(FORTUNE Magazine) – VITTORIO CASSONI sometimes lies awake at night worrying that orders for AT&T's computers aren't piling up fast enough. Small wonder. When the fashionably tailored native of Parma, Italy, arrives at his office in Morristown, New Jersey, his job is to transform AT&T's computer business from a ''nightmare,'' as Chairman James E. Olson once called it, into a shiny exemplar of the corporation's future. The nightmare is a screamer. When AT&T plunged into the computer market in 1984, soon after winning the freedom to do so by divesting its local telephone monopolies, Olson declared that the move would ''redefine the industry.'' Company officials predicted that by 1990 AT&T would rank second in data processing, behind IBM. So far the plunge has produced little more than a stupendous splash of red ink. AT&T refuses to discuss the matter, but insiders say that in 1985 the computer division lost $500 million, before taxes, on sales of $2.4 billion. Last year sales slumped to $2.2 billion, and pretax losses more than doubled to a staggering $1.2 billion. To be sure, sagging computer sales and billion-dollar losses won't break AT& T. With 1986 corporate revenues of $34 billion, the company is making more money than ever in its core business, long-distance telephone service, where it has savaged competitors like MCI. Nor will the losses unseat AT&T's top management, which answers to an unusually lenient set of shareholders. But the computer fiasco is a bitter disappointment to investors and to the chairman, and it clouds AT&T's future. The company's stock, recently trading at $24 a share, has risen only 20% since the January 1984 divestiture, compared with 76% for Standard & Poor's 400 industrials. While shares of IBM and other computer companies sell for around 20 times their earnings per share, AT&T's multiple is a lowly 8, typical of a utility. That's not the kind of business Olson has in mind. His grand plan is to make AT&T a front-runner by combining its dominance of long-haul telephone service with growing strength in computers. The combination is supposed to enable AT&T to lead the world into the age of so-called IM&M, or information movement and management: a futuristic state where voices, data, and images intermingle freely and flow over giant networks of computers. The computer operation also plays crucial roles in AT&T's current business. Roughly one-third of the division's sales come from other parts of the company that use data-processing equipment for telephone switching and administration. Besides, AT&T needs computer sales to fill up its mammoth semiconductor mills, such as the $400-million factory it built last year in Orlando, Florida. And it expects the computer division to help pay for the corporation's most lavish plant of all, the idea factory known as Bell Laboratories, which spent more than half its $2-billion budget in 1986 on electronics research and development. Balancing those short-term duties against his division's strategic goal is the biggest dilemma Cassoni has faced since becoming senior vice president of the Data Systems Division last November. Long term he must win over computer buyers at major corporations, for they will do the bulk of the spending to build the brave new world Olson envisions. But the costly, painstaking cultivation of corporate customers is likely to yield lower sales and continuing losses for the time being, even if it eventually succeeds. In April, Cassoni assured security analysts that AT&T will keep its computer revenues level with last year and will cut losses by two-thirds. If all goes well, he added, the division will break even in 1989. But insiders say the division's 1987 plan actually shows sales tumbling from last year's $2.2 billion to $1.7 billion, while pretax losses are forecast to shrink by less than half, from $1.2 billion to $650 million. Cassoni may not break even until the early 1990s. Results like that won't fill chip factories or pay for Bell Labs' research, so Cassoni may face enormous pressure to try pumping up sales soon, a move almost certain to swell the tide of red ink. He would have trouble resisting such pressure. Cassoni's pared-down crew of 5,000 marketers and product developers must negotiate with other, much larger, divisions for what it needs to do business: a sales force to deal directly with customers and an army of customer-service technicians, as well as factories and research labs. To a large extent Cassoni's fate hangs on whether Olson and other senior managers can do a better job than they have in the past at sticking with a clear computer strategy and keeping other parts of AT&T from undermining it. In its first three years as a computer company, AT&T embraced no fewer than four strategies and often pursued conflicting ones simultaneously. Lack of focus helped keep the computer unit in constant flux. Asked to describe its organization, one major supplier used the word ''kaleidoscopic.'' The division's turmoil was magnified by the organizational and cultural upheavals that have roiled all of AT&T since its historic divestiture. The first strategy, hatched within Western Electric, AT&T's $11-billion-a- year manufacturing division, was to commercialize Western's 3B series of minicomputers, originally developed for telephone switching. They had the virtue, from AT&T's viewpoint, of using hundreds of in-house chips, but the 3Bs were hardly hot boxes. AT&T priced them as high as Digital Equipment Corp.'s popular minicomputers. But the AT&T minis lacked applications software, instruction packages that make computers do useful work like tracking inventories. Some of AT&T's own marketing executives opposed launching the 3Bs. But the company ignored its Cassandras to follow an unlikely prophet. Jack Scanlon, an articulate, obsessed engineer from Bell Labs who became the computer operation's first boss, thought he saw a way to sell hardware. He proposed that AT&T change the rules of the computer game by spreading the gospel of Unix, an innovative computer operating system developed by Bell Labs and used in the 3Bs. Operating systems are behind-the-scenes software programs that computer makers jealously guard. Enormously expensive to develop, they manage the moving and storage of data within a computer. Because they determine which applications a computer can run and usually work with only one brand of computer hardware, they are also the stuff of market dominance. Business- computer companies lock customers in by convincing them to invest heavily in applications that work only on their brand of operating system and hardware. UNLIKE IBM or Digital Equipment Corp., Bell Labs had given its operating system away practically free for years, licensing Unix to anyone who asked. Scanlon saw licensing as a competitive weapon. In contrast to most other operating systems, Unix is designed to adapt fairly easily to any computer, from micro to mainframe. Encouraging dozens of computer makers to adopt Unix, he reasoned, would make the system into a new industry standard. Programmers would flock to write applications software for Unix in preference to writing for the IBM and DEC operating systems. Though hardware to run Unix would become a commodity, Scanlon imagined that AT&T's design skills and manufacturing brawn would enable it to dominate the business. Quixotically, Scanlon turned up his nose at the engineering and scientific marketplace, where Unix had legions of followers. Business computing -- IBM country -- was the market thought ''worthy of an AT&T,'' he says. But Unix was virtually unknown in that market. Moreover, Scanlon had no salesmen to call on corporations. All he could do was to recruit resellers, independent firms that create applications for the computers and sell the bundle of hardware and software into niche markets. A massive reorganization ended Scanlon's reign after a mere six months. To run a vastly expanded Computer Systems Division, AT&T picked James Edwards, a red-haired, preppy, ex-IBM planner whom Olson himself had groomed. Edwards inherited a panoply of computer and communications gear that he arrayed as ten product lines. His legacy also included Scanlon, who stayed on as chief technologist; Scanlon's Unix strategy; and a second, overlapping strategy that AT&T marketing executives had set in motion on their own. Their idea was to < make AT&T-built office telephone systems double as networks for computers. As part of the plan, AT&T would replace the telephones on executives' desks with hybrid devices for talking and exchanging computer data at the same time. Convinced demand would be enormous, the marketers had ordered specially designed voice/data terminals and computers from Convergent Technologies, a California-based supplier. Recalls a Convergent manager involved in the negotiations for tens of thousands of machines: ''We said, 'Maybe half as many would be more reasonable?' They kept asking if we were sure our factory was big enough.'' By connecting these products through their office phone systems, customers were supposed to build their own networks of desktop computers. But most customers chose instead to install networks designed expressly for computers, bypassing the phone wires. Convergent's hardware piled up in AT&T's warehouses, and last year AT&T was forced to write down some $100 million of the odd-duck equipment. The one part of the desktop strategy that worked came not from AT&T's marketers but directly from the corporate office. Edwards was told to start selling clones of the IBM PC, mainly through computer stores and various resellers. Behind the plan was a deal with Olivetti, the booming Italian maker of office equipment. In order to get marketing muscle in Europe, AT&T had bought 25% of Olivetti for $260 million in 1983. As part of the alliance, AT&T later agreed to sell a minimum of 20,000 Olivetti-made clones per month in the U.S. Marketed as the AT&T PC 6300, the unassuming, me-too product became the computer division's surprise success, far outselling its more ambitious AT&T cousins. Dealers liked the machine because it was well built, carried an AT&T nameplate, and cost them little enough that they could make fat profits. The 6300 won a 5% market share practically overnight. Though AT&T earned almost nothing on the machines, clone sales boosted Olivetti's earnings by $80 million in 1985. The price of Olivetti stock soared, and the value of AT&T's stake quadrupled to $1 billion. TRYING TO RECONCILE his division's two contending strategies, Edwards, an incisive thinker, came up with a better one. In presentations to corporate buyers, AT&T began to emphasize that its gear could be used to link together IBM equipment that already occupied customers' desktops and computer rooms. Known as ''data networking,'' this approach was supposed to give AT&T's diverse lines of computer products a common purpose, one that would differentiate the company from its competitors. From the start, however, Edwards's division was mired in AT&T's postdivestiture muddle. The division took six months simply to assemble payroll records for the employees it had drawn from other AT&T divisions. Basic logistical systems for keeping track of products, unnecessary in Ma Bell's old monopoly world, had to be built from scratch. For example, computers at first lacked serial numbers because in predivestiture days AT&T had no need for warranty records: It simply serviced everything it built. Information about manufacturing and distribution, which product managers needed to make decisions, was very hard to uncover. Says one former executive: ''There was no cost data, no revenue data, no discounting data, no real data about products.'' Corporate upheavals continued to buffet the division through the end of last year. One cutback of 24,000 employees virtually wiped out the young service technicians AT&T had retrained to handle computers; a consolidation of sales organizations announced in the middle of last year left so many salespeople frightened for their jobs that computer orders dropped 35%. The ongoing turmoil led a computer industry veteran whom Edwards had recruited as head of product management to bolt after eight months on the job. EDWARDS, it seems, was temperamentally unsuited to crack down amid what John Segall, AT&T's chief strategic planner, dryly calls ''a high level of entropy.'' Instead Edwards applied brilliance and enthusiasm that at times made the chaos worse. Out of his division came a proliferation of products, distribution schemes, exotic marketing joint ventures, and development deals with hardware and software houses. The computer boss was crippled as well because he couldn't control his division's sales. Edwards depended on a separate AT&T sales division for roughly half his revenues. These 3,000 salesmen, who were supposed to push AT& T's computers to large corporations, were mainly sellers of telephone equipment who had been hastily retrained in data processing. In 1985 they came up $650 million short of the $1.5 billion in sales they had promised to deliver to Edwards. Not only did Unix machines lack applications software that would appeal to corporate computer buyers, but AT&T's salesmen lacked access to those buyers. Their closest ties were to telecommunications managers, who could do little to influence decisions to buy Unix computers. Says one former executive: ''Trying to sell Unix to telecommunications managers was like charging the hill where all the machine guns were.'' Months of 1986 sales were lost as Edwards and the sales division wrangled with headquarters over quotas. The computer division's total sales shortfalls -- $600 million in 1985 and $700 million in 1986 -- hit particularly hard because few of Edwards's expenses were under his control. Until this year AT&T used an unwieldy system of cost allocation left over from Ma Bell. The system based operating budgets and cost allocations on start-of-the-year sales projections, then cast the budgets in concrete. Though actual sales always varied from the forecast, the budgets and cost allocations never did. Thanks to that anomaly, other AT&T units left heavy footprints all over the computer division's income statement for 1986, according to sources familiar with the unpublished numbers. Corporate overhead, at $340 million, soaked up 15% of Edwards's revenues, and charges for servicing computers in the field came to a staggering $400 million, or 18%. The separate sales division billed Edwards $300 million, or 35% of the revenue it had produced, about twice as much as IBM pays in sales expense per dollar of revenue. The computer division's own generous spending included R&D costs of $400 million, or 18% of revenues, more than double the ratio for most large computer companies. Arbitrary allocations and legitimate costs combined to drive operating losses to $800 million in 1986. Write-downs brought the total to $1.2 billion. Last October, AT&T shifted Edwards to a staff job. Cassoni, his replacement, was head of Olivetti's North American computer business and says AT&T drafted him into the job, a mixed blessing since he had planned to spend a lot more time at home with his infant daughter before Olson came calling. An ebullient marketer who smokes Marlboros and talks out of the side of his mouth, Cassoni, 44, earned a degree in electrical engineering from the Politecnico di Milano. He spent his first 13 years in computers at IBM, mostly in Europe, but including a short stint in Minnesota, and his last six under the tutelage of Carlo de Benedetti, Olivetti's charismatic and highly effective boss. Cassoni moved quickly to quash the computer division's Camp Runamok culture. He merged the ten product groups that reported to Edwards into three, refused to renew contracts with outside suppliers such as Convergent Technologies, and cut the payroll from 8,000 to 5,000, mostly by shipping people off to other parts of AT&T. He earned a reputation for political adroitness by arranging for the computer division to shrug off some of its corporate burdens. Instead of paying to improve Unix for AT&T's telephone operations, Cassoni has managed to shift the cost to the division that needs the new software. Such simple changes will cut $175 million from the computer division's development costs this year. Cassoni has not hesitated to grapple with the challenge of winning credibility for AT&T's computers while dissuading his bosses from insisting on quick, high-volume sales at any cost. His current strategy, the computer division's fourth, tempers AT&T's ''data networking'' sales pitch with European pragmatism. As he proclaimed in a polished March debut before computer industry analysts, AT&T wants to concentrate on certain high-growth market segments between now and the mid 1990s. Instead of covering the field, AT&T will try, for example, to capture a multibillion-dollar stake in ''departmental systems,'' or office minicomputers designed to serve groups of ten to 60 users. In language aimed at data-processing professionals, Cassoni explained that AT&T will sell ''general-purpose computing platforms'' for each market segment. Plainly put, Cassoni wants his computers, terminals, networks, and software to perform the same role in office automation that mayonnaise does in a sandwich -- not so much to add flavor or nourish as to make the various parts of the sandwich work together more smoothly. The functions such platforms provide are not those that ordinary users see, but utilities designed to make life easier for computer bosses, helping them cut costs and get better mileage out of their investments in other computer brands. One such AT&T product is software that eases the exchange of electronic documents between word processors from Wang Laboratories and office automation programs that run on DEC minicomputers and IBM mainframes.

STRESSING such conveniences may be a way for AT&T slowly to build credibility, provided it can tolerate the drawbacks of the mayonnaise sell. Each time AT&T hooks its equipment up to that of other manufacturers, it risks becoming mired in custom-engineering work, where costs can soar unpredictably, shriveling profits. A second drawback: Cassoni's platform products are not likely to generate huge revenues at the outset. To boost sales, Cassoni needs to find niches where AT&T's goods have special advantages. One roomy niche could be telephone-related applications, such as computer systems that help telemarketers. Last year selling goods and services by phone was a burgeoning $100-billion business, according to National Telemarketing, a New Jersey-based market researcher. AT&T is supplying systems that enable telemarketers to extract maximum work from their clerks. American Express, for example, asked for help in reducing the cost of dunning delinquent cardholders. AT&T supplied a $1.1-million computer-and-switching system that dials the cardholder, verifies that someone has answered the phone, and then switches the call to an available agent, simultaneously filling the agent's screen with a record of the deadbeat's account. According to American Express, labor cost savings will pay for the installation in less than two years; the company has already ordered four more systems. In the short term, whether Cassoni can succeed depends on his ability to cut better deals internally than Edwards did. So far, say insiders, the results have been mixed. Though Cassoni wanted his own sales force for calling on customers, his wish ran counter to a huge sales force consolidation AT&T completed last January. The shakeup created a 20,000-man organization that will sell both long-distance service and telephone and computer equipment to corporate customers. This sales force will include 600 computer specialists who report indirectly to Cassoni; otherwise he commands only a 500-man marketing team that helps salesmen devise pitches for data networking. Cassoni has been granted his second wish: that AT&T jump quickly into the hot market for personal computers based on Intel's 80386, the microprocessor that is also at the heart of IBM's newly announced top-of-the-line PC. AT&T's version, which Cassoni says will appear later this year, will be a small, extremely powerful computer to be made by Olivetti. Though it will run programs written for IBM's existing PCs, the new machine will be designed mainly for running Unix. To ensure that plenty of alluring applications programs for Unix will be available when his 80386 machine goes on sale, Cassoni in February ended a long-running fight between AT&T's Unix managers and Microsoft, the dominant supplier of operating-system software for personal computers. Microsoft markets its own derivative of Unix, called Xenix. Before Cassoni made peace, AT&T spurned Xenix customers as heretics, even though they account for nearly half of all Unix sales. Now the two companies will collaborate to tailor Unix to the 80386 chip. The resulting software will enable Cassoni's new machine to accept both Xenix and Unix applications programs, broadening its appeal. CASSONI'S THIRD WISH is to avoid serving as an outlet for too much home- grown AT&T hardware. At technical conferences, Bell Laboratories has been proclaiming the wonders of its next-generation microprocessor. Though samples of the experimental micro have already reached AT&T product designers, the company has announced no plans to manufacture the chip or sell computers based on it. If it does, Cassoni will face a chorus of AT&T chipmakers clamoring for him to sell the new computers in high volumes. For the foreseeable future, Cassoni's best chance lies in the opposite direction: modest but profitable sales and the freedom to carry products from outside AT&T that seem more likely to sell, such as Olivetti's 80386 machine. Revenue from computers in telemarketing and other niches where AT&T enjoys instant credibility may help fill the chip factories and pay for Bell Labs. If it doesn't, the sensible strategy would be to scale back the factories and the R&D budget until AT&T has learned how to market. Otherwise, Cassoni will be undone by the grandiosity and impatience that have already cost the company billions.

CHART: NOT AVAILABLE CREDIT: SANDRA HIGASHI CAPTION: Sorry, Wrong Numbers AT&T's losses in computers vastly exceeded its forecasts, which project more red ink this year. Insiders gave these numbers to FORTUNE. DESCRIPTION: AT&T's forecast and actual losses for 1985-1986 and forecast for 1987. Color illustration: telephone connected by wire to computer screen.