altogether. Hardware doesn't guarantee trust anymore.

The trust users need to put Moore's Law to practical use, the trust the industry needs to make good on the InfoBahn and the Age of Connectivity, is built into software: into agency. Microsoft doesn't get it, and, even if they did, no one would buy it. Others get it, but lack what you have -- an understanding of the way to reach the broad market and stick there.

Microsoft is as vulnerable as it has ever been to the company that can build a software agent that embodies the hope for a relationship with our desktops, laptops and notebooks: a relationship that Windows will never be able to provide because -- even if Windows '95 were the lean, mean, DOS-free, plug-and-play desktop operating system we were promised -- Redmond is what Redmond is: a franchise technology supplier, driven by technology and driven to technology. Not a trusted partner-in-computing. Not the company I'm turning my desires over to.

As goes Apple, I think, so goes software agency. Here's hoping Apple picks up trust, dusts it off, and puts it back into play. If it does, I know a few hundred thousand people who'll volunteer for early beta, and pay for it.


Last updated on 06-22-97 by Marc Demarest (

The authoritative source of this document is