Sign up for our newsletter - Navigating the horizon of business technology​
Technology / AI and automation

A TALE OF TWO INFERENCES

Back in August we asked if help-desk software vendor Inference Corp might be in need of some help itself (CI No 3,235). That plight, as we’ll see, sadly hasn’t got any better for the Novato, California based outfit – but with an ironic twist: for as Inference seems to be spinning slowly into oblivion, its neglected orphan offspring, Brightware Inc, is doing surprisingly well, and may end up outclassing its helpless parent. This is a somewhat complicated story, involving both eighteen and two year old companies which are actually one and the same thing, one no-longer fashionable technology (Artificial Intelligence, AI), one extremely fashionable technology (the web and e-commerce), and two radically different uses for the same base engine, one of which seems to be running out of steam while the other goes from strength to strength.

By Gary Flood

Our tale begins in 1979, when Chuck Williams co-founds a company he calls Inference, to keep faith with the then trendy technological phenom known as AI. Inference builds a complex, Rolls-Royce of software development environments for building and maintaining knowledge based systems, ART (Advanced Reasoning Tool), which garners praise and sales and kudos and all that nice stuff – in the very narrow market that existed through until about the end of the 1980s which thought little of buying $100,000 Lisp workstations and the like to explore the bounds of KBS and the like. Around 1990 or so Williams had been moved aside as Big Enchilada by one Pete Tierney, previously of Oracle Corp (who isn’t in IT these days?). Tierney recruited some other Oracle types, implied history would repeat itself and Inference would be a big deal company like unto the Death Star, itself. By the time that the basic AI market starts to sag, though, Williams and his rocket scientists have come up with a new mousetrap, an application of an esoteric AI research topic called CBR (Case Based Reasoning), which is all about using past examples stored in a knowledge base to work out the likeliest answers to a current problem. Inference then starts selling CBR as an application, begins to get success, in the meanwhile trying to turn ART into a high-end client/server development tool. In the end, CBR wins, and ART*Express, as the nascent tool has been called, gets sidelined (time frame is now roughly 1994). Tierney then realigns the company as a maker of intelligent server software backbones for help desk systems, which is a smart move, and the company finally begins to make money… at which point Williams and he part ways. Williams takes the ART*Express side, or the Inference Front Office Solutions division, which really holds the last of the AI element of the company, and turns it into Brightware Inc, which will sell ART stuff. Tierney takes the CBR/help desk part of the company public in June of 1995, and does very well. So Tierney got the better part of the deal? Sure looked that way for a while, as Inference made money for its first two years as a public company (net of $3.8m on sales of $29.4m for fiscal 1995, net of $2.4m on $36.0m revenue for fiscal 1996), while Brightware mooched about, vaguely telling this reporter in April of 1996 that it was doing $10m plus in sales and concentrating on selling AI expertise to the financial services and insurance industries, where it had some long standing clients. Ho hum.

Victim of its own success

White papers from our partners

Then things started to get interesting. On the one hand, from the third quarter of last year on Inference has been in trouble. The company has been in some ways a victim of its own success – by trying to compete with the mainstream help-desk providers, it has only exposed its weakness as a more high-end product, perhaps even opening up doors to these other fellows: so Tierney is desperately trying to turn his tanker around, attempting to refashion it as a provider of so-say self-service and knowledge management software instead of trying to maintain his co-existence with the CIM (Customer Information Management) vendors. According to the gripes of some of his investors on the bulletin boards, he has lost some hearts and minds in the process. The company has seen its share price fall from over $20 a year ago to $4.50 today, and its sales are still shrinking ominously. Its third quarter figures (released last month) showed its problems haven’t gone away – its revenue for the first nine months of its fiscal were $20.8m, down 20% from $26.0m for the same period in fiscal 1997 – and that on an operating loss of $4.4m for the period, compared to operating income of just $300,000 last time. In net terms, that is down to a razor thin $500,000, from net of $800,000 last time. For the third quarter itself, sales slipped from $7.5m to $7.0m year on year, and unless Tierney had made money on a sale of an investment of $3.8m, he would not have been able to report net income of $2.4m. On the other hand, Williams had the nous to spend some time trying to figure out an application that he could make money out of – and he seems to have chosen the right one. Starting in about April of last year the company has refocused as a provider of a sales server for web sites, shipping its first such product, the eponymous Brightware 1.0 this August.

Round the clock selling tool

This is claimed to turn a company’s static-ish net presence into a true round the clock selling tool, by enabling customer email requests to be fielded by an inbound marketing [intelligent] agent claimed to do similar stuff to a human telemarketer. By using a combination of AI techniques (fuzzy matching of natural language input, good old CBR – ART is the inference engine in the product, which is written in a combination of C++, Java, and Visual Basic), Brightware has released a product that knows typical words and phrases applicable to many standard business enquiries, though customizable for any particular organization’s needs, and thus automating replying to freeform web and electronic mail enquiries. The thing is a Windows NT offering, priced at $95,000 per year leasing or $190,00 straight sale, with native links to Sybase and Oracle, ODBC for other databases, a C++ API for other applications, and support for standards including MAPI for Microsoft Exchange and Lotus API for Lotus Notes. Brightware claims you can get the thing up and running in six to eight weeks. The thing is being hyped as moving the web beyond the publishing paradigm to a true sales paradigm – Too many web sites are just marketing brochures, online brochureware, with less than 1% of visitors ever staying to buy anything says Williams. But it does seem to have won real customers, including Wells Fargo & Company, Fannie Mae, and American Finance and Investment (you can see an example of Brightware in action at http://www.loanshop.com).

Shooting for an IPO

Based on this sort of customer win, the company has been able to supplement a modest first round of investment (for $3m in May of 1996) with a second, much better, $11.5m round in October. And according to The Red Herring that may be the company’s mezzanine financing, with the firm shooting for an IPO within the next six months. Obviously Brightware isn’t the only player in this area – Gartner Group would certainly offer Trilogy Development Group Inc and Calico Technology Inc as examples of what it is dubbing unassisted selling systems vendors – but Williams claims he has never had to compete with them so far on a deal. But nonetheless many organizations will surely be interested in any technology that helps make the web do something for business other than be another medium to print the catalogue. Williams himself says the Herring is too aggressive on that time frame, and the company is highly conservative in terms of such affairs – but the fact remains that Brightware is going in one direction and its erstwhile owner seems to be going in another, at least for the time being. Packaged applications are just the best way for us to exploit our core technology competencies in AI, he notes. Will the verdict of history (and the market) be that he by luck chose the right application – and Tierney, ultimately, the wrong?
This article is from the CBROnline archive: some formatting and images may not be present.

CBR Staff Writer

CBR Online legacy content.