From Software Futures, a sister publication
Expert systems! Intelligent applications! Founded in 1979! Oh boy, are these Brightware guys for real! Well, yes indeedy, says Gary Flood, editor of Software Futures.
Mission: To help businesses automate transactions that require analysis, knowledge, and good decisions. Nothing so threatening about that little statement, is there? No need to get nervous. About what? Oh, about vast expenditures on ultra-ultra-boutique software, sold and tended by folks with very faraway looks in their eyes, who can’t seem to utter sentences that don’t include at least one paradigm, knowledge window, or heuristic. You know – whisper it – the software market that dare not speak its name – AI. While the artificial intelligentsia has probably been whittled down to its true hardcore fan base now, it’s not gone away, and many organizations are still getting value out of the arcane stuff. Every year, there is a wealth of interesting application success stories that are carried in the conference’s Innovative Applications of AI stream. In the seven years that the organizers have been spotlighting AI success stories in corporations, one company has been used in 38 of those applications – more than all its major competitors combined – and it’s the company whose mission statement we quoted above. And that company, amazingly enough, has roots going back 17 years – it’s Brightware, now of Novato, but previously of nearby El Segundo, California, but possibly better known for nearly all of that time frame as Inference Corporation. Inference has divested itself of its original seed core – as in the technology and tools around ART (Advanced Reasoning Tool), which has been carried off by Inference founder Chuck Williams since May of 1995 into a new company dedicated to the AI toolset that software represents, rather than the technology of Case Based Reasoning (CBR), which as we know Inference has been applying so successfully to the help desk market.
Playing ball
So, Chuck, we asked; seems to us like you got ticked off with Oracle refugee cum Inference president and CEO Pete Tierney, and decided to take your ball to play with kids in another park? Williams laughed and agreed that from an outsider’s viewpoint it might look that way. But we had been building CBR Express before Pete and all those other Oracle guys got there. Williams’ reference will be familiar to those of you who’ve been tracking Inference/Brightware, but for those listeners who’ve just joined us; check out a corporate backgrounder on Inference circa 1989-90 and you’ll find excited comments about the fact that lots of people who’d worked with Pete Tierney at Oracle – a successful and rich company! – had all moved to Inference. Conclusion of this syllogism; therefore Inference would soon be a successful and rich company too!! Ahem. Didn’t happen. Well okay, since Inference in its new post-vanilla AI incarnation has had an IPO and is doing pretty well financially, in a way they were right. But we had a long haul getting there, and a wrong turn or two. Again, software archaeologists, look up the background stuff on Inference and what by then was dubbed ART*Enterprise in about 1992 – and you’ll see that Inference was gonna kick Oracle’s ass in the client/server development tool/high-end 4GL market. Ahem, ahem. That didn’t happen either. We’d developed a high-value client/server development tool – just before Powersoft and all those guys really got started, laments Williams now. The attempted move into tools just wasn’t successful, so we had to go back into delivering solutions. This move inevitably led to the Inference-Brightware fission. Because the company had two solution areas; this AI tool and related consulting stuff, and the CBR/help desk thing. John Knightly, director of corporate marketing for Brightware, comments: Really we had two very different businesses in one company, with a single sales force trying to sell both. The company divisionalized, and what Knightly candidly says was a stable, slow, at one point even declining AI tools and consulting revenue stream perked up almost at once. But how come this slow progress didn’t destroy the company altogether? Same reasons it survived the AI Winter of the late 1980s, according to Williams. We consistently delivered value to our customers, we had great functionality including advanced object-oriented, rule, and Case Based technologies, and we simply have the world’s leading knowledge engineering facility.
Oldies but goodies
OK – so solutions, not tools, were the answer. Those solutions, by sheer gravitational force, had to be more like the ones that had won all those Innovation awards. But to some of us, these sound suspiciously like Motown oldies but goldies – as in, didn’t I read about that application back when AI was a Good Thing, and it was still Morning In America? Take American Express’ Credit Assistant application. You know, that expert system they have that helps Amex work out if you really can afford to buy Jack Kennedy’s golf clubs on-line, so that you don’t make Sotheby’s call the police? Now set me afire and call me a Zoroastran, but that’s been around since Kruschev was still a big cheese. And lo and behold, there it is in Brightware’s press kit – claimed productivity gains of 20%, minimum $1.4m annual savings, reduction of 22 IMS transactions for a review to one, and imposition of a global credit policy. We tried to tease Williams about this, but he laughed it off very convincingly. As a matter of fact, that system has been in continuous successful deployment since Christmas 1988, when it had 1,000 rules in its knowledge base and only dealt with the American Express [green] personal card. Now it’s up to 5,000 rules, runs on Suns and RS/6000s, and covers all cards issued by the company. Hmm. Not convinced, we wanted to know what the forward path for the company can possibly be. Can Brightware win new customers and not just live off the royalties from its successful hits of the early 1960s? To answer that, consider what it has to play with, post-Inference spin-off. The company claims $10m+ in revenues, and says it focuses on the financial services, insurance and telecommunications markets. AI’s been quietly building steam [here] behind the scenes, adds Knightly. It’s always been giving high ROI [return on investment] for many of these companies. In financial services AI as a term actually has very positive connotations. Its investors and backers include JP Morgan, Ford, and Lockheed, presumably in the aspect of once having been users (Ford certainly is, as is IBM, Canon, Pac Bell, the US Navy and NASA). Brightware says its user base is over 200 separate organizations, and its ART family, including the original ART, ART-IM and ART*Enterprise, has over 35,000 licensed users worldwide. ART*Enterprise, the main product, due for an upgrade to version 2 later in the year (it’s slipped the original date of last month for the Windows and June for the Unix releases), offers the builder of knowledge-enabled solutions a toolbox that gives him/her what Sybase-nemesis Curt Monash describes as the strongest inferencing capability of any full-fledged [application development tool].
Easing the bottleneck
Be that as it may, if you look in the job ads in the back of the computer magazines you won’t see many full pagers asking for folks with extensive knowledge engineering skills these days. (By that, we mean the process of extracting/collating/codifying the knowledge of a domain expert in a form that can be used by a program to assist a decision-maker – in a sales negotiation, diagnostics and dispatch, or network management scenario – with an intelligent back up. The terms also implies the ongoing updating and maintenance of said body of rules of thumb or conditional statements, hence the term knowledge base.) Knightly acknowledges the problem – and proffers the Brightware patented solution. One of the issues with AI was that for the largest companies in any given industry hiring the engineering staff necessary to build and maintain knowledge-based applications was not a problem. And in many cases they were – and are still – very, very successful with these applications. But below the top ten or twenty, the cost of hiring knowledge engineering staff and skills is an issue, even for the top fifty or a hundred corporations. To ease this bottleneck and extend the appeal of such systems, Brightware is rolling out knowledge management frameworks – a componentized approach, if you like, to the problem of enhancing knowledge bases. Tools are offered that sit atop ART*Enterprise and allow business experts to capture knowledge and extend such applications. Note that some programming skill in AI is still expected at the beginning – the framework must be properly specified at the outset – but the idea is that such programming resources won’t be so necessary in the later adding or changing of that ruleset. Knightly offers the example of the mortgage banking market, where lenders resell loans to a secondary market which depends on a lot of complex rules/guidelines to determine who actually can resell. He claims policy analysts are using frameworks to keep these rule bases up to date, and also claims that analogous progress is being made with ART applications in insurance and manufacturing. How different, though, can these tools be from the kind of standardized, generic expert system tool builders of the late 1980s? Well, the Brightware take on that is by focussing on vertical markets (basically in and around diversified financial services), more domain-specific rules for areas like risk management can be built, which tend to be more useful than some of those earlier superficial horizontal rule editors. As an exemplar company he pointed us to a Manhattan legal knowledge based systems VAR of Brightware’s called Counselware, founded by a technologist, a lawyer, and a judge. The application area is a framework helping determine legal risk in the financial services area. Specifically, if a stockbroker advises clients to take positions that later result in loss, the agent may well face a lawsuit. Most of these actions don’t ever really come to anything, it seems, but some do, and the idea is to offer a system that can sift the scores of complaints a firm might face to identify the ones which really do need close attention.
Underhyped
Fred Parnon, president of the 18-month old company, told Software Futures that in his view the technology represented by ART*Enterprise have gone from being ridiculously overhyped to ridiculously underhyped. Counselware is also interesting in terms of Brightware since it intends to offer its products over the Web for Internet or Intranet access, and this is very much in tune with Brightware’s own stated direction. At the moment most Web commerce consists of little more than signing up for subscriptions, or giving out a credit card number, says Knightly. But the future lies in much more complex transactions that can’t be done simply by filling out a form. Think of a health provider that might want to offer direct access to very specialized care; or buying a PC and configuring it on- line so you’ll be sure it works when it’s delivered. So Brightware wants to be the Quicken of Web commerce? Actually, we’re going to help the banks compete with Quicken. Up until now, AI has been kind of hidden away in the processing centre dungeon. But we think its real added value is going to be at the point of customer interaction, be that over the phone or over the Internet, replies Knightly. As companies do more and more business on the Internet, they’re beginning to understand that they’re being represented more and more by the computer. So that computer better be very intelligent, reasons Williams. That’s going to be very important for the player in the virtual economy. Thus the commercial AI of old – standalone, CPU intensive, never performing well (or acceptably!) other than in batch – is being transformed into the Web-enabled client/server of today which is finally the right platform in his view. But is Brightware the right sort of company to exploit that platform and help those complex Web-enabled commercial applications to achieve cruising altitude? Parnon offers a mixed assessment: They’re guys who have hung in there and didn’t give up, and they are rocket scientists. Which means that they’re sometimes so smart they spend perhaps more of their resources on their technology than they really should. I think they need to work more on their marketing side. To which Williams and the rest of his team would probably reply with a wry shrug, and explain that it’s as an enabling technology supplier that Brightware might best succeed. We’ve learnt the lesson about trying to be all things to all people, notes Knightly. So we’re no longer allowed to snicker up at our sleeve at Brightware saying it’s going to be the next Oracle; but what’s much more believable is that it’ll help the Oracles and other Fortune 500 companies deliver the kind of full, interesting, worthwhile Web applications that might even make this Information Hypeway thing happen. And hey, that’s not such a bad thing for a clapped out old AI company, now is it?
