(FORTUNE Magazine) – The last thing that Laura McCann wanted to be two years ago was a bare-knuckles boss. She was a graduate of New York City's Parsons School of Design, which trained her to be an artist, not an autocrat. Moreover, she had just ended one troubled business relationship. So McCann went out of her way to be accommodating to her two new partners when she joined forces with them. She took the title of CEO at the new company, International Product Options, because her business was larger. But she agreed that the three of them would be equal partners, each with a one-third ownership. One for all, all for one.

And one colossal mistake.

As the 35-year-old McCann tells the story, she was the boss on paper only. Anytime one of her partners disagreed with her, he did what he wanted anyhow. And because her two partners had worked together before, they generally supported each other at her expense. (McCann's partners declined the opportunity to tell Fortune Small Business their side of the story.) Now International Product is deeply in debt, and the partnership is being dissolved--a situation McCann blames largely on the usurpation of her authority. "Today, I would never give up an ounce of power and control," says the reformed egalitarian. "I would treat even partners as employees."

McCann's outburst of frustration and regret is the way a partnership often ends when it goes bad because of confused authority. When two or more entrepreneurs join to form a business, few issues are more vital or more sensitive than who is going to be the leader. These days the matter is more urgent than ever as groups of friends, college classmates, and business colleagues rush to form companies that can quickly take advantage of opportunities in e-commerce.

Who will be the boss? Or does the company really need a boss? Why can't everybody vote on everything? After all, in this enlightened era the word "boss" is a bit tainted. In progressive companies there are no queen bees and drones; everyone is an "associate." (You can thank Wal-Mart for making that one popular.) If it's essential that someone be boss, what talents does he or she need for the job? How do you decide who among your partners has the right stuff? And what do you do if none of you has the skills or the stomach to pull it off?

The answer to the first question--Does one person have to be boss?--is a resounding yes in most cases. "At the end of the day, somebody has to be accountable," insists Karen Kerr, managing director of ARCH Venture Partners of Chicago. The yes is even more emphatic when a new company is raising money. Daniel Denison, professor of management at the International Institute of Management Development in Lausanne, Switzerland, describes a typical startup in which three enthusiastic crusaders launch a new company. "They get $6 million from some venture capitalists, and guess what?" asks Denison. "The venture capitalists don't want the Three Musketeers onboard. They want only one musketeer, and they want a decision-making structure." That doesn't mean that the chosen musketeer should be tyrannical and the others submissive.

At Revelwood, a New York City software company, the three partners have worked out a successful power-sharing structure. Ken Wolf, 33, is CEO and president. Partners Rob Gordy, 33, and Daniel Bernatchez, 34, report daily to Wolf just like the half-dozen other department heads. "Every so often, however, Rob and Dan and I put on our partner hats," says Wolf. Together they plan budgets, set financial targets, and determine compensation for employees and themselves. "At those times, we are completely equal," Wolf continues. "We know when we're wearing one hat or the other, and we don't cross over. It would be inappropriate for Rob to walk into marketing, for example, and say, 'Do this' or, 'Do that.'"

But Revelwood, we'd argue, is the exception. It isn't likely that a group of, say, three partners is going to produce three folks capable of being CEO.

So what does it take to be the boss? Ray Smilor of the Kauffman Center for Entrepreneurial Leadership in Kansas City, Mo., identifies three outstanding qualities (see box for more). The first, says Smilor, is vision: the ability to articulate the purpose and direction of the enterprise to outsiders and employees. The second is the ability to inspire and motivate, to be the cheerleader even in hard times. "There has been research on optimism," says Smilor, "and optimists accomplish more than others. Successful entrepreneurs encounter disaster and continue to come back again and again with something positive." The third quality is financial know-how. "The leader doesn't have to be an accounting wizard or a financial genius," he says, "but he has to know how to track the company's key financial issues and communicate their importance to others."

Unfortunately, the statistics suggest that no matter who becomes boss, it is almost even money that he or she will not measure up. A survey of 826 companies a couple of years ago by Manchester Inc., a research organization in Jacksonville, revealed that more than 40% of new leaders failed in their first 18 months. There is no single reason, although Kerr of ARCH Venture says that when she has to change the management in a startup it is often because the CEO isn't hiring good people. Steve Bengston, director of emerging-company services at PricewaterhouseCoopers in San Jose, echoes Kerr's observation. "A lot of CEOs think hiring talent is something that ought to be left to human resources," he says. "But the old expression is really true: 'A' people hire 'A-plus' people, and 'B' people hire 'C' people. In early-stage companies, great CEOs spend about 30% of their own time recruiting."

Companies that fail to establish strong leadership might survive for a time, especially if they stumble into an industry that is experiencing hypergrowth, such as e-commerce. But in the end, poorly led companies will fail. Here are the tales of four groups of partners who faced the challenge of picking the boss who would turn their bold idea into a disciplined company. One partnership, Candide Media Works, seems on the verge of finding that leader after a couple of years of power sharing. Another, International Product, has failed. The third, QuickBuy, picked a chief with surprising ease. And the fourth, Training Resources Group, came up with effective governance after years of experiments.

One of the toughest things for owners to do is admit to themselves that they have neither the skills nor the desire to be a chief executive--and then make the decision to hire outside the family. That's what Candide, a producer of Internet documentaries in New York City's Silicon Alley, is finally doing after two years of drifting as a loosely run democracy. Six electronic journalists with backgrounds in the arts and social sciences--and no managerial experience--joined forces to produce digital documentaries, often as companion pieces to TV shows. (One of its early projects was a Website for the Discovery Channel that enabled auto-racing fans to keep track daily of the 1997 Beijing-to-Paris motor rally.)

Like many partnerships, Candide was born of democratic idealism. Power would be shared. Everyone would have a vote in every decision. Early on, they even developed a method--typical of the company's management style--that allows all the partners to express their feelings while limiting the pain those feelings might inflict on others. Each partner has a series of masks at his or her disposal during meetings. Instead of Mike Bettison's accusing Robert Thomas of being too commercial and not artistic enough, for example, Bettison holds an Ernest Hemingway mask in front of his face to remind the group that they started the business to be storytellers, not to get rich.

A nice touch, all in all, but Candide's attempt at communal power hasn't worked as planned, resulting in a collective nonleadership that the partners think has kept the company from progressing. "We didn't consider the problems inherent in not having a leader," says Christina Bennett, 31, one of Candide's six founders. "We're creative individuals who have been acting more as though we're a rock band than a company." When several partners work together on a project, everyone--and therefore no one--is in charge. They have to reach consensus on matters as routine as when to draft a storyboard and whether the background should be blue or green--decisions an executive producer would make efficiently in a traditional organization.

Because all the partners are involved in individual projects, no one has time to plot Candide's long-term strategy. "We all have vision," says Bennett, "but it's hard to keep an eye 100 miles ahead when you're hunkering down on your daily work." What Candide needs is fewer but larger projects, and to do that it needs a leader who can publicize the company, market it to big clients, and, if necessary, raise more capital. No one among the founders emerged to take that role, so last summer they ran help-wanted ads for a CEO on the Websites of the top business schools. (At press time, the company was on the verge of hiring Paul Fraser, 40, a vice president at Ogden Entertainment in New York City.)

Candide seems to have solved its problem. Our friend Laura McCann--she of the company with the difficult partners--wishes she had a solution so simple. The lesson of her cautionary tale? Once you do have a CEO in place, make sure he or she has the authority to run the company.

The fundamental flaw in the structure of McCann's International Product was that although she had the title of CEO, the other two partners did not treat her as the first among equals. "When three people decide to allocate the stock evenly, that's often a bad sign," says Bengston of PricewaterhouseCoopers. "That may be the easiest thing to do, but the chances of all three people being equally valuable to the enterprise is low." The CEO is the most valuable, adds Bengston, and that ought to be reflected in the size of his or her share of ownership. Until she is clearly failing, she should be given daily command.

Here's how the arrangement fell apart for McCann: As an agent for large U.S. retailers, International Product's role was to find a suitable Asian manufacturer when, for example, Victoria's Secret wanted to bring out a new line of negligees. International Product, headquartered in New York City, prospered in its first year, as revenues reached about $2 million. But in its second year the company was overtaken by a wave of cost cutting in the garment industry. Then it lost a client that had accounted for 25% of its business. So the partners agreed on a two-pronged strategy: slash expenses, and open a Hong Kong office to be closer to the manufacturing base.

One partner went to Hong Kong to run the operation there, and soon, as McCann tells it, the Hong Kong branch was operating almost as if it were autonomous. "We would eliminate a position in New York, but then Hong Kong would hire someone, so we couldn't get overhead under control," she says. A major problem, although far from the only one, was that both offices were simultaneously servicing the same accounts, at great cost in time and money. Moreover, both offices tracked orders on two computer systems. McCann says she tried to bring the wasteful and confusing duplication under control but couldn't.

The business was sinking under its costs, in such deep water that McCann wanted to bring in outside help. "For me, it's clear that when you are in trouble, you bring in experts," she says. "But my partners felt that outside help was too expensive and that we could fix things ourselves." Even though McCann was CEO, she couldn't override them. The three partners had agreed at the start that all decisions had to be unanimous, a condition McCann had insisted upon to keep the other two from ganging up on her.

International Product would have had a formidable challenge under the clearest and sturdiest management structure. With the vague setup, however, failure became almost certain. By the end of 1999 the company was unable to service $1.4 million in bank debt or pay some of its other bills, and the three partners were in the process of closing down. By early February the partners split up and shut the company's door. "What has kept me from falling apart over this," says McCann, "is that I'm developing an Internet version of the business--matching retailers and manufacturers online--called ZWeave.com." This time she's partnering with her mother, Leah Poller, a 30-year veteran of international business. Poller will be president, and McCann, chairman and CEO.

One way to avoid McCann's problems is to pick a partner so different from you that it's obvious who will be the boss. The inclination in forming a company is to surround yourself with people just like you. In other words: a recipe for a destructive rivalry. A business might run better when the two partners are so unlike each other that who will do what is very clear.

So it is with Kenneth Knowlton, 68, a senior introvert, and Gary Miliefsky, 31, a junior extrovert, who teamed up two years ago to create QuickBuy, a dot-com located in Tyngsboro, Mass., that expects to launch its product, the Buycon, sometime this year. A Buycon is a portable icon that will allow online shoppers to avoid the annoyance of filling out address and credit card forms each time they visit a new Website. The shopper can store the information in his Buycon and drag it with him from Amazon.com to Priceline.com to toysrus.com.

Knowlton, an MIT grad, spent 20 years as a scientist at Bell Labs, then proceeded to Silicon Valley, where he was one of the pioneers in computer graphics. In the early 1990s he was working at floundering Wang Laboratories. That's where he met Miliefsky, a Wang engineer with a talent for Internet innovation. Miliefsky first suggested an alliance four years ago. "Ken, we've got to stop doing things for other people," he told the older man. "Why don't we leverage my Internet experience with your brains, your patent genius, your computer graphics knowledge?"

Did it bother Miliefsky that his prospective partner was about to become eligible for Medicare? "My parents taught me to respect my elders" is his politically correct answer. "I try to listen to people who have been on this planet longer than I have." But the real reason the arrangement has worked is this: Because of the difference in their ages and temperaments, the division of labor and titles evolved naturally. Miliefsky, the energetic extrovert, was clearly the right Mr. Outside, the partner who enjoys taking QuickBuy's dog-and-pony show to hundreds of meetings with venture capitalists. Understandably, the VCs want to meet the boss, so Miliefsky assumed both the title and function of CEO. Knowlton is a natural Mr. Inside, one whose passion and genius is for writing algorithms. Moreover, he had reached a stage of life when he didn't want to work more than 20 or 30 hours a week. He took the modest title of vice president for research. Knowlton does show up at meetings with VCs when Miliefsky assures him that the VCs are on the verge of coming across with the money. Then the two give the VCs a persuasive one-two punch. "I am gung ho," observes Miliefsky. "Ken is wisdom." In 1999 they raised $7 million with that approach. "I don't think the VCs would have funded the company without the two of us," says Miliefsky. "We're different. We're complementary. We're us."

"Different" might best describe Training Resources Group. It certainly is a model for those partners who are convinced that democracy is the way to go. At least it worked for them. Their story: When a half-dozen young idealists, many of them former Peace Corps volunteers, started TRG 25 years ago, they wanted everyone to have a chance to participate in running the company.

An early experiment was "manager of the month," in which each partner took the controls for four weeks at a time. There was little continuity. TRG, located in Alexandria, Va., has annual revenues of about $4.5 million and a clientele that includes the World Bank and the U.S. Agency for International Development (AID). When AID pays for sewer-system construction in Sri Lanka, for example, it might hire TRG to send trainers to the site to help blend Sri Lankan engineers and their American AID supervisors into a team. Negotiating and implementing such contracts goes on for months, and time wasted away while TRG's July manager tried to catch up with what June's manager had done to carry out the May manager's plan.

So TRG modified the democratic impulse--but didn't eradicate it. The company created an unusual corporate structure in which a traditional CEO serves for only five years and then goes back into the field. "We have a whole lot of folks in the company who have leadership skills," explains Ed Salt, who succeeded to the position of CEO in 1995. "It didn't make sense to have one person remain CEO until retirement." Anyone who stays with the company for five years and demonstrates enthusiasm for management issues is eligible to be on the board and run for CEO. A year from now, Salt will return to the ranks, and those interested in succeeding him will raise their hands.

TRG developed a governance system that works well within its distinctive culture. It wouldn't necessarily function elsewhere. There is no template for creating a successful structure and hierarchy, and there are enormous variables from enterprise to enterprise. A firm of professionals, such as lawyers or accountants, who band together but still work independently to a large degree, needs a different kind of governance than a manufacturing firm that requires careful and constant coordination of many parts.

But what's clear from these examples--from Candide to TRG--is that every organization needs a leader at some point, someone who can say to the world: "The buck stops here."

COMMENTS? TIPS? QUESTIONS? sb@fortunemail.com

PLUS: A LESS OPPRESSIVE IRS? -- DOT-COM MASCOT MANIA -- ECCENTRIC EMPLOYEES