Continuing the series of Computer Business Review’s profiles of European high flyers we examine the prospects for the UK-based web search engine specialist, Autonomy.
Few companies fit the profile of a hot technology company start- up quite like Autonomy Systems Plc, of Cambridge, England. It is just one year old, staffed largely by software engineers and scientists from Cambridge University, plans for a listing on the US Nasdaq exchange, and has already raised $45m in venture capital, from Apax and the English National Investment Company. But above all, Autonomy has genuine, ground-breaking technology in a sector that is likely to show strong growth: software tools for making it easier to search the web more effectively. Already, Autonomy has attracted considerable attention, and has signed an agreement to provide software which will enable Rupert Murdoch’s News Corp, in conjunction with British Telecommunications Plc, to offer personalized news services. Autonomy, which was spun off from pattern recognition specialists Cambridge Neurodynamics in June 1996, has combined neural network techniques and mobile agent technology to produce a method of intelligently automating web and newstream searches. In doing so, the company believes it has achieved many of the goals which General Magic Inc and Verity Corp have described but have yet to achieve.
Legs and brains
We make whole agents, with legs and brains, says Dr Mike Lynch, managing director. The ‘brains’ in Autonomy’s software comes in the form of neural network technology. The software recognises what type of material the user likes and does not like, and learns to mimic the behavior and choices of the users. Once this has been done, an agent is sent out over the web – or it can trawl through newstreams, establishing hot links to find similar end results. Effectively, the process of web surfing is automated. Unlike, for example, Verity’s Topic text retrieval engine Autonomy does not work using key words, but by recognizing successful search patterns employed by users. While free text retrieval engines are essentially non-dynamic, searching material as it comes in or using specific lists set up by the user, Autonomy’s agents make decisions on the fly, going into Web sites, corporate intranet servers and other agent servers looking for patterns. The point is that the agent should be left to get on with it, says Lynch. Autonomy launched its first product, Agentware, to considerable acclaim last year. This enables users to set up agents which act as personalized secretaries, collecting information as directed. After this, Autonomy’s products become more advanced or specialized: its Push Content Server enables an advertiser or information provider to automatically build up profiles of a user or customer and then push only the information they are shown to like. Autonomy’s server based software enables agents to securely operate in an internet service provider’s site while the client is offline. Autonomy, in its first full year, has yet to report a full year’s profits, although, says Lynch products are shipping and revenues exist. The company, which has offices in California, expects to go to initial public offering in 1997 or 1998.