(FORTUNE Magazine) – On a warm Sunday afternoon in Palm Beach, Fla., at the end of Thanksgiving weekend, Jack Welch climbed aboard a General Electric corporate jet and told the pilots they would not be flying to New York as they had thought. What he was about to do required such secrecy that even the pilots could not be directed in advance. We're going to Cincinnati, Welch said.

They went, landing in darkness and rain. What came next, Welch had been dreading for months. In a private room at the corporate aircraft hangar, Welch met James McNerney, the highly regarded head of GE's aircraft-engine business and one of three finalists in the race to be GE's next CEO. Welch's message: You're not getting the job. "We had a very nice talk," says Welch. "He was terrific."

Welch walked back to his plane, where the pilots thought they would now be going to New York. Wrong again, Welch said: We're going to Albany. There, in similarly dismal weather, in the lobby of the corporate-aircraft building, he repeated the same bad news to another finalist, Robert Nardelli, chief of the GE business that makes turbines and generators for electric utilities.

Having given two superstar executives "the worst news of their careers," Welch finally flew to New York, arriving about 10:30 P.M. "I felt all kinds of things," he says. "Happy for my successor. Sad about giving bad news to friends. Relieved."

Relief was understandable. After six years, five months, and two days, the formal process of choosing Welch's successor was over. The next morning at eight, GE would announce that Jeffrey R. Immelt, 44, head of its medical-systems business, would become the next chief executive of the world's most valuable company.

Thus ended the most closely watched, highly anticipated, and frequently second-guessed corporate-succession drama ever. Also the most tightly guarded. Welch had simply refused to talk about it. When FORTUNE asked about the process last spring, he wouldn't even say, No comment; he just sat in silence. While the scuttlebutt proved right--Immelt was the rumored front-runner for two years--Welch was determined that not a whisper of the real deliberations or the final choice would leak.

It took some doing. When Welch called Immelt with the good news on the Friday evening before the Monday announcement, he asked Immelt and his wife and daughter to fly immediately from South Carolina, where they were spending Thanksgiving, to Welch's Palm Beach home. But how? Not on a GE plane--the whole company would know within hours. So they took a corporate jet from a fractional-ownership program, with Immelt posing as James Cathcart, son of GE's longest-serving director, Silas Cathcart. After Immelt and Welch met all day Saturday, the plan was for a celebratory dinner, but they couldn't risk going to a restaurant. If one of Palm Beach's movers and shakers spotted the Welches and the Immelts laughing and popping corks, they might suspect something. So dinner was catered at Welch's home--but not by a regular GE caterer. Too risky.

The next day, while Welch made his bad-news tour to Cincinnati and Albany, N.Y., the Immelts flew to New York on another time-share jet. The cloak-and-dagger precautions worked. When the announcement went out Monday morning, not a word had been leaked.

Now that the confidential exercise is over, Welch and several GE directors have lifted the veil and told FORTUNE, exclusively and in detail, how they chose the man who would succeed the world's most admired CEO. Welch showed us handwritten notes from his presentations to the board, including his lists of CEO candidates over the past six years. He did so on the condition that we observe certain restrictions on use of the names, and he made something else clear from the start: He will never tell us or anyone else exactly why he and the board chose Immelt over McNerney, 51, and Nardelli, 52. He says he can't explain it himself--and he couldn't even tell them. He had to make a choice, and he clearly found it agonizing. Though he and the board won't say why the choice went the way it did, a number of possible reasons emerge from the interviews.

As they tell their story of the process, the big theme is that Welch and the board broke most of what passes for the rules of corporate-succession planning. They never named a chief operating officer or other heir apparent. They never looked at an outsider. They formed no strategic vision of the next ten years or used any common template for measuring the candidates.

What they did was spend an extraordinary amount of time getting to know the contenders, and more time talking with one another about them. Various best-practice guidelines for boards outline succession processes that may require fewer than 100 director-hours. GE's board spent thousands. All the directors we spoke with agreed they had never participated in, or even heard of, a succession process in which so many directors spent so much time. This most striking feature of the process--a staggering amount of human interaction--was also the most important feature, Welch believes. Outsiders are often curious about the mechanics, but Welch says mechanics aren't the key: "The process is all chemistry, blood, sweat, family, feelings."

It's obviously too early to say how well the process worked, but it's worth understanding for at least two reasons. (1) An extraordinary number of U.S. companies have recently failed in their choice of a new CEO. In just the past year Lucent, Coca-Cola, Xerox, Campbell Soup, British Airways, Gillette, and Procter & Gamble have said goodbye to short-tenured CEOs. Something is apparently wrong with the way many companies are picking the boss. (2) Though we cannot know yet whether GE chose the right man, this is a company with a remarkable record of choosing successful CEOs, and Welch has an unmatched record at nearly every other part of his job. So the way he and his directors made the most important decision of their GE careers deserves a close look.

GE's succession process began formally on June 24, 1994. Of course the directors had thought about succession before then. Every December they conduct a thorough review of executive talent; part of the drill since 1981 had been to put a name in an envelope in case Welch got hit by a truck. But with Welch approaching age 59, it was time to move beyond contingency planning.

So at the June 1994 meeting of the board's management development and compensation committee (the MDCC), the agenda was all about succession. Working from a handwritten list, Welch discussed 24 candidates, in three groups. Under "Obvious Field" he listed the seven men then running GE's largest businesses. They had to be considered because of their positions, but some could be eliminated from serious consideration because of age. None of the seven would make it to the group of three finalists four years later.

Under "Contenders," Welch named four executives just below the top tier. If not obvious, they were not exactly surprising. As things turned out, none of them would become finalists either. (It's worth noting that 20 years earlier, when GE's "executive manpower" department gave then CEO Reg Jones a list of ten possible successors, Welch wasn't on it; how he got there is another story.)

Under "Broader Consensus Field," Welch named 13 other executives who had caught his eye in widely varying positions. His heart was clearly in this list; probably nothing in Welch's professional life gives him more pleasure than spotting talent. This list included all three finalists. From that point on, says Welch, "All key decisions in the careers of these people were made with succession in mind."

None of the men on the lists (they were all white males) was ever told that he was a possible successor. Most figured it out, of course, and through joshing, the top echelon could acknowledge what they all knew. So when Nardelli lost 30 pounds several months ago and starting buying his clothes at Faconnable, an upscale French retailer, Welch and others joked, "It's only for Gerry Roche"--the famous executive headhunter who ended up placing Nardelli and McNerney in their new jobs (see boxes).

Over the next four years Welch tried to fill the gaps in all the candidates' resumes and test their ability to grow. For example, candidate David Calhoun (who now runs Aircraft Engines) was put in charge of Transportation Systems, which makes locomotives; Welch calls it "the best training job in GE" because you have to deal with governments, unions, communities, and the CEOs who buy your products. Nardelli held the job before Calhoun, and another candidate, John Rice, held it after. Other important stops for CEO candidates were Asia, where GE is expanding aggressively, and GE Capital, which generates more than 40% of the company's profits. Welch is careful not to claim that the talent-building process was precise or entirely systematic; he was trying to grow leaders but also to hit earnings targets. He says, "It was a combination of personal development and responding to crises in the business that got us where we needed to be."

Welch wanted the directors to get to know the leading candidates, and not just through formal presentations. Directors needed a feel for the human side of the contenders, which Welch thought the board hadn't received when he was a candidate. Back then the directors knew him and his rivals in board settings, but "we were kept at bay" otherwise, he recalls.

Welch devised several recurring occasions at which directors and candidates could mix. Each April before the Masters Tournament, he invited them to the Augusta National Golf Club in Georgia, of which he's a devoted member. The day before each July's board meeting, he invited them to play golf at a club near GE headquarters in Fairfield, Conn., and attend a dinner afterward. The evening before the December board meeting, they dined and danced in the art deco Rainbow Room on the 65th floor of the GE Building at Rockefeller Center in Midtown Manhattan.

These were festive occasions with a serious purpose, and Welch took them seriously. He personally worked out the golf foursomes and the dinner seating. "I seat them carefully," he says. He kept records so that he could mix things up from year to year. He also found targets of opportunity, such as the Olympics (televised by GE's NBC), for bringing together directors and candidates. Over the 6 1/2 years of the formal succession process, these occasions created thousands of director-hours of interaction with the would-be CEOs.

Welch also encouraged candidates to call board members directly when they thought it useful. For example, a number of them called former Senator Sam Nunn when they needed advice on Washington matters. Bypassing the CEO on such calls is unusual, but it yields more grist for the mill.

The directors' impressions and insights became far more valuable when they could discuss them. The main opportunity was the December board meeting, when Welch and human resources chief William Conaty presented a detailed review of the company's top 20 to 30 executives, including all the CEO candidates. The directors received a book containing pictures of the executives and summaries of their job history and performance. Then, again working from handwritten notes, Welch would give his candid evaluation of each executive. Director Douglas "Sandy" Warner, CEO of J.P. Morgan, says, "For me, those discussions are the most memorable part of being a GE director." (Several directors agree that Welch's assessments were especially powerful and immediate. His notes were handwritten for security; these presentations were never going to be on anybody's hard drive.) Working through the book "is not a page-flipping exercise," says Conaty. "We might spend an hour on the first page. It's not what most people are used to."

Welch would conduct another in-depth review each June with the MDCC, the board committee most responsible for succession. This committee convenes before every meeting of the full board, about ten times a year. In all, these board events amounted to hundreds of director-hours each year devoted to discussing potential successors.

After a couple of years of tracking the candidates, Welch and the MDCC decided that committee members needed to know more about the top ones. So from 1996 to 1997 the committee, en masse and without Welch, visited several GE businesses. This is almost unheard of in the corporate world. "Most CEOs are not so self-assured that they'd want the board digging underneath the operations," says director Gertrude G. Michelson, a retired R.H. Macy executive. Some candidates weren't sure what to make of it. They called Welch to ask what they were supposed to do. "I didn't tell them anything," he recalls. "I said, 'It's your show.'"

That, of course, was the idea. Each contender was simply invited to host the MDCC for a day. "We'd start at 8 A.M., and they'd put on a pitch," says committee member Silas Cathcart, former CEO of Illinois Tool Works. "They could bring in three associates or 23. We could see how they interacted, what the rapport was, how they delegated. Did they have good people or so-so people? What were their strategies? Were they happy with their organizations?"

Because Welch's retirement was still years away, the committee members didn't want to be explicit about why they were visiting, so they adopted a cover story. "Our guise was that we wanted to understand their businesses," says director Frank Rhodes, former president of Cornell University. That ruse seems thin, but another bit of info camouflage may have been more effective. "We didn't always visit the true comers," recalls Michelson. "We didn't want it to be that obvious. So we'd visit seven or eight places, when only three or four were really of interest."

During the intensive talent review at the December 1997 board meeting, Welch and the directors decided they now knew enough to cut the field to eight. GE won't confirm the names, but internal and external speculation at the time focused on the three eventual finalists plus David Calhoun, head of Lighting; David Cote, chief of Appliances; CFO Dennis Dammerman; John Rice, running Transportation Systems; and Gary Rogers, head of Plastics. Even then, not all these candidates were equal; some were decidedly long shots. In addition, Dammerman was older than most of the other candidates and apparently was on the list not as a potential successor when Welch retired but as the name in the envelope during this period.

Six months later, at the June 1998 board meeting, Welch and the board concluded that it was "the beginning of the endgame," says Welch. They would leave all contenders in their jobs until the winner was chosen. It was time to watch and winnow.

Developments followed at a steady pace. Dammerman became chairman of GE Capital in December 1998, after Gary Wendt resigned. (In light of revelations that Welch and Wendt could scarcely bear each other, it's a shocker to realize that Wendt was reportedly on Welch's early list of CEO candidates.) The rumor mill already had Immelt leading the race, but Welch insists there was then no consensus between him and the board about who was out in front. Still, the internal buzz built to the point that FORTUNE predicted a year in advance that Immelt would be the winner. Top-tier headhunters, hearing the same tom-toms, began circling McNerney and Nardelli.

Calhoun left Lighting to run the Employers Reinsurance business within GE Capital in July 1999, a move outsiders interpreted, correctly, as meaning he was no longer in the running. David Cote resigned four months later to become president of TRW (in mid-December the company named him CEO). About this same time, Welch first disclosed exactly when he would retire: at GE's annual meeting in April 2001, precisely 20 years after becoming CEO. That implied a succession announcement in the fall of 2000.

As the field narrowed, conventional practice might have been to bring the top contenders to headquarters jobs, where the CEO and the board could get a closer look at them. That's what GE had done 20 years earlier--and Welch had hated it. He was one of three vice chairmen (the others were Edward Hood and John Burlingame) duking it out for the top job. The atmosphere became political and poisonous, says Welch: "It was awful to go to lunch and sit across from the guys you were competing with." People throughout the company chose sides. When Welch won, Paolo Fresco, who became one of GE's most important executives, walked into Welch's office with a letter and solemnly announced, "I would like to give you my resignation." Welch asked why. "I supported another candidate," answered Fresco, who figured the only honorable course now was to quit. Welch told him, "I don't care--now tear that thing up and get outta here!"

Welch wanted nothing of the kind repeated. "So I kept [the finalists] running their own businesses, hundreds of miles from each other." The three front-runners became--and remain to this day, they all insist--good friends.

By mid-2000, Welch was ready to effectively announce the names of the finalists. In June he named seconds-in-command at the businesses run by Immelt, McNerney, and Nardelli. (The moves also confirmed the out-of-the-running status of Calhoun, who was named McNerney's No. 2, and Rice, who became Nardelli's.) The announcement created what Wall Street likes best, a feeling of comfort. It said: Here are our three candidates, and whomever we choose, we have backup leaders in place, and we're telling you who they are well in advance. Analysts praised the move as a masterstroke. "I was kind of proud of it," Welch allows.

Six weeks later, a day ahead of the July board meeting, Welch held his last major meeting with the MDCC before the final decision. Despite dozens of hours of discussion already, in meetings and by phone, they talked about succession for four hours that Thursday morning. Then, as usual in July, they played golf and had dinner. The next day they spent another hour and a half on succession. "It may seem like overkill," says Welch, "but it's an iterative process." Suggest to him that a key part of the process seems to be talking about it for hours on end, and he smiles: "The word we use is 'wallow.'"

Though it may seem hard to believe, all involved say that in six years of discussion up to this point, no one had put forth a name as the proposed winner. The idea was to keep all possibilities open. But now it happened. "I did it, finally," says Frank Rhodes, who suggested that Immelt looked like the one. "Jack said, 'Good--that's what I think, and that's what Dennis [Dammerman] and Bill [Conaty] think.'" But with four months until the announcement, they still weren't ready to pull the trigger. Conaty says, "Even in the summer and fall we said we've still got time. And the other guys were doing fabulous things."

Around this time Welch began to face the reality that he would have to give bad news to two executives he admired enormously, and it tore at him. "He hasn't been sleeping well since summer," said Sandy Warner shortly before the process ended. Jack Welch? Troubled by delivering bad news? He'd done it hundreds of times, but this was different. "I've fired people my whole career--for performance," he said a few weeks before the end. "Here I've got three guys who've been great. They've delivered great results. They've never dickered me. They've never tried to put the blocks to me or to each other."

Welch got useful guidance while golfing with Sam Nunn, who told him to stop tormenting himself. Nunn's message: You've done great things for these guys; you've groomed them and made them famous executives. It was a legitimate point. The contenders had been written about in scores of articles. Disappointed they might be, but there weren't any divisional executives at any company in the world who were nearly as well known.

As the process steamed toward conclusion, a surprise event seemed to throw it off course: In late October, GE announced its biggest acquisition ever--Honeywell, for $45 billion in stock--and Welch said he would stay on through 2001 to oversee the integration. Skeptics thought they saw a too-familiar scenario, a desperate ploy by yet another long-serving CEO who would have to be pried away from his desk. To reassure everyone Welch promised he would name a successor by year-end, as planned. The successor would take over about eight months later than originally scheduled, at the end of 2001 instead of late April.

Eight days after the Honeywell announcement, GE's board met in Greenville, S.C. (site of a major Power Systems manufacturing operation), and moved a big step closer to a final choice. For the first time, Welch asked every director to speak about the candidates and rank them. It took two hours. At the end, Immelt was the clear favorite. Still, Welch did not ask for the formal vote needed to make it official. He just asked the directors to think about the choice.

With the decision essentially made, the considerations now mainly involved timing and logistics. Determined to keep a tight lid on the process, Welch figured Thanksgiving weekend would be an excellent time to wrap it up: The nation was resting, shopping, traveling, and, as it happened, obsessing over the vote count in Florida. The day before Thanksgiving, Welch called the MDCC members and said he intended to convene a phone meeting of the full board on Friday to name his successor. Were they agreed on Immelt? They were, unanimously.

On Friday, at 5 P.M. Eastern Time, Welch told the full board from his Palm Beach home that he and the MDCC recommended Immelt as GE's next chief. Did the board concur? It did, unanimously. At 5:30, Welch called Immelt, setting in motion the events that culminated in the Monday-morning announcement.

Now that it's all over, a couple of questions remain. Most obviously, why Immelt? Welch and the directors resolutely refuse to answer this directly, apparently out of a sincere wish not to criticize, even implicitly, McNerney and Nardelli. But a few relevant facts are clear. Immelt's youth was certainly an advantage. A company is swinging for the fences when it names a CEO with 20 years ahead of him; GE hit a home run with Welch and wanted to try again. In addition, Immelt is well liked and has a good shot at keeping many managers who might be tempted to leave when Welch does; if this had been a popularity contest, he would have won. More profoundly, he apparently demonstrated a superior capacity to grow, which was the most important criterion in the choice. Welch and the directors knew they could never envision the challenges a CEO might face 15 years into the job. They just knew he would have to rethink and reinvent GE. Immelt demonstrated, in Frank Rhodes' phrase, the "most expansive thinking."

Would the result have been any different without the elaborate process? What if Welch had just made the decision himself? The question is unanswerable, of course. All one can say is that the process was genuine. The discussions were substantive and candid, and everyone involved believes they were invaluable. They began so many years before the choice of Immelt that no one can guess what the decision might have been without them.

Inevitably, a process like this doesn't just stop. Only three weeks after anointing Immelt, the directors had another board meeting. It might seem reasonable to conclude that maybe, just this once, they could skip their elaborate December talent review. Conaty asked Welch about it. He said, We'll do it just like normal. After all, with three top executive changes, there was lots to talk about.

And so, on a chilly Manhattan evening, GE's directors and top executives, and their spouses, dined and danced at the Rainbow Room. Reg Jones was there--as were McNerney and Nardelli, who like Immelt received standing ovations. The next day the board--including its newest member, Immelt--gathered as usual. They received their books from Conaty and reviewed each business and staff leader, and their primary backups. Welch gave his handwritten assessments of all of them. Conaty presented his annual update on officers--who left and why, best bets to become officers and senior executives, leaders in GE's Six Sigma quality and e-business initiatives. They reviewed the results of the latest employee survey. They did all the things they usually do at the December board meeting.

And at the end, they put a name in an envelope.

FEEDBACK: gcolvin@fortunemail.com