When Two Heads Are Better Than

8/9/2019 When Two Heads Are Better Than

1/20

W h e n Two (or M ore)Heads are Betterthan One:T H E PROMISE A N D PITFALLSOF SHARED LEADERSHIP

James O'TooleJay GalbraithEdward E. Lawler, III

At thie turn of the Miiiennium, A mazon.com listed some 4 00 curren tleadership books on its web site. A quick review of the subjects cov-ered in those books reveals an impressive range of styles of leadersfrom the take-charge "man on a white hors e" type to the spirituallyinspiring sort, a la M other Teresa. Nonetheless, an a ssumption com mon toalmost all these books is that leadership is a solo acta one-person undertak-ingregardless of whether the organization being led is a nation, a global cor-poration, or a scout troop. However, every rule has its exception: In 1999, the

publicat ion of Co-Leaders: The Power of Great Partnerships, by David Heenan andW arren Bennis, presaged a new way of thinking a bout corporate leadership."This iconoclastic book broke with the traditional concept of leadership as anindividual endeavor and, instead, treated the activity as a shared effort. At thetime the book was published, our research at the Center for Effective Organiza-tions w as leading us to a similar, if m ore radical, conclusion: leadersh ip is asmuch an institutional as it is an individual trait.^ Hence we concluded that "greatminds were converging" on a new model in which leadership would come to bethought of as a team sport (or, at least, as "doubles"). In fact, nothing of the sorthas happened.Without creating a ripple, Co-Leaders disappeared into that vast sea ofunread leadership tomes. Our own research received a hearing in 2000 at theWorld Economic Forum in Davos, Switzerland, but the findings were receivedwith indifference. As we see it, this resistance to the notion of shared leadershipstems from thousands of years of cultural conditioning. We are dealing with anear-un iversal m yth: in the po pular mind, leadership is always singular. Four


2/20

When Two (or More) Heads are Better than One

wisdom and tru th. L ater efforts by Plato's pupil, Aristotle, to demonstrate thatwisdom is never the sole province of one person fell ondeaf ears. The die hadbeen cast and, besides, Plato's view coincided with the kind of leadership mostpeople saw inpractice: one-man rule.Thus, for most people, shared leadership is counterintuitive: leadershipis obviously and manifestly an individual trait and activity. When we speak ofleadership, the likes of Mohandas Gandhi and Martin Luther King, Jr. springto mind. We don't immediately remember that, during the struggle for Indianindependence, Gandhi was surrounded and supported by dozens of other greatIndian leaders including Neh ru, Patel, and Jinn ah, with out w hose joint effortsGandhi clearly would have failed. We forget that, far from doing it all himself.King's disciples included such impressive leaders in their own right as JesseJackson, Andrew Young, Julian Bond, Coretta Scott King, and Ralph Abernathy.Ditto Winston Churchill, Franklin Roosevelt, Thomas Jefferson, George Wash-ington, and almost every other great leader. When the facts are fully assembled,even the most fabled "solitary" leaders are found tohave been supported by ateam of other effective leaders.'

Moving to the business world, the identities of American corporationsare often viewed as mere reflections of the personalities of their leaders: entireorganizations are portrayed as shadows of the "Great Men" who sit in the chiefexecutive chairs. Fortune and BusinessWeek cover stories are more likely to beabout Jack Welch, Bill Gates, William Clay Ford, or Larry Ellison than about GE,Microsoft, Ford, or Oracle. Investors join jou rnalists in the personification ofcorporations, focusing on the characters, biographies, and "charisma" of a singleleader. Wall Street, inparticular, dem ands to see someone clearly in charge atpublicly traded corporations. That's one reason why both the media and TheStreet love Berkshire H athaway: for all intents and purposes, that corporationis Warren Buffett.

Business schools dutifully conform to the common wisdom: Leadershipis studied, and taught, in the singular. Shared leadership is largely ignored inthe academic research literature inwhich corporations are assumed tohave oneleader who is clearly in charge: the CEO. If team concepts are taught at all, itis no t in leadership courses. MBA students are schooled to believe that CEOsbehave as solo operators (over the last decade the dom inant role mod el hasbeen Jack Welch). In terms of best practices, it is assumed that a single personmust be held accountable for performance of a corporation, and for each of itssubsets. This individual focus is often appropriatefor example, when studyingthe role of an entrepreneur who is running a company that he or she hasfounded. Indeed the issue isn't tha t solo leadership is always wrong, or evenusually w rong; rather the problem that th e traditional view blinds us to the exis-tence of other models and causes us tooverlook a t remendous amount of valu-able corporate experience that runs counter to the received wisdom.


3/20

WhenTwo (or More) Heads are Better than One

Change at the TopIn fact, the trend over the last half-century has been away from concen-tration of power in one person and toward expanding the capacity for leadership

at the top levels of corporations. In the early days, publicly traded corporationswere run by a President and a Vice President (whose prim ary role was to step inif the President became incapacitated). Since the end of WWII, countless combi-nations and permutations of such roles as Chairm an, CEO, Vice Ch airman, COO,and CFO have been created and proliferated. The reasons for this increase ofshared leadership are relatively obvious. History shows that businesses depen-dent on a single leader run a considerable risk: If that individual retires, leaves(or dies in office), the organization may well lose its continuing capacity to suc-ceedwitness the performance of General M otors after Alfred Sloan, ITT afterHarold Geneen, Polaroid after Edwin Land, and Coca-Cola after RobertoGoizueta.

Frequently, organizations learn the hard way that no one individual cansave a company from mediocre performanceand no one individual, no matterhow gifted a leader, can be "right" all the time. As the CEO of Champion Paper,Richard Olson, explained at a recent conference at the Harvard Business School,"None of us is as smart as all of us." Most simply, in large corporations there isjust too much work for one person to do, and no one individual is likely to haveall the skills needed to do it all. Since leadership is, by definition, doing thingsthrough the efforts of others, it is obvious that th ere is little that a businessleaderacting alonecan do to affect company performance (other than tryto "look good" to investors).

Amana Corporation chief executive Paul Staman recently explained toau tho r Joe Tye the benefits of joint leadership: "it allows m ore time for leadersto spend in the field; it creates an internal dynamic inwhich the leaders con-stantly challenge each other to higher levels of performance; it encourages ashared leadership mindset at all levels of the company; it prevents the traumaof transition that occurs in organizations w hen a strong CEO suddenly leaves.""Significantly, Staman is one of our co-leaders atAm ana. The company has n othad a single CEO since 1995 wh en leadership of that complex organization(with business units in farming and forestry, utilities and construction, manu-facturing and retail, and tourist services) was divided among a team of four co-equals along industry lines. Until that restructuring of leadership, Amana hadnot made money since it sold its famous refrigerator line years earlier. Now, thecompany is making steady profits even though it is involved in cyclically volatilebusinesses. W hen Am ana's Gang of Four are asked w hat is most imp ortant inmaking their unusual arrangement work, according to Tye they identify "ashared set of guiding principles, and a team inwhich each mem ber is able toset aside ego and 'what's in it for m e' thinking."


4/20


Partnership at th e TopThe first thing one must understand about shared leadership is that thepractice is neither new nor unusual. Despite the ancient and broadly held view

to the contrary, h istory is full of examples of shared political power. In its declin-ing years, the Rom an E mpire was ruled by two "Caesars," one based in Romeand the other in Constantinople. In imperial China, real power was wielded notby the person on the throne but by the meritocratic Mandarin class of leaders.As those tw o examples illustrate, the ways inwhich leadership is shared varieswidely. Indeed, it is through understanding the reasons for such differences thatone can began toevaluate where, under what conditions, and towhat extentshared leadership is appropriate in the corporate world. As shown inTable 1,there are many past and present examples of shared leadership at the top ofmajor U.S. corporations. Most of the examples are of co-leaders, but sharedleadership also has involved th ree and mo re individuals. Because the aura (orPR) of great, high-profile solo operators like Jack Welch, Henry Ford, and HaroldGeneen has caused the general public to believe that leadership is an aria sungby one prima donna , we need to remind ourselves from time to time of th efamous duos of the business world, HP's Hewlett and Packard, Berkshire Hath-away's Buffett and Munger, ABB's Barnevik and Lindhal. Then there is Intel,which has been r un by leadership teams from day one: various com binations ofMessieurs Noyce, Moore, Grove, and Barrett (see Case I). We tend to overlookthese counterexamples because we are misled by the fact that joint leaders sel-dom are identified as such. Instead, the m emb ers of leadership teams usuallycarry differentiated titles (CEO, Chairman, President, COO, Vice-Chairman, andthe like) to make it easier for outsidersand for those in the investment com-munitywhose traditional views of leadership cause them to insist on the orderof formal hierarchy.

The fact that shared leadership exists doesn't make it a good practice, ornecessarily bette r tha n th e solo variety. Indee d, as is indicated inTable 1, someof the most visible examples of shared leadership have ended in failure. So,when are two (or more) leadership heads better than one? The simple answeris when the challenges a corporation faces are so complex that they req uire a set of skillstoo broad to be possessed by any on e individual. That was recognized as t r ue as ear lyas the 1940s w he n strategist Charles Merrill teamed up with implem enterWinthrop Smith tomake Merrill Lynch, Pierce, Fenner and Smith the longestrecognizable n am e in corporate America.' In 1969, Avis CEO R obert Tow nsend'sbestseller. Up the Organization, touted the advantages of joint leadership and out-lined how two executives can divide the tasks of the CEO.* Townsend explainedthat the best duos are like yin and yang: they say, "neither of us is very good,but ou r weaknesses (and strengths) may be com pensating." The trick, Townsendsaid, was for co-leaders to "split up the chores, check in advance on strategicmatters, and keep each other informed after the fact on the daily disasters."


5/20


T A B L E I . Evaluating Shared Leadersh ip (Some Selected Examples)

Company(approximate years)

Effectiveness o fTeam over Tim eCo-Leaders

Citigroup (1990s)Apple (1980s)Apple (1990s)Schwab (1990s)Microsoft (current)Morgan Stanley (1990s)Boeing (current)Patagonia (1990s)Intel (1970s)Intel (1980s)Intel (1990s)HP (1950s-1980s)Goldm an Sachs (1970s)Goldm an Sachs (1980s)Goldman Sachs (1990s)Disney (1980s)Disney (1990s)Berkshire Hathaway (1980s-1990s)ABB (1980s)Ford (1980s)Arco (1980s)Asda (1990s)Oracle (1990s)TIAA-CREF (current)Motorola (1960s-1980s)

SanfordWeillSteve JobsJohn SculleyCharles SchwabBill GatesPhil PurcellPhil GonditYvon ChounardBob NoyceGordon MooreAndy GroveBill HewlettJohn WhiteheadSteven FriedmanHank PaulsonMichael EisnerMichael EisnerWarren BuffettPercy BarnevikDonald PetersonR. O.AndersonArchie NormanLarry EllisonJohn BiggsBob Galvin

John ReedSteve WozniakSteve JobsDavid PottruckSteve BallmerJack MackHarry StonecipherTom FrostGordon MooreAndy GroveCraig BerrettDavid PackardJohn WeinbergRobert RubinJohn CorzineFrank W ellsMichael OvitzCharlie MungerGoran LindhalRed PolingThornton BradshawAllan LeightonRay LaneMartin LeibowitzMitchellAWeitz/Fisher

During the mid-1970s, there were several American companies with twoleaders, the most visible of which was Atlantic Richfield, headed by the entre-pren eurial oilman, Robert O. Anderson (who directed the compa ny's acquisi-tions, growth strategy, and its board) and the scholarly, inspiring ThorntonBradshaw (a former Harvard Business School professor who oversaw internaloperations and external stakeholder relations).In the 1980s, a nearly bankr upt Ford was saved by the dynam ic duo of"finance whiz" Red Poling and "car guy" Donald Petersen. This joint leadership


6/20

When Two (or M ore) Heads are Better than One

Case I: The Intel Office ModelThe members of the Intel Office were founders o f the company-Bob Noyce, Gord onMo ore, and Andy Grove, all of wh om became the CEO. All three owned sizeable stakes inthe company.The Office is also used to gro om successors.The original Office had Bob Noyce as CEO and Chairman, Gordon Moore as President, andAndy Grove as Executive Vice President.This Office lasted only a couple of years. Bob Noycewas pro ud of having fired himself as CEO and for taking the position ofV ice Chairman.Noyce, the co-inve ntor o f the integrated circuit, knew he was not a manager So after thecompany achieved a certain size and its financing was stable, he turne d th e CE O role o ver toothers. In this case, Gord on Mo ore became Chairman and CE O and And y G rove became thePresident and C O O . Noyce said he would fire himself the day that he walked do wn th e hall,saw a new employee and did no t take the time to introduce himself That day was the day theorganization had so many people, tha t he co uld n ot know everyone's name. It was also a tim ewhen the company was of a size that it needed to be run by formal systems, plans, and bud-gets. Noyce knew he was not the person to build and run these processes.In the Office that took shape. Bob Noyce became Mr Outside. He was Intel's face to theworld. A n attractive and articulate man, he testified before Congress on th e Japanese threatIn semiconductors. He represented Intel to the customer and to the Semiconductor IndustryAssociation. W he n Intel was having a tough com petition over an attractive jo b candidate, itwas Noyce who did the last interview and made the offer Few could resist.Go rdon Mo ore was the C EO and the long-term thinkerT he originator of Moore's Law(double computing pow er every eighteen m onths), Go rdon though t about the evolutionof technology He took care ofTechnology and Finance. He was a quiet and introverted man.However; he was always at ho me wi th a young engineer with a new idea. Andy Grove was onthe opposite ends of the time, emotion, and management spectrums relative to the other tw o.Grove was short-term and hands-on. He held people accountable. Everyone had quarterlygoals called 0-o ne s or 0- tw os . He was in fact the C O O . While Bob Noyce hated formalsystems. Grove loved them. He built and ran the Strategic Long Range Planning process, theCouncil system, the performance management process, the practice of I on I's, the ma trixorganization, and so on. He was the organization designer and wro te books a bout the IntelW ayW hil e Mo ore and Noyce w ere quiet and calm. Grove was volatile. Du ring a business review.Grove would explode,"That's nonsense!" A heated discussion would ensue.The episode wouldtypically end with a thoughtful summary and proposed solution from Moore .The reviewwould then proceed until Grove erupted again. Grove wo uld surface the issues (albeit no tvery diplomatically) that required discussion and could profit from the insights of Moore andNoy ce. W hil e no pushovers, M oore and Noyce's style wo uld not necessarily have surfaced thetough issues.The three members of the O ffice were opposites. Yet they we re able to conve rt their differ-


7/20


Grove w ro te about "constructive con fron tation " as a means to surface tough issues and dis-cuss the m fro m all sides.The thre e met tog ethe r every Friday m orning, probably at Grove'sinitiative.They shared inform ation, developed their shared positions, and decided o n th eagenda for theTuesday Executive Staff meeting. No one really knew what took place on Fri-days, but they probably worked through their differences. In the end, all thre e g ot the ir chanceto be Ghairman and GEO.There are two main lessons from the founders of Intel.They were able to combine their dif-ferent skills into a combination of complimentary skills. Each one had his own specialty anddid not compete for credit. A lot ofthe credit goes to Bob Noyce and his low ego needsand realistic assessment of his own skills. Moreover; after tw o successful startups , he becam e amentorThe second lesson is that they worked hard and talked together to maintain a unifiedposition on major issues. If there we re differences, the Friday meeting was the approp riateforum for discussion and resolution.

had characterized their long careers at Fordand publicly acknowledged theircom plemen tary contributions that positive change started to occur in the orga-nization. Eventually, after m uch counseling and effort, P&P's public truce grewinto genuine mutual respect for each other's skills. Only then did the silo-basedconflict that had characterized Ford's culture finally recede. As one Ford execu-tive explained to Richard Pascale, once Ford's leaders got over worrying about"Are we w inning against each othe r?" they w ere ready to focus on "Are we w in-ning against the Japa nese ?"'

The Ford experience highlights both the promise of joint leadership andthe obstacles that stand in the way of its success. It is commonly assumed, asTownsend did in the 1970s, that the biggest problem is "dividing up the work."In fact, experience has proved that to be just one determinant of success. That iswhy there are as many examples of failure of shared leadership (Sanford Weilland John Reed at Citigroup) as there are examples of success (Steve Ballmer andBill Gates at Microsoft). Below we identify the factors that frequently explainwhy the co-leader model works in some cases and not in others. Based on ourobservations, that success depends on how co-leadership roles originate, howcomplementary the skills and emotional orientations and roles of the leadersare, wh ethe r they are selected as a team or as individuals, how they w orktogether, and how they involve others on the management team.

Origins of Shared LeadershipShared leadership can originate in different ways and for different rea-

sons: Co-leaders arise from corporate mergers of equals, from co-founders, fromthe practice of two individuals sharing jobs, and from invitations from sitting


8/20


as a "merger" in the real world: there are only acquisitions. Hence, to pretendthat two former CEOs are going to be equals in a newly formed company is thet r iumph of public relations over reality.A second source of difficulty arises from the fact that merger-createdco-CEOs have no history of working together (except in "doing the deal"). Therelationship between such relative strangers doesn't have a basis of trust to buildon (a rare quality that needs to be created slowly from scratch, decision by deci-sion). Worse, mergers of any kind require win-lose decisions to be madequicklywhich usually means laying off people and closing facilities: Whichmerger partner gets the CFO job? Whose branches are to be closed? Trust is anunlikely by-product of such a rushed and high-stakes decision-making process.The negative examples of Reed and Weill at Citigroup and Purcell and Mack atMorgan Stanley are the un fortunate result of such "mergers of equals."Finally, theCEOs of the m erged companies rarely have a history of work-ing in a shared leadership structure and often don't have the skills or desire towork in one. Often they see the merger as just one more challenge in their life-long pursuit of becoming the sole CEO of a major corporation.Not all merger-originated forms of shared leadership are doomed to fail.At Boeing, CEO Phil Condit (originally from Boeing) and Harry Stonecipher(ex-CEO of McDonnell Douglas) have managed towork together seamlesslyafter the merger of their two giant companies (that is, after Boeing's acquisition

of McDonnell Douglas). One might expect to find that these two powerful indi-viduals had divided their responsibilities along the lines of the company's twomain businesses, commercial and defense. While that is partly true, the full storyis more com plex. The two hav e a relationship that is often described as "symbi-otic." Insiders say they work in tandem, each picking up different parts of aproblem.Condit is described as the "visionary" who provides spiritual and long-range leadership because he is seen as embodying the ideals of Boeing (he was,after all, the father of the 757, 767, and 777 projects). In contrast, Stonecipher,

coming from the hard-nosed culture of GE (by way of McDonnell Douglas),watches the numbers and holds people accountable. Yet, it is Stonecipher whohas taken the lead in the "soft" area of executive development. Even there,however, Condit plays a supportive role as well. Apparently, they have eachlearned to take a step back in the areas where the other takes the lead. Thesecret here may be that because Stonecipher is much older than Condit (andabout to retire), there is no competition betw een them for power or acclaim.The creation of co-CEOs from co-founders has a better track recordbecause the individuals involved have typically chosen each other as partners.

While they may start with a good working relationship because of their successas entrepren eurs, this doesn't m ean they can learn to become co-CEOs of a large


9/20


after having learned that he wanted no part of running a big comp any (he left tofinance rock concerts). When Paul Allen let his start-up partner Bill Gates havefull managerial control of Microsoft, hedid so because of his health problems.While Allen has remained on the company's Board of Directors, it is extremelyrare for a founder of a company to remain as a contented and productive m em-ber of the board after stepping down from a leadership position. Moretypicallyat Patagonia, for exampleco-founders end up going their separateways because irreconcilable differences about the future of a company oftendevelop shortly after an initial public offering (IPO). Sometimes such partingsare voluntary but, more often, one of the founders dominates the Board andforces the other out.Just like individual founders, co-founders often fail as co-leaders becausethe skills needed to start a company are not the same as those needed to run it.The late Robert Noyce, co-founder and first Chairman of Intel, was a rarityamong successful entrepreneurs; he understood his strengths and weaknesses(Case I). Early on, he fired himself and became Vice Chairman of the company.Fellow co-founder Gordon Moore became the Chairman (later, he turned overthe reins toano ther co-founder, A ndy G rove). Noyce was fortunate tohave aco-leader with com plemen tary skills who m he trusted and with who m he couldwork. Most entrepreneurs, in contrast, are loners who end up getting the bootfrom the board when it discovers they haven't got what it takes tobe CEOs.The most visible example of co-founders becoming successful co-CEOsis William Hew lett and David Packard. They both focused onstrategy, but theydifferentiated their other leadership roles. Bill Hewlett became the heart of th ecompan y (to this day he is the patron saint of HP's engineers). In contrast,David Packard was the guts of the organization, the hard-nosed businessmanwho made the tough calls. It was Packard who removed the non-performers.Together "Bill and D ave" were a great combination, yet HP's board did notreplace them with another set of co-leaders. Instead, the y w ere succeeded byJohn Young, a man very much in the same mold as Bill Hewlettwarm, caring,and respected by all. However, he could not manage the conflict that occurred

as a matter of course inHP's highly decentralized structure. Packard was forcedto come out of retirement and run the company while a new CEO was groomedfor the job (who also failed to live up to the high standard set by the co-founders). The lesson at HP seems tobe that failure to replace all of the manyand diverse skills of the founders has led to continual turmoil in the executivesuite. HP may be one company that requires more than two heads to run.Some companies have a tradition of shared leadership. Investment banksin general, and Goldman Sachs, inparticular, use co-leaders as a matter ofcourse, inpart as a residue of their historical partnership structures (most of

which have now been abandoned). In 1976, w hen the CEO of Goldman Sachsdied, senior partners John Whitehead and John Weinberg put together a succes-


10/20


became the name of the game at Goldman for the next twenty-five years. Typi-cally, one of the company's CEOs has come from the sales and trading side of th ehouse, and the other from investment banking.

In 1984, when Whitehead retired to join the Reagan Administration,Weinberg appointed Steven Friedman and Robert Rubin to serve as co-COOs.In 1990, they became co-CEOs, and served until Rubin joined the ClintonAdministration. They were succeeded in 1994 by anoth er set of co-CEOs, HankPaulson and John Corzine. Even though Goldman's shared leadership model haslasted for a long time, it has had its rough m om ents. In 1999, when Corzine leftthe firm, there was great controversy and debate over whether Goldman shouldgive up its partnership structure and go public, which it did. Yet, in general, theshared leadership model has worked well in investment banking where it isoften extended down the line to co-heads of major business units, such asinvestment banking and equities.The co-CEO ap proach seems towork best wh en an existing CEO createsit. For example, Charles Schwab invited his long-time collaborator, David Pot-truck, to becom e his co-CEO, and Bill Gates invited his collaborator, SteveBallmer, to join with him in various join t leadership relationships before finallyturning over the CEO job tohim. Significantly, Gates remains co-leader ofMicrosoft as Chairman of the board and head of software research. At bothSchwab and Microsoft, long histories of working together had built trust andrapport between the principals. The Schwab-Pottruck relationship showed somesigns of unraveling in the 2001 recession when Schwab took back some of th eauthority he had ceded toPottruck and began m ore directive, personal leader-ship of the company. Nonetheless, this partnership still appeared strong as thisarticle went topress in mid-2002.

Different RolesThe odds on the success of shared leadership appear togo up w hen the

individuals involved play different and complementary roles: HP's "heart" wasBill Hewlett and its "guts" David Packard. Charles Schwab is often described asthe "voice of the customer" and the "company conscience," while David Pot-truck provides task leadership. Steve Ballmer provides emotional leadership atMicrosoft, ditto John Weinberg at Goldman Sachs, and Phil Condit at Boeing.These "emotional leaders" are (or were) matched by "task leaders" Gates, White-head, and Stonecipher in their respective com panies. In the 1980s, ABB becamea world-class corporation under the economist and strategist, Percy Barnevik,and a leadership team that included a people-person (Goran Lindhal) and acomputer guy (Jurgen Centerman)both of whom, significantly, would even-tual take their turns as CEO of ABB. By having shared leadership atABB, it waseasy to switch the focus at the top to meet changing business conditions. Simi-


11/20


technical w hiz kids as Joh n M itchell, George Fischer, an d William W eisz ranoperations.

Power at the top can be distributed inmany ways: In the 1970s, SimonRamo (the "R" inTRW) worked as team lead er with several different Chairmen,CEOs, and Presidents (his highest title: Vice Chairm an in charge of innovation,a position much like the one Bill Gates occupies today atMicrosoft). Galvin, apeople manager, and Ramo, a technologist, each com pensated for his respectiveweaknesses by sharing pow er w ith o thers who se skills could help them toachieve their goals. Similarly, the charismatic and entrepreneurial Larry Ellisonfinally made a real go of Oracle in the 1990s wh en he brought in the analyticaland managerial Ray Lane. Then Ellison fired Lane in 1999, taking back thepower he had shared, much as (but far less gracefully than) Charles Schwabreclaimed power from David Pottruck. If nothin g else, these two examples illus-trate the fragility of shared leadership, and the fact that one member of everyteam of tw o is usually more equal than the other.

Because of the increasing complexity ofmost businesses, shared leader-ship has become almost a necessity when it comes to leading chang e in largeorganizations. For example, the successful transformation of England's giantAsda supermarket chain in the late 1980s was led by two individuals who bothadmit that neither could have accomplished the feat as a solo act. The analyticalHarvard MBA, Archie Norman, provided the strategic framework and financialacumen, while the more open and accessible Allen Leighton brought the peopleand marketing skills needed for the tasks at hand.Intel exemplifies the combining of different roles into a total leadershippackage: Bob Noyce was Mr. O utside, Intel's face to the world; Gordon Moorewas the long-term strategist and thinker; and Andy Grove was the short-term,han ds-on , "make you r num bers guy." Grove was volatile, while Noyce andMoore were calm and rational. Inmeetings, itwas said that Grove would eruptlike a volcano, surfacing important issues. These instances gave Noyce andMoore the o pportunity to offer their calmer and more long-term perspectives,and then Moore would lay the issue to rest with awell-reasoned summary andresolution. The meeting would continue until the next Grove eruption. In short,as a team these leaders were able tomanage and value their com plementaryskills, temperaments, and perspectives.The situation at the top of GE has been more fluid with respect tomem-bership and roles (see Case II). The Office of the Chairman atGE was createdwhen Reg Jones selected Jack Welch as Chairman along with two Vice Chair-m en towork with him. They were selected as a complementary team both interms of skills and chemistry. Indue course, Welch brought Larry Bossidy tothe Office togain oversight of GE Capital. When Bossidy later left the company,Welch added Dennis Dammerman, the former CFO, to fill the same role. Finally,Paolo Fresca was put on the team tobring international experience when GE


12/20


Case II: The General Electric ModelThe GE model has a Chairman and CEO and tw o or three Vice Chairmen.They do not use aChief Operating O fficer (CO O) .Th e model has been shaped by Jack Welch, the CEO for thelast twe nty years. Initially it was shaped by Reginald Jones, the CEO before Jack Welch.Wh e n Reg Jones was selected as CEO, his tw o com petitors for th e Chairman's ro le becamehis Vice Chairmen.These two competitors stayed on in their roles and the Office was charac-terized by a great deal of discon tent.Therefore , whe n Jones was choosing Welch, he did agreat deal of interviewing of the candidates, their peers, and the corporate staff. He wanted toknow t he chemistry and trus t betwee n the candidates.Then whe n he selected jack W elch asChairman and CE O, he simultaneously selected the Vice Chairmen with wh om Jack couldw o rk He selected a team of three to move into the O ffice based on trust and chemistry inaddition to competence.The O ffice of three w orke d fairly well at GE. One of the m embers, John Burlingame, took onsome projects such as selling Utah Intemationai and then left. He was replaced by one of Jack'sfavorites, Larry Bossidy wh o built GE Capital.The tr io of Ed Hood, Jack, and Larry workedquite effectively until Ed retired and Larry became CEO of Allied Signal. Jack has selected aseries of people to work with him in the Office. On all occasions these are people who canwo rk with Jack and the other Office members.The new m embers are also selected in part onthe needs of the O ffice and in part to free up development op portunities for young talent.When GE was promoting international growth, Paolo Fresca,the head of International, becamea Vice Chairman. Jack had little international experience and wanted the skills of the Office tobe rounded out. Paolo is now the CEO of Fiat.The Office before the a ppointm ent of Jeffrey Imelt as the new C EO consisted of DennisDammerman and Robert Wright. Dennis was brou ght in because Jack wan ted some oversighto f GE Capital after his falling ou t w ith th e fo rm er GE Capital Chairman. Dennis, the former GECFO, became the C hairman o f GE Capital and Vice Chairman of GE.The move freed up theCFO role, which w ent t o a talented young forty-year-old. Robe rt W rig ht moved to the ViceChairman role and freed up the leadership of NB C to some new ta lentThe Office concept works at GE because the members are selected to contribute neededskills and also because they can becom e a mem ber of Jack's Office team .The Office wo rkswell in part because the m embers of the Office know that they will never be the CEO of GE.Robe rt W rig ht left GE to become the CEO of Cox Communications. W he n C ox was takenoven he came back to GE to run NBC.The members of the Office are often elder statesmenw ho are probably m ore interested in leaving a legacy than getting to be CEO.The O ffice under We lch wo rked in a fiuid manner In addition to new m embers, the w or k wasconstantly reshuffled.Typically Jack took a couple of businesses that were undergoing the moststrategic change.The other members then took the other businesses and functions. Jack typi-cally to ok the strategy fun ction and Human Resources. For exam ple, if Aircra ft Engines wasundergoing a big change. Jack wanted the business to rep ort t o him with no filters in between.


13/20


R&D w ould also repor t to Jack. He wou ld be C EO and C O O for those parts of the companyundergo ing the mos t change. He w anted those businesses to be able to make decisionsquickly If this meant an other $ 100 million for R&D and Capital, so be it. After eighteen monthshad passed and Aircraft Engines was clear on its de velopm ent. Jackwo uld take charge o f Light-ing and Major Appliances.The o ther members o f the Office w ould then sort out their po rtfo-lios in turn.The re are tw o ma jor lessons from GE's Office expe rience. First, since chemistry and tru st areimpo rtant, do not leave them to chance. GE selects for chem istry and trust and manages themunder Jack We lch. Second, whe n you become a mem ber o f the Office w ith Jack, you are avalued person but you are no longer a candidate for the CEO of GE. Candidates run busi-nesses. Vice Cha irmen are valued e lder statesmen.

went as Welch needed their skills (and needed to free up their positions downthe line to test young talent).

Se lecting a Lead ership TeannIt is often forgotten that, in the process th at led to the selection of JackWelch as GE's CEO, Reg Jones (the retiring CEO) vow ed to choose a team ofleaders and not just an individual. The process Jones used hecam e a H arvard

Business School case study, and it is worth going back to it to see how he con-ducted it thorough interviews with all of the CEO candidates and with theirpeers and others.* Jones was primarily interested in issues of trust and chem-istry. He knew the skill profiles of the various candidates, hut the interpersonalrelationships between them were hidden from him in his role as Chairman. Thelesson of the case is that it is hetter to select a team of co-leaders who have rap-port and com plementary skills tha n to choose a group of talented individuals.Goldm an Sachs appea rs to use this sam e process of joint selection stressing, asJones did, the importance of a combination of talent and chemistry.Chem istry, like leadersh ip ability, is often overlooked in executivesearches.' The late Frank Wells succeeded as Michael Eisner's co-leader at Dis-ney while Michael Ovitz would later fail in the same role. Why? Not because theearlier team had done a better job divvying up power and authority, but becausemembers of the latter team competed for time "in front of the mirror" (attentionand credit in the pre ss). Erank Wells was satisfied tha t Disney's executives andboard knew he was the cerebral leader of Disney (while Eisner was the publicand emo tional leader of the co mp any). In contrast, Ovitz needed everybody's loveand adulation and quickly began to compete with Eisner.Who should choose co-leaders? Jack Welch, Bob Galvin, and Bill Gatesselected their own co-leaders; hence, there is some evidence that it is wise to


14/20


to select someone with whom to share the leadership of Disney. However, hav -ing the current leader make the selection is most likely a necessary, if not suffi-cient, condition for success.W orking Together

Once selected, co-leaders need to learn to work together. Eor the sakeof accountability, tasks must be divided. However, perhaps more importantthan the division of tasks, co-leaders need to learn how tohandle the divisionof credit. The hardest thing for individuals considering joint governance tounderstand is that their biggest challenge is not practical or technical. Instead,it is managing their egos: Can one of them step back and let the other take thebows? Can they come onstage and take their bows together? In the final analy-sis, co-leadership has worked at Intel and TIAA-CREE because executives atthose companies are able to share credit, and it has failed atDisney andCitigroup because of the egos that were rampant in the executive suites ofthose companies.In addition tomanaging the allocation of credit, the success of sharedleadership depend s onhow effectively the individuals involved communicate,han dle crises, allocate and reallocate join t tasks and decision mak ing, anddevelop common positions on key issues. The roles and tasks at the top can be

divided invarious ways: Mr. Inside and Mr. Outside; Mr. Business Line A andMs. Business Line B; Ms. O perations a nd Ms. Acquisitions. A lternatively, taskscan be divided by interests (innov ation vs. operation s), skills (technology vs.people), or personality bent (strategy vs. implementation). Indeed, roles andtasks can be divided along as many lines as ther e are individual skills and inter-ests onone axis and organizational needs and opportunities on the other.In some cases, the most effective approach to task division is tohave afluid approach towho takes responsibility for leadership tasks. Eor example, atTIAA-CREE (managers of over $600 billion inpension assets), CEO John Biggsonce had the investm ents function reporting tohim when he shared leadershipwith a president wh o ran operations (who has since left the com pany). How-ever, no w Biggs manages strategy, new businesses, and extern al affairs wh ilehis co-leader, Vice-Chairman Martin Liebowitz, runs equity investments (wheremost of the risk is in the business). Wh at is impo rtant to understand is that titlesand turf are relatively unimportant to both the two leaders of TIAA-CREE and totheir followersBiggs and Liebowitz are co-leaders who share equally in thecredit and blame that is attached to the executive suite.In the final analysis, it doesn't matter so much how responsibilities aredivided, as itmatters that the individuals involved are clear about their roles andhonest with themselves and each other about their respective contributions andfor


15/20


How are w e going tocoordinate and communicate with each other so we don'tstep oneach other's toes? How can we make sure we send the same message?"Coordination and alignment begin with communication. When there are

only two co-leaders, comm unication often takes place spontaneously and con-stantly. Our colleague Jay Conger reports that Schwab and Pottruck were intouch continually throughout the day. However, some sharing of thoughts andideas needs to be formalized. In the case of the Midwest Manufacturing Com-pany (see Case III), the co-CEOs have a monthly dinner scheduled indelibly ontheir calendars. Very few co-leaders are as well-coordinated and aligned as wasthe team of Noyce, Moore, and Grove at Intel (see Case I). The trio would meetevery Friday m orning to share their thoughts, debate issues, develop commonpositions, and plan the agenda for an Executive Staff Meeting that took placeevery Tuesday. The opposite was true at Citigroup where insufficient communi-cation among the co-leaders led to an irreparable divergence of views and opin-ions. Itwas reported in the press that Reed and Weill actually avoided eachother. Subordinates accentuated the rift by "shopping ideas" with the two lead-ers, which led to further conflict and a resulting decay in the level of trust.

Wo rking W ith OthersThe division of tasks among co-leaders often affects their subordinates'work. For example, questions arise concerning attendance at executive teamforums: Whose meeting is it? Who should attend what meeting? Such questionsare relatively easy toanswer at companies organized like Amana where the fourco-leaders each head s a relatively independent husiness unit. At the otherextreme is the "Midwest Manufacturing Com pany" (Case III) wh ere th e co-leaders have divided responsibility based on their interests and skills. Becauseco-CEO "Sue" looks after all sales and distribution activities, including thosein the businesses run by her co-leader, this makes everyone else's work highlyinterdependent. Issues of responsibility for inventory, prices, and customer prior-ities constantly need to be resolved. Because the co-CEOs did not involve others

in their decisions concerning the allocation of their tasks, the responsibilities ofothers had to be worked out in the heat of day-to -day issues. This was clearlythe wrong way go. Co-leaders need to discuss these issues and resolve the ambi-guities themselves, rather than leaving them to their subordinates. Managersshould be spending their time dealing with business issues not with politicalissues caused by the co-leaders' role ambiguities. The lesson is this: the moreinterdependent the work of co-leaders the more input they should solicit fromaffected others, and the more they need to coordinate between themselves.

Leadership Institutionalized


16/20


Case III: Midwest Manufacturing Connpany"Midwest Manufacturing Company" (a pseudonym) is a 5,000-employee equipment manufac-turer. It sells its products through distributors and at the time of this study had seven differentproduct lines. It also has the typical functions of sales, marketing, R&D, finance, HR, IS, and legal.The company is an old one and is family owned. A t the tim e of this study Midwest Manufactur-ing had recently shifted its organization design. In the old structure, the co-CEOs "John andSue" (pseudonyms fo r the children o f the founder) were in a two-in-a-b ox arrangement Theydid virtually nothing separately and all reporting relationships were to both of them.Therewere some differences in how they spent their time since Sue seemed to be particularly skilledat dealing with distributors and enjoyed meeting customers more than John.This led them to,change from their two-in-a-b ox structure to one in which they had separate repor tingrelationships.Rather than go to a C O O /C EO reporting s tructure, each of them t oo k certain lines of busi-nesses and certain s upp ort functions.The ones they to ok seemed largely to be dictated bytheir preferences and knowledge about certain areas rather than the interdependency thatmight ex ist among the functions and business units. In essence, they decided to take prima ryresponsibility for things they knew the most ab out and most wa nted to manage. W he n th echange occurred , little input was sought from oth er mem bers of the organization. In essence.Sue and John were th e major owne rs o f the company and little resistance occurred. John andSue have maintained a good personal relationship. Indeed, once a m onth th ey mee t w ith theirrespective spouses to talk about how well they are working together and the impact of theirwork ing together on their spouses.The most important reason for the change was efficient use of t ime. Both Sue and John feltthat they were simply extra wheels at certain meetings and that they could collectively havemore influence and provide more expertise to the organization if they operated indepen-dently in certain situations.This, of course, is also consistent with their having different skills anddifferent interests. Further there were some mem bers of their direct re por t group that pre-ferred w orking w ith one o f them so the new structure was seen as a positive by them.The major advantage of moving to the new structure was better utilization of both John andSue's tim e and skills.They simply could do m ore and could be m ore com fortable wi th the newstructure. In general, their direct reports preferred the new relationship because it gave themmore opportunity to get guidance and discuss issues with one of the leaders of theorganization.The major weakness of the new structure was the tendency of subordinates to shop decisions.Even though there was a clear reporting relationship there often was enough ambiguity so thatindividuals could take issues to either John or Sue. N o t surprisingly, they preferred to takethem to the one they thought was most likely to give them the decision they wanted. In somecases, they too k it to both if they didn't like the answer from th e otherAnother problem with the direct reports concerned communication between Sue and John.


17/20


question of whether Sue explained to John what decision had been made (or vice versa).Because o f amb iguity in this area, subordinates ofte n felt like they had to explain Sue's deci-sions to John (and vice versa) when only one of them was present at the meeting where adecision was made.Finally the situation led to some increased com petition betwee n John and Sue. Both wan tedto per form the ir jobs well so as result they ten ded t o push the ir divisions and the ir functionsparticularly hard.They also more often felt left out of issues and decisions and raised issuesabout wh ethe r they should have been included in a particular me eting that might have someimplications f or th e areas for w hich they we re responsible.When Sue and John changed to the new structure they did iittle work in the area of groupprocesses and communication. Because of this, their subordinates were often unclear as tohow John and Sue planned to w or k toge ther and indeed how they should treat them in termsof decision making and comm unication. It was often not clear what was a group activity andwh at was a one-on -one activity, particularly given the history of Sue and John w ork ingtoge ther and dealing w ith v irtually e very issue. Much of this could have been resolved if thetop management team had spent some time talking about their decision process and commu-nication.The structure made co ordina tion betwee n John and Sue particularly critical. By making differ-ent assignments of reporting relationships to both Sue and John it would have been possibleto reduce the importance of coordinating their activities. In essence, the reporting relationshipsto the m w ou ld need to be based on the degree to which functions and business units areinterde pen dent The objective wo uld be to identify tw o groups of relatively independent func-tions and business units.As John and Sue developed the new rep ortin g relationships, they te nde d t o speak less and lessfrequently as united leaders.This began to cause problems and certainly c ontrib uted to thetendency of their subordinates to decision shop.The c ritical learning is that wh en leadershiptakes the form that it does in Midwest Manufacturing, the two leaders need to be sure thatthey frequently make joint presentations and representations of what the company is about,what its basic strategy is, and wha t it stands fo rOverall, the structure tha t Midwest Manufacturing used was judged to be effective. Moving t odifferent reporting relationships for John and Sue did seem to represent an improvement overthe earlier structure. An important note of caution is appropriate here, howevenThe successof any co-leader approach very much depends on the relationship between the two individu-als and how it is perceived by the others in the organization. In the case of John and Sue, it wasperceived that they were fundamentally on the same page with respect to the direction of thecompany; and since they had different skills, having different reporting relationships passed thecredibility test.


18/20


these companies, we discovered that man y of the key tasks and responsibilities ofleadership were institutionalized in the systems, p ractices, and cultures of the organiza-tion.''" Without the presence of a high-profile leader (or "superiors" goading orexhorting them on), people at all levels in these organizations:

acted mo re like own ers and entrep rene urs than employees or hired hand s(they assumed owner-like responsibility for financial performance andmanaging risk); took initiative to solve problems and to act, in genera l, with a sense ofurgency; willingly accepted accoun tability for me eting com m itmen ts, and for livingthe values of the organization; and created, maintained, an d adhered to systems and procedures designed to

measure and reward the above behaviors.Among other things, this discovery helps to explain some persistent con-tradictions to the dominant model of leadership: If leadership were solely anindividual trait why is it that some companies are able to renew their strategiesand products and outperform competition in their industries over the tenures ofseveral different chief executives? Why is it that some CEOs who have succeeded inone organization often turn in so-so performances in the next? It is our conclu-sion that the reason is found in such key organizational variables as systems,structures, and policiesfactors that are not included in research based on a solo

leadership model.Conclusions

Despite continued assertions that co-leadership does not work, there areenough examples of such successful combinations to reject that conventionalwisdom. Instead, the issue is to identify the factors that improve the odds that aparticular combination of leaders will succeed. Joint selection, complimentaryskills and em otional orientations, and m echanisms for coordination are am ongthose key factors. While these may seem to be common sense, they are not com-mon practice. Research and analysis should continue to test these factors andothers. It is important to do so because collective leadership is here to stay.

Notes1. D. Hee nan and W. Benn is , Co-Leaders: The Power of Great Partnerships (New York,NY: John Wiley & Sons, 1999).2. J. O'Toole, "When Leadership Is an Organizational Trait," in W. Bennis, G. M.Spreitzer, and T. G. Cummings, eds.. The Future of Leadership (San Francisco, CA:

Jossey-Bass, 2001).3. J. O'Toole, Leadership A to Z:A Guide for the Appropriately Ambitious (San Francisco,


19/20


5. Heenan and Bennis, op. cit.6. R. Townsend, Up the Organization (New York: Knopf, 1970).7. R. Pascale, Managing on the Edge: How the Smartest Companies Use Conflict to Stay

Ahead (New York, NY: Simon & Schuster, 1990).8. CA. Bartlett, F.J. Aguilar, and K.W. Elderkin, "General Electric: Reg Jones andJack Welch," Harvard Business School Case #9-391-144, 1991.9. W. Bennis and J. O'Toole, "Don't Hire the Wrong CEO," Harvard Business Review,

78/3 (May/June 2000): 170-176.10. O'Toole (2001), op. cit.


20/20

Documents

When Two Heads Are Better Than