A Study of the Service Industry—Functions, … Schumpeter tracedthegrowthandtheimpactoftechnologiesover thelasttwocenturies.(Figs.3,4).Theﬁguresshowthe ... (TheEconomist,Dec.31,

IEICE TRANS. COMMUN., VOL.E83–B, NO.5 MAY 2000885

INVITED PAPER IEICE/IEEE Joint Special Issue on Autonomous Decentralized Systems

A Study of the Service Industry—Functions, Features and

Control

Chitoor V. RAMAMOORTHY†, Nonmember

SUMMARY We study the evolution and the dominance ofthe service based functions and their distinguishing features. Asthe service industry matures, its functions bear many similaritieswith the software development processes, such as intense man-machine interaction, knowledge intensive activities, flexibility inthe organization, control and execution of tasks. In this paperwe discuss wide range of interconnected topics, emphasizing themulti-faceted nature of service functions. These include the evo-lution of service industry and their products, the consumeriza-tion of high tech products based on their large-scale adoption andthe consequent creation of implicit requirements; the technologytransfer processes; the error proneness due to intense and pro-longed interaction with computers and some methods of mitigat-ing error incidence. We argue that by proper ‘humanization andpersonalization’ of interactive systems and by the use of teamsof computer supported professionals, we can prevent such errors.We discuss some useful team types, models of their behavior andtheir control aspects. As the cost of communications shrinks likedue to the Internet, we conclude that a fully decentralized systemcontrol provides a flat, flexible, and fair and friction-free organi-zation for large team based service systems.key words: service engineering, service industry, software en-

gineering

1. Introduction

The service industry, whether it is travel, leisure, en-tertainment or finance, involves personalized activitiesrequiring interaction and intervention between humansand machines. The U.S. Bureau of Labor Statisticsestimates that more than two thirds of all workers inU.S. are involved in service functions at present andtheir numbers are increasing rapidly (Fig. 1(a)).

The dictionary defines ‘service’ as an organizedsystem of appliances, products, personnel and other re-sources to supply activities needed to satisfy a publicor private need. Insurance, banking, catering, lodg-ing, travel and entertainment activities are tradition-ally considered as service industries. TV repair ser-vice, product maintenance services, and even massageservice in recent times are listed as service functions.However, the goals and the meaning of the service func-tions have broadened. In modern parlance, service isthe work performed directly or indirectly to satisfy theneeds required by customers. Service employees of-ten supported by such resources as telephones, mes-saging devices, TV, computers, etc. that help support

Manuscript received January 18, 2000.†The author is with Computer Science Division, Univer-

sity of California, Berkeley, CA, USA.

their interaction with customers. The service activi-ties are generally initiated by customers and generallycontrolled by them.

Since the industrial revolution in Europe manuallyintensive functions have given way to mentally or intel-lectually intensive activities. As a part of this trans-formation, we have seen an increasing dependency onthe high technology, actually information technology,mostly on PC’s, and networks. The manual chores havereduced and are replaced by automated devices androbots. Software agents and tools support the mentalactivities.

Service functions are delivered or discharged byservice providers at the request of customers. Serviceproviders who cater to customer needs and requests sat-isfy the preferences, needs and conveniences sought bycustomers. Services must be easy to call on, responsiveand dependable. They must be competitive in costs,easy and reliable to use. Service offerings must be upto date. They must pour out new products, keeping upwith customer’s insatiable frenzy for functional noveltyand technological curiosity. The service providers indi-vidualize/customize the service packages and productsto suit their customers, collectively, selectively and in-dividually, where possible.

The service providers often may work at homelike their clients, reminiscent of the ancient crafts-man of pre-industrial Europe, who practiced his craftfrom home. Service providers generally work in smallteams—some co-located and some distributed in des-perate places (virtual teams) but all work collabora-tively or cooperatively as members of a single team.‘Small service firms with fewer than 100 workers ac-count for one half of the American workforce and areresponsible for 2 out of 3 new jobs. These servicefirms seldom have more than 3 layers of hierarchy—Theworkers are provided with more variety and responsibil-ity (through self governing teams and the like).’ (TheEconomist, Jan. 29, 00)

Service functions are major components of everyindustry. Examples of service function-dominated in-dustries include finance, travel, telephone and commu-nications, power and other utilities, health care, enter-tainment etc.

These mentally (knowledge) intensive activities areshown to be error prone if the operator or user isstressed emotionally or physically. Studies by Andersen

886IEICE TRANS. COMMUN., VOL.E83–B, NO.5 MAY 2000

(a)

(b)

(c)

Fig. 1 (a) The growth of service industry functions in US. (b)Service function. (c) Evolution of service functions and serviceorganizations.

Consulting and others have shown that about 40–80%of the failures may be due to improper human reac-tion due to unnatural circumstances, including lack ofknowledge, stress and confusion created by informationoverload and abundance of useless information (com-monly known as data smog).

Service functions aim to satisfy and facilitate thegoals of their customers. In doing so, service providers

design their service functions (also called, service tasks)so that the customer feels comfortable and convenientin using them. The service providers try to precon-dition, specialize or personalize their offerings to sat-isfy the customer’s needs, desires, preferences and fan-cies. This personalization or preconditioning is basedon or derived from user’s requests, perceived needs andgoals primarily through interactions. The interactionsare between humans (with their support resources likecomputers, networks, etc.), between teams of humansand the application environment. Customers and userscommunicate their service needs through their interac-tions. Preconditioning of service functions may also bebased on other aspects such as community, environmentetc., but we shall consider not them in this discussion.

The dictionary defines interaction as ‘a two waycommunication between source of information and auser who can initiate or respond to queries.’ Every ser-vice function has a ‘convenience’ dimension, i.e., adapt-ability to help users in facilitating towards their goals,but also be a silent monitor who observes the user’s ac-tion so that it could serve the user better in the nextsequence of tasks. Therefore preconditioning is basedon ongoing dialog (interactive session) between the userand the servicing system.

We classify the interactions into three categories:1) interactions between service functions and individu-als. 2) interactions between service functions and teamsand 3) interactions between service functions and cus-tomer applications.

By personalizing and humanizing our service prod-ucts we enhance users’ convenience and reduce the inci-dence of errors and thereby improve their productivityand quality of their work.

In the case of interactions between and amongstteams, we can show that decentralized form of organi-zation and control facilitates the effectiveness of teams.This is somewhat similar to personalization of servicefunctions and service products for the individual cus-tomer. Decentralization of teams provides a flat orga-nization (or something close to it, like the 3-layers ofhierarchy as quoted from The Economist above) withthe assumption that as communications costs becomenegligible there would be lesser need for rigorous hierar-chical or centralized control. Therefore there would bemore encouragement for individual initiatives and ex-change of ideas during decision making and all of whichfacilitating sharing, learning and creation of knowledge.

The third type of interaction is between the servicefunction and the application objectives. Here the ser-vice products are preconditioned to provide high perfor-mance, reliability and safety to that specific spectrumof applications. We shall discuss these in the subse-quent sections.

The technology evolution has seen manually inten-sive tasks being replaced by a combination of mentallyand manually intensive tasks which in turn being re-

RAMAMOORTHY: A STUDY OF THE SERVICE INDUSTRY—FUNCTIONS, FEATURES AND CONTROL887

placed by a combination of mental and interactive tasksFig. 1(c). Currently, some of the services are possiblyperformed by individuals at their homes. Figure 1(c)also depicts the evolution of team organization andtheir control. The service system moves from an iso-lated cluster of non interacting entities, towards a cen-tralized, then to a partially decentralized and finallyto a fully decentralized fully connected entity. Whenthe last stage of evolution is reached, the system is or-ganized as a flat (one layer) organization where eachcomponent or entity may act by itself in full agreementwith other members of the system by mutual consulta-tion or rules of behavior protocols. We shall defer ourdiscussion to later sections.

In this paper, we study, very briefly, the evolu-tion of the service industry and the role played by au-tomation We shall also touch on the product improve-ments by customization, humanization and personal-ization that support service functions.. We shall alsocomment on the process of technology transfer, and theevolution of information technology products that sup-port the service functions. We show that this evolutionhas created a new set of requirements, called the im-plicit requirements.

Since many service activities involve teams of peo-ple and machines, we shall study some significant trendsin the organization and control of service teams. Lastlywe develop some models for team-oriented systems andcomment on their usefulness. We shall show that decen-tralization is the growing trend of team-based servicesystems of the future.

2. Distinguishing Features of Service Func-tions

The distinguishing characteristics of service functionsinclude the following:

Human (customer) needs drivenKnowledge intensive, high mental effortAutomation intensive to reduce manual effortHuman interaction intensiveInformation technology intensive andTeam based.

We shall next comment on some of these.Since they cater to human needs, the service func-

tions are human-machine interaction intensive, knowl-edge intensive or manual.. They are mostly information(software) technology driven. Teams execute most ofthe functions. Teams could include clusters of humanprofessionals and computer resources.

There exist several parallels between the serviceindustry and the software industry. They both arehighly knowledge oriented and are heavily dependenton collaborative efforts between humans and their net-worked computers. Collaboration is needed because thesize, complexity and intellectual intensity of the service

tasks. Just like software development operations, theservice operations tend to be flexible, easily modifiableand requiring to variable lengths times for completion.Also, the modern day service functions are focussed onfacilitating the goals their clients and these goals canchange as the services are being rendered. By contrast,some service functions may involve manually intensiveoperations, whereas the software functions do not.

2.1 Human Needs Driven

The human needs can be of three types:

Personal—humanistic and individual.ProfessionalSocietal (community)

One major personal need is to make the machines reactmore like humans so that the operator or user is alwayscomfortable, and not stressed, surprised or confused atany time particularly in an emergency. We call this the‘humanization’ process. Individualized personal needsmay depend on our disabilities, such as eye and hear-ing impairments, left handedness, short attention spansetc. They may also depend on our personal preferencesand styles. We call this ‘individualization.’

Professional needs of service functions call for ex-celling in the job, and improving one’s productivity andperformance at work. This generally requires customiz-ing work equipment and the work place to fit the ap-plication and the employee so that the performance ofthe system and the productivity of the professional areimproved.

Societal needs of the service functions emanatefrom community and environmental consciousness andresponsibilities, thereby eschewing altruism—befittingthe theme of ‘The Three Musketeers,’ namely, ‘all forone and one for all.’

2.2 Knowledge-Technology Intensive

Service activities generally involve choosing among op-tions, and solving problems with the help of comput-ers, technology and knowledge. By knowledge, wemean primarily technology oriented scientific knowl-edge. Davis and Boskin [1] have conjectured that sci-entific knowledge grows exponentially, doubling everyseven years. It raises the technological and educationallevels with its growth. By Ross Ashby’s law [2], this hasthe rising tide effect; that is, the rising tide (of knowl-edge) lifts all the boats (improvements in functionalityand quality of service). Specifically, as customers be-come more educated and familiar with new technolo-gies, they demand better, faster and more intelligentservices and products. Instead of walking or request-ing a rickshaw to take her from San Francisco to LosAngeles, the customer likes to fly in a jet plane at thecheapest fare.


Knowledge growth implies a parallel growth intechnology from which it is derived. This is evi-dent from the semiconductor (primarily the microchip)technology’s exponential growth as postulated by theMoore’s law, which stipulates that chip performancedoubles every eighteen months while the cost remainsthe same.

The streams from the knowledge generation pro-cess flow across the knowledge-technology-services-products chain, or the technology transfer chain—allthe way from knowledge creators, technology develop-ers, service providers, product developers, to fat, thin,and famishing clients and users—as products and ser-vices.

The driving forces in technology belong to threecategories namely, the current technology, the support-ive and integratable (collaborative) technologies, andthe so-called disruptive technologies. We follow simi-lar arguments like those of Clayton Christensen (‘TheInnovator’s Dilemma. When New Technologies CauseGreat Firms to Fail,’ Harvard Business School Press,1997). As the services and their related products growlike in the semiconductor industry, the supportive andintegratable technologies will help to sustain or boostthe current technology by enhancing its native perfor-mance and capabilities. Some examples include thecopper wire technology, which has enhanced the semi-conductor microchip performance, and the laser tech-nology, which supports the established semiconduc-tor technology by providing important computer ad-juncts like printers and fiber optic networking. Thusthe supportive and integratable technologies lengthenthe life (strangle hold?) of the current technology be-yond its physical limits because a changeover couldmean another Y2K problem, (or will it be a fizzle or aChernobyl?—that will be the question!) but more mas-sive disaster. A disruptive technology such as the onebased on superconductivity or quantum physics mayhave a devastating effect on the current technology. Anexample from the olden times is the uprooting of thevacuum tube and the magnetic core memory technolo-gies by the invention of the transistor, which heraldedthe current semiconductor industry.

2.3 Human Interaction Intensive

There exist many types of human interactions; viz.,amongst humans, between humans and machines, andwithin and across teams of humans and machines. Wewill be concerned with two primary types of interac-tions, one of which is knowledge intensive and the other,heavily oriented towards simple keyboard interactionsand bookkeeping. Knowledge intensive interactions arethose in which humans have to think and interact withthe computer, like when one composes a report or aletter on the computer. We shall simply label this as aknowledge intensive (K) activity since mental activity

is primary and machine interaction is secondary. Onthe other hand, when someone does pure transcriptionlike copying a letter onto a terminal for transmission,this is considered as a simple machine interaction, orsimply an interactive bookkeeping operation (IB).

2.4 Information Technology Intensive

The service functions depend on very heavily on in-formation (computer and communications) technology(IT) with the software discipline as the major player.Every aspect of the service function from customer re-quests to service delivery, and service evaluation bycustomer satisfaction-metrics is IT driven. There arestriking similarities between software development pro-cesses and service functions. These include their knowl-edge dependency, their collaborative development andthe ease and flexibility in modification and execution.Just like the software activities, the service functionsshow great variability of interactions, namely, human-human, human-machine, team-team etc. The two dis-ciplines depend on decisions and choices made by usersand developers but always facilitating the goals theclients would want to achieve.

2.5 Automation Intensive

Automation is the replacement (or substantial reduc-tion) of manual, mental and human interaction effortsby automated systems, computer and knowledge basedmechanisms. Automation includes the use of intelli-gent robots to tend the manual service tasks, and in-telligent software agents to take care of intellectual ormental functions. As we shall see later, the evolution ofservice industry and its functions parallels the growthof the automation technologies. Automation has pro-duced in its wake momentous improvements in produc-tivity, performance, and product and service quality.But most significantly, automation has been very ag-ile and resilient in accepting new ideas and new tech-nologies rapidly—thus accelerating the creation of newproduct families and facilitating their use.

Manual (physiological) effort is tiresome andaccident-prone. Automation reduces manual effort andthereby helps prevent human induced failures. Simi-larly, computers, knowledge-based software tools anddecision support systems relieve intense mental (psy-chological) effort and mitigate error incidence in men-tally intensive activities.

2.6 Teams and Teaming

Teams of people supported by machines who work col-laboratively or cooperatively perform service functions.In serial chains, or assembly lines, teams work on a taskin a sequential fashion. Teams can also be structuredto perform parallel operations functions in the form of


parallel threads. We shall discuss about teams in a latersection.

3. Service Functions

Service functions are embedded in all industrial, busi-ness and economic functions, and indeed, in every hu-man endeavor. Included amongst them would be theadministrative services, and bookkeeping activities thatare prominent in business and industrial enterprises,and the maintenance services that are primarily asso-ciated with complex manufactured products. We canclassify the activities into four principal categories. Thefirst category is mostly manual (M). The second ismostly man-machine interactive but requiring no in-tensive mental effort. We include in this the book-keeping and administrative activities which generallyrequire simple interactions with computers. We shalllabel them as IB (Interaction-Bookkeeping). The thirdcategory is mostly mental or knowledge intensive ac-tivity (K) which includes human to human discourses,man-machine interactions requiring intellectual effortlike playing chess with the computer or composing aletter on the computer terminal. Human thinking anddecision making are central in this type of activity. Ex-amples include planning, searching, selection and find-ing closest matches to the desired routing, and develop-ing cost effective schedules subject to customer’s desiresThe fourth type of activity is associated with networkedtransactions (NT). This would involve accessing web-sites, databases and performing data-processing func-tions as well as Internet and web-based activities. Theinteraction based activities (IB), the knowledge basedactivities (K) and manual activity (M) take variablelengths of time for execution. The execution times fornetwork-transaction (NT) type of activities will dependon the vagaries of database processing times, delaysdue to network congestion and transmission. But for-tunately, they can be estimated reasonably well. Book-keeping and administrative operations or the IB typeof activities can be done in parallel with other typesof service functions with which they are generally over-lapped and interleaved.

We shall illustrate the service activities by meansof a simple example. The example does not include anymanual operations. A customer decides to take a tripand interacts with his travel agent to discover the bestchoices for the journey. The customer provides a list ofspecifications and constraints such as the places he likesto visit, preferred travel dates, his budget etc. This isa knowledge intensive (K) activity. We shall call thisStep 1. The travel agent consults an air line reservationsystem to provide a list of feasible options. The data en-try part is an interaction-bookkeeping (IB) operation.This is Step 2. Next, the travel agent accesses the air-line reservation system and she receives a list of feasibletravel choices. This (Step 3) is a networked transaction-

Fig. 2 The sequence of service operations in air line reservationexample.

based (NT) operation Then the customer makes hischoice. This is a knowledge intensive (K) operation.This is Step 4. The agent enters the customer’s choiceinto the networked airline reservation system. This isStep 5 and is a data entry (interaction-bookkeeping, IB)operation. The reservation system updates its recordsand confirms the customer’s itinerary. This is Step 6and is a network transaction (NT). The customer thenpays for the fare using his credit card. The travel agententers the payment information into the system. This isStep 7 and is an IB operation. The reservation systementers this into its file. This is Step 8 and is a networktransaction (NT). The agent confirms the itinerary andprints it out. This is Step 9 and is an interaction-bookkeeping (IB) activity. We have broken down thisservice transaction into a number of simple steps.

We shall illustrate the sequence of activities by asimple Gant-chart like sequence diagram (Fig. 2).

As an example of service using the Internet-Webdomain, consider a typical transaction of the on-linebookseller, amazon.com. When the company acceptsa book purchase order, it performs several interactive-bookkeeping (IB) and Network-Transaction (NT) op-erations. We shall discuss only a few. It performs anetwork transaction (NT) on the customer’s credit cardaccount, transmits the book dispatch order (NT) to thewarehouse. It also requests UPS to do the pick-up anddelivery to the customer’s address. This is an NT typeactivity. The delivery of the book by UPS is a mostlymanual activity (M). We will not go into details of thisfunction. We can represent all the service componentsin this ‘e-tailing’ transaction using the activity elementswe discussed above. In parallel, the company may alsorecord customer’s book preferences, tastes for topicsand authors for its future business. It may also mon-itor the performance of their service personnel. Theseare all parallel bookkeeping (IB-type) operations.

4. Evolution of Service Industry

The Austrian born American economist, Schumpetertraced the growth and the impact of technologies overthe last two centuries. (Figs. 3, 4). The figures show thewaves of innovation of dominant technologies over the


Fig. 3 The growth of new technologies (Europe).

Fig. 4 The growth of service industry and automation.

last two centuries. We have included in it the informa-tion technology industry, encompassing the computer,communication and artificial intelligence technologies.Each technology growth wave rises out fast, matures,and then declines, just when the next technology inno-vation wave starts to move up and takes hold. The in-teresting fact is each technology spawns new industriesless and less manually intensive than its predecessorsand accepts more automation as it matures and there-fore requires less manual labor than it started out with.This also implies a tremendous growth in productiv-ity. Figure 4 shows the marked decline of manual effortand the growth of the service oriented functions andthe growth of automation as a function of time. Onecan attribute the decline of manually intensive func-tions to the surge of automation. Service industriesseem to have benefited by the advances in the informa-tion technology, which is, by far, the most importantenabler of manual and mental automation. Economiesof U.S., Japan, Singapore, and Western Europe seemto fit well with the Schumpeter’s model.

The rapid growth of the service component of theindustries implies an increase in activities largely de-pendent on manual, and mental interactions with ma-chines. This, unfortunately, may provoke human gen-erated errors. Product design and development is com-

Fig. 5 Knowledge technology transfer (e.g., semiconductorindustry).

plex and mentally intensive activities partly becauseit is cross disciplinary. Products and product familieshave to be periodically enhanced by adding new fea-tures and upgrading the technology to meet the chal-lenges of competition. At the same time, the design andthe operational life of the product must be lengthenedand the costs must be kept low. These activities, par-ticularly in the software area, come under the bannerof maintenance. These could be horrendously detailedand mentally taxing, therefore become bug ridden andveritable sources of future problems.

5. Knowledge Utilization and TechnologyTransfer

The delivery of services depends upon products, pro-cedures and human interactions. Service products areproduced by technology. Human needs, knowledge andexperience create technology. Knowledge, in our con-text, implies a specific aspect of the discipline that isneeded to satisfy a product technology. The ultimateaim of technology innovation is new product creationto satisfy some human need. Knowledge utilizationand technology transfer is an important part of thisprocess, which we consider as a very important servicefunction that supports human progress. It consists ofthe following four broad phases: a) knowledge genera-tion, b) technology derivation to satisfy a broad arrayof human needs and c) new product family specifica-tion, and the development of design methodologies, andthe related computer-aided tool development and d)product implementation, and manufacturing. For thesake of brevity, we have omitted important steps liketesting, product quality control, marketing etc. Fig-ure 5 illustrates the essential phases of this process anddisplays the processes associated with the growth ofknowledge, technology, tool development, and imple-mentation, over time. We have extended the graphicaldepiction of Dr. Wilfred Corrigan, CEO of LSI Logic,


and we have augmented it by adding the knowledgegeneration (creation) phase and introducing the con-cept of phase latency. We define latency as the timedifference between two successive phases during a spe-cific instance of growth of a technology (Fig. 5). For ex-ample, the latency between knowledge creation and thecorresponding technology development at some specificinstance of growth is x1 (Fig. 5). Latencies, in practice,cannot be measured precisely, but only estimated basedon experience. As shown in the Figs. 5 and 6, the laten-cies in the semiconductor industry between successivephases seem to follow a transitive relationship, whichis x1 < x2 < x3. To phrase this in words, the time totransfer knowledge into technology (x1) is shorter thanthe time to develop support tools from technology (x2),and which is shorter than time to use the tool to developthe product (x3). However, this may not be true for alltechnologies. Sometimes, innovations in one technol-ogy may have to wait for a long time for an enablingtechnology to evolve before marketable products can berealized.

6. The Kozmetsky Effect

Prof. George Kozmetsky of the University of Texas,Austin, an internationally acclaimed business leaderand a National Medal of Technology recipient, hasshown the existence of a strong interaction betweenknowledge and technology growths. According to him,the rapidity of the knowledge growth exerts a strongsynergistic attraction on the technology growth andpulls technology growth curve towards it (Fig. 6). Inother words, the technology transfer latencies get re-duced or compressed with a maturing technology. Or,the technology transfer time gets shortened with a ma-ture technology. As alluded earlier, there are two rea-sons for this. First, knowledge and technology interactvigorously, thus accelerating the corresponding tool de-velopment. The knowledge technology curves get closerand closer to the tool development and implementa-tion curves. Secondly, as knowledge saturation in aparticular technology takes place due to physical lim-its, we only encounter incremental advances but no bigbreak-through. Thus the tools and the manufacturingprocesses may not change significantly In effect, thetechnology progresses ever so slowly and incrementally.Due to this convergence of the knowledge-technologytransfer curves the phase latencies shrink and may ul-timately coalesce. This implies that the technology hasreached a saturation point and progress will be diffi-cult, like trying to squeeze water out of a desert rock.We shall call this convergence of knowledge-technology-transfer phases as the Kozmetsky Effect. This createsan inflection point in the technology growth curves sim-ilar to the Dr. Grove’s inflection points [3] in the growthcurves of hi-tech enterprises. (The Economist, Dec. 31,99 defines the inflection point as ‘an event with poten-

(a)

(b)

(c)

Fig. 6 (a) The compression of the technology transfer phases(the Kozmetsky effect). (b) The convergence of technology trans-fer phases. (c) Technology growth inflection point.

tial to change competitive landscape so fundamentallyso nothing can be same again.’)

Tight marketing deadlines and product releaseschedules may hurt the quality of the final system(product or service) because of the haste associatedwith the development. In the service industry, this canbe inadequate preparation in the new training coursesor maintenance procedures. In the software industry,


tight deadlines are often blamed for inadequate prod-uct testing, and therefore, for the dismal quality andfailures of the product in the field caused by residualerrors still remaining in the system. This may explainwhy our high tech products undergo several incremen-tal corrections—humorously called revisions, versionsand releases—over their product lifetime.

We conclude this section by commenting on twomajor characteristics of knowledge-technology syner-gism. First as stated earlier, knowledge doubles ev-ery seven years, i.e., knowledge grows exponentially[1]. Secondly, technology too may be similarly grow-ing in an exponential fashion. Moore’s law, associatedwith the semiconductor technology, hypothesizes thatthe performance of semiconductor chips doubles every18 months while the cost remains the same, and there-fore, follows an exponential growth. One sees the sim-ilarity and the relationship between these two empiri-cal laws. But Moore’s law cannot be sustained indefi-nitely due to physical limitations Thus, as the technol-ogy growth reaches its peak and begins to slow down, anew fledgling technology may take over. We can justlysay that ‘old technologies, like old soldiers, never die,but they are just left behind’—a saying attributed toDr. Andrew Grove of Intel.

7. High Tech System Evolution

In this section we shall consider the evolution of hightech service products like microprocessors or applica-tion programs that are the harbingers of the new ser-vice era. These products are initially produced in smallquantities and sold at high costs. They are consid-ered as ‘commercial products.’ They may become verypopular, and create a large consumer demand. Theseare, then, produced in large quantities and sold cheaply.They are called ‘consumer items.’ Consumer items areimplicitly subject to stricter safety; reliability, inter-operability (open-systems) requirements and environ-mental constraints set up by the industry than a com-mercial product. If these products become essentialneeds in our lives, they become ‘commodity items.’These items are mass-produced and would be availablein many retail stores. Commodity items are subject toboth to industry standards and to governmental andenvironmental regulations. The implicit requirementswill be stricter than those of similar consumer items.An important class of consumer items is the digital ap-pliance. This is a simple, generally, single function-based, easy to use, inexpensive but reliable digital de-vice. Examples of digital appliances are pocket cal-culators, digital thermometers, or digital watches andorganizers. A consumer item reaches the status of acommodity when its use is widespread, and its needbecomes essential, like clean air, water, food grains,gas and electricity—items that we cannot live without.In essence, it fulfills a major service function. A long

list of products have received the commodity status inthe high technology field, such as digital watches, cell-phones, TV, VCR’s, PC’s etc. (Do you remember thetime when you misplaced your TV remote controller,and how much inconvenience you had until you foundit.)

When high tech products reach the consumer orcommodity item status, they will be sold in large quan-tities in very competitive markets, their manufacturersfeel intense pressure on profit margins. The customerswill often focus on price, while taking the things like re-liability, safety, quality, etc., for granted. Thus, as theevolution proceeds from a system prototype stage to asalable product, then to a consumer item, and then to acommodity item, the consumer expectations rise, creat-ing stricter implicit requirements in their wake. The de-signers must remember and factor these during the de-velopment of architecture of the product family. Theseimplicit requirements are distinct from the explicit re-quirements which are generally the functional require-ments initially used by the designers during the firstproduct creation. A good example of implicit require-ment is the electromagnetic radiation standard imposedon TV sets, computers and cell-phones. Product de-signers have to foresee these and create product familyinfrastructures to easily accommodate these evolution-ary (implicit) requirements during the life cycle of theproduct family. It is somewhat similar to requirementimposed on high-rise building designers in Missouri andOregon to consider the possibilities of seismic catastro-phes even though there never was a major earthquakein those states. (The lack of a spare tire is no con-cern if you do not get a flat, so said Alan Greenspan.)Figure 7(a) summarizes these findings.

We shall extend the evolution of implicit require-ments by considering the some new trends in softwareproduct family design by referring to Fig. 7(b). Herethe rows refer to the commercial product types and thecolumns refer to the various standards and regulationsin the order of their strictness. We summarize whatis depicted in the figure as follows: The initial com-pany (proprietary) product is based on the core func-tional requirements plus some company preferred qual-ity and performance characteristics. When the productbecomes a consumer item, it is subjected to industrystandards in addition. When it becomes an open sys-tems product, it is regulated by the industry for inter-operability. When it becomes an open software sourcesystem, like LINUX, its design and code is open to allusers and developers, who have a say in any future mod-ification, change and enhancement. It is thus regulatedby the industry and its developers and users. When itreaches the status of a commodity item, it is governedboth by industry and government standards and regula-tions. Ultimately, the products may be subject to com-munity’s environmental and ethical concerns. Theremay come a time when we may require every service


(a)

(b)

Fig. 7 (a) Product evolution and implicit requirements. (b)Extended view of product evolution and implicit requirements.

system and software developer to take an oath like thephysician’s Hippocratic Oath which says, ‘I shall do noharm,’ or equivalently, ‘I shall not introduce any ma-licious viruses and activities.’ This possibility is indi-cated in the last column.

8. Surety Engineering

The implicit requirements, as we have seen in the previ-ous section, grow more stringent as the product evolvesfrom a simple prototype to a consumer or commodityitem. Of course, the nature of these requirements willdepend on the applications. Sandia National Labora-tories developed a conceptual model based on ‘Suretyengineering’ [4], which helps us to classify applicationsbased on the nature of the requirements. We have ex-tended and in some cases modified their ideas, to fitour aims. ‘Surety is the methodology by which anysystem can be designed to operate exactly as planned,every time in every circumstance. Surety is a way ofapproaching just about any systems problem that con-siders why and how a system fails, then anticipates andprevents such failures’ [4]. Products based on theseconcepts are guaranteed or assured to be safe, reliable,secure over specified lifetimes, under pre-defined con-

ditions of use. Surety engineering concepts are par-ticularly attractive for specifying the implicit require-ments of high tech services and their products, as theyevolve from purely lab prototypes to marketable prod-ucts, then to consumer items (or appliances) and thenultimately reaching the star status of commodities ornecessities of life.

Our extended version of surety engineering distin-guishes five levels of product safeguards. Each levelincludes the functions of its lower level predecessor, thehighest level being the most stringent. This followsthe original concept of Sandia, ‘the greater the con-sequences, the higher the surety level.’ We shall useSandia’s examples in the following discussion.

The lowest level, Level 0, personifies products thatare cheap (often called throwaway or non-maintainableitems) for which the manufacturer provides little or noguarantee or safeguard except it works as intended, atleast at the time it is sold. Cheap information appli-ances correspond to this category.

The next level or Level 1 safeguards are primarilyreactive. When the system recognizes or detects a sys-tem failure, it reacts to it, and it may take some correc-tive or evasive action to mitigate the effects of the fail-ure. It is suitable for applications, where a failure wouldnot create catastrophic or collateral damage. The sys-tem is based on very reliable design and uses very reli-able components and where failures are highly unlikelyand infrequent. Most importantly, it is assumed thatthe system failure is recognizable after it happens, re-pairable and the system operation is restorable, withoutcreating undue inconvenience. Level 1-surety measuresmay include consumer protection under product liabil-ity warranties, insurance against property damage, etc.

Level 2 surety is proactive. Its purpose is to avoidfailures as much as possible by taking precautions andinstituting surveillance measures like metal detectors atairports to prevent sky-jacking and rigorous training ofpilots and crew. It would include the measures outlinedfor Level 1 surety because it provides all lower levelsafeguards.

Level 3 surety is preventive. Its purpose is to pre-dict and lower the odds of a potential system failure bycareful redesign of the system. Because the automobileaccidents kill, manufacturers incorporate seat belts andairbags, not just as add-ons but as integral parts of to-tal vehicle design. It would also include the safeguardsrequired of Level 2 surety.

Level 4 surety is applied to systems where failureis not acceptable or unthinkable, that is, surety safe-guard is fundamental. To ensure absolute surety, thedesign is made highly reliable and fault tolerant so thatsafe predictable behavior is assured under anticipatedfailure conditions. Examples of applications fitting thiscategory include nuclear power plants and weapons sys-tems.

The surety levels which we associate with implicit


Fig. 8 Surety levels.

requirements on services and products follow a tran-sitive relationship, in which the implicit requirementsin Level 0 are included in Level 1, which are also in-cluded in Level 2 and so on, through Level 4. Thisis portrayed in Fig. 8. The modeling of the utilizationof high tech services and products illustrates the factthat as these systems or services move up in use andpopularity, their providers and manufacturers have tosatisfy increasingly stringent but unwritten surety con-ditions. When Intel Corporation introduced the Pen-tium microprocessor, it initially ignored a design errorin its floating-point divide command. The companyfelt the resulting numerical error was very negligibleand the frequency of use of the operation was small.But after an academic user discovered the error, a largenumber of customers demanded an immediate remedy,even though it did not affect their applications. Intelrealized that its product had become a consumer itemwhere trivial errors are not tolerated. It redesignedand reissued a corrected version of the Pentium at acost upwards of $200 million.

9. Errors in Service Engineering Activities

Service activities require intensive interaction amongstpeople and machines. It is very vulnerable to inten-tional and unintentional errors. In the following sec-tions, we shall explore ways of mitigating such errors.As stated in the beginning studies of Andersen Con-sulting and others have shown that about 40–80% ofthe failures may be due to improper human reactions.

To help reduce accident and error incidence re-sulting from man-machine interactions, designers haveconsidered two general methods applicable to a broadrange service applications. One is the process of hu-manization, personalization and application specializa-tion. And the other is the use of teams—groups of pro-fessionals (usually a small number) working together—to accomplish the tasks. We shall discuss these in thesubsequent sections.

10. Humanization, Personalization and Appli-cation Customization

We work long hours in intense interaction and close

working relationships with their computers. We pre-fer to treat these intelligent machines as human-beingsand expect to receive human-like responses from themWe can ‘humanize’ the machines to imitate human-likereactions under selected circumstances, so that theirhuman users are not confused or stressed. This built-in‘naturalness’ makes it easy for us to understand, learnand use them better. We shall call this ‘humaniza-tion.’ Examples of this include providing voice or visualadvice, guidance, and warning, before or when an op-erator inadvertently commits an error. The messagescould carry emotional content! These are common inPC publishing software.

Individuals differ from one another by gender, age,education, and physical and mental ability. One can‘personalize’ the target service system to conform toindividual user’s disabilities, preferences, and fancies.We call this ‘personalization.’ Examples of personaliza-tion are TV remote controllers for left handed people,telephones with volume and tone control for the hardof hearing, and use of large letter fonts in computerdisplays for those with impaired vision. Thus ‘human-ization’ of the machine makes it easy to use since itsresponses will be less confusing to the humans, whereas‘personalization’ individualizes the system to the par-ticular human user to compensate for physical disabil-ities or to support his/her preferences

‘Customization’ adapts a general-purpose com-puter or a machine which is generally, a cheap, mass-produced consumer or commodity item to performmore effectively and efficiently in a specific class of ap-plications. This is generally done by adding specializedinstructions, or ROM-based firmware or software ad-juncts. Intel’s Pentium MMX microprocessor comeswith 57 special instructions to support multi-media ap-plications. Application customization is very commonin the PC domain, particularly in visual graphic sup-port, in database accelerators, and in sound cards forenhancing audio performance.

In summary, we ‘humanize’ the machines so thattheir responses are natural and human-like. We ‘per-sonalize’ the machine responses to fit the user’s individ-ual preferences or disabilities. We ‘customize’ or pre-condition general-purpose machines to perform special-ized application-based tasks efficiently and effectively.

10.1 A Layered Hierarchical Model for ApplicationCustomization, Humanization and Personaliza-tion

To develop system architecture based on the ideas out-lined in the previous section, we shall propose a lay-ered model much like the one used in communicationprotocols generally known as the ISO Model. In sys-tems engineering, complex designs are often modeledand constructed by hierarchical layering. Each layer isassociated with a set of service processes. The layer


Fig. 9 Layered hierarchical model.

at the lowest level interfaces and interacts with thefast electronic and physical elements. The top mostlayer generally interfaces with applications and users.Through a series of layers, the top-level (the highestlevel) tasks are translated into machine recognizableelectrical signals at the bottom-most layer [5].

Our model consists of four layers (Fig. 9). Thelowest layer will be the general-purpose microprocessoror micro-controller. The next higher layer will be theapplication customization layer, which preconditionsthe machine to perform specific application orientedtasks efficiently. This is similar to the enhancementsand the add-on packages (or boards) achieve speed-upin database accelerators or graphical visualization sys-tems. The next higher layer will be the humanizationlayer, which helps to the system to react in a human-istic fashion. This could be also be dependent on theapplication tasks but it will focus on general humanperformance and reactions The top or the fourth layeris the personalization layer which tries to help out inthe personal disabilities or caters to user’s preferences.Individual personalization is left as an option to be ex-ercised by the user and therefore, some manual controlsmust be provided where warranted.

10.2 Comments on Humanization

The purpose of humanizing a computer or a machineis to make its interactions as natural and human-likeas possible to its users. By humanizing, we try to re-duce the circumstances where the human user will besurprised, uneasy, inconvenienced, confused or stressed.Computer and other intelligent machines have becomeour constant companions. We crave for human-like re-lationships with them. We expect human-like interac-tions with machines and we apply subconsciously socialrules when we interact with them, such as politenessand respect. We prefer a female voice for soothing af-ter making a mistake and a male voice to admonishus before committing one. Unfortunately, our currentsystems are not designed to be humanistic and behavein a predictable way. This is a likely source for a largefraction of machine operator’s mistakes. The majorityof our responses during man machine interactions fallinto one of the following categories:

a) Responses based on inherited and built-in naturalinstincts and reflexes-often called the knee-jerk re-sponses. These are fast responses where no think-

ing is required.b) Responses based on acquired knowledge, experi-

ence, education and training. As knowledge ac-cumulates and evolves over time, our habits andresponses tend to change. These are thoughtful ordeliberate responses, which require introspection.

c) Responses to new and unanticipated situationswhich have to be dealt without prior training,preparation, experience and knowledge. While re-sponses to these may create risk but they alsoprovide exposures to new learning experience.These require research and more exploration orknowledge-gathering.

For the first two categories, we can respond effectivelyusing our built-in instincts, experience and knowledge.For the last category, we are left in the dark. Peoplemay look over their neighbor’s responses and do (copy)the same, much like the way we follow the main trafficflow when we are lost in a fast freeway. Or succumb toTurtle’s casino effect—the addictive habit of continuingto gamble even if one has lost heavily. Or, moving inthe direction of increasing illumination when we are lostin the dark, or doing what may be considered as polit-ically correct, even though it is against our conscienceor moral judgement.

We next summarize or quote some eminent so-cial scientists on the man-machine interactions withoutcomment. These include Profs. Sherry Turkle of MIT,Byron Reeves and Clifford Nass of Stanford, W.C. Fred-erick of University of Pittsburgh and journalists likeThomas Pitzinger Jr. of the Wall Street Journal [6].

Information technology has changed the way weinteract with the world at large and with the machinesin particular. People have become the connoisseurs ofinformation. One can go on-line, learn more about theproduct one plans to buy, make a rapid price compari-son between different stores and then make a purchase.There is an innate yearning for the human touch thatwe believe will be provided by the computer. There isa lack of predictability when we deal with machines.“The future has so much uncertainty about it that weneed to prepare our children to see the future in a dif-ferent light.” Prof. Turkle tells about the Furby doll, afuzzy computerized mechanized doll that talks, blinks,sleeps and asks to be fed. The actions of this doll aretotally unpredictable. Compare this with a Barbie dollthat behaves predictably. Playing with the Furby dollmakes the child get used to unpredictability and un-certainty. The child will not be uncomfortable or illprepared in the future, when dealing with unexpectedcircumstances or interactions with machines whose be-haviors are not predictable.

All living things harbor an impulse to economize,to accomplish more with less. Economizing process isonly way to survive, grow, develop and flourish. ‘Thegenes that create us have programmed us for business-


the foundations of which are trade, technology and di-vision of labor.’

11. Teams and Team Evolution

Since service functions are enabled and executed byteams, we shall briefly comment on some pertinentcharacteristics, specifically their types and functions.The word ‘team’ is defined in the Webster’s dictionaryas a number of persons associated in some joint ac-tion like a team of experts, or ‘draft animals’ harnessedtogether to draw a vehicle. P. Senge [7] explains themeaning of a team as, ‘an alignment of people when itfunctions as a whole. When a team gets more aligned,a commonality of direction emerges, the individual en-ergies harmonize.’ Teams in our discussion are madeup of people supported by machines to achieve certaingoals during some specified period of time in a coop-erative or collaborative fashion. Teaming includes theprocesses of planning, organizing, controlling and exe-cuting the required tasks by teams.

Teaming is a primitive (herd) instinct inherited byman from his animal ancestors, practiced by predators,hunters and scavengers amongst animals, and used bystreet gangs and the military amongst the humans. Inancient days small teams (typically from 2 to 7) workedtogether as craftsmen at the home of the merchant-craftsman. This is the forerunner of home-office andtelecommuting concepts. Their organization was tightknit and hierarchical. The 1770’s marked the start ofindustrial revolution in Europe funneled by the steamengine technology. It used the division of labor, and thework place shifted from the home to the factory. Herethe teams employed anywhere from 200 to 2000 people.With the growth of assembly lines in U.S. (thanks toHenry Ford), the teams became more specialized andmanually intensive—using more brawn than brains. Inthe future, with the advances in information and com-munication technologies and the ever-spiraling incon-venience of rush hour traffic grid-locks, telecommut-ing and home-offices will become popular, where teamswill virtually work together from the more pleasant sur-roundings of their homes at different physical locations.Lo and behold, our ancient craftsman has returned!

As the technologies begin to interact, the systemsand their designs become more complex. Teaming be-comes tools for design, implementation, prototypingand design verification. Team members learn from eachother, complement and supplement each other’s exper-tise and share the developmental burden they have beenvery effective in the detection and correction of errorsbefore the designs are implemented. These activitiesare done by groups of people since no one single personcan do them alone. Teams, when properly organizedoften produce innovative and bold ideas.

Gersting and Ives of Anderson Consulting (So-lution Integration Magazine, July 1999), state that

87% of failures and costly mistakes are caused by lackof knowledge on the part of designers, operators andusers. Several of the catastrophic disasters are tracedto operator errors during unexpected and unanticipatedevents. Teams of knowledgeable operators, users andexperts can, with proper computer aided support canhelp reduce this. Oftentimes, the critical decisionsmade by properly qualified teams are superior thanthose made by a single human. Poole [8] provides argu-ments and instances taken from safety critical applica-tions to support the inherent error detection and errorprevention effect of teams. Quality Circles used in soft-ware development in Japan utilize well-chosen teams ofprofessionals for early detection, location of errors inlarge software systems. The IBM’s Clean Room soft-ware development procedures advocate the multiple ap-plication of inspection teams for the same purpose.

A well-coordinated group of people was found tobe very effective when working with large complexcomputer-communications systems in FAA, US Navy’sNuclear Submarines and aircraft carriers.

A cardinal advantage of any team-based activity isthe learning experience gained by its members. Cross-disciplinary team members provide both supplementaryand complementary support in accomplishing a com-plex intellectual activity. We shall cite a recent exam-ple.

The chess match that culminated on May 11,1997 pitted Chess Champion Garry Kasparov againsta human-machine team that included the IBM super-computer, Deep Blue. The Champion lost The newsmedia came to the conclusion that the machine has de-feated a human champion, ignoring the fact that thecomputer and its programs were all developed by asuperb group of cross-disciplinary experts and math-ematicians. Indeed one can say that a team of humanswith support from machines can solve more complexproblems than it is possible by single human alone. Amachine by itself cannot solve a given structured prob-lem unless it is provided with intelligent software toolsencapsulating the chess-players’ knowledge and tactics(in the form of algorithms). In the case of the ChessChampionship, the Deep Blue’s programs (developedby several teams) looked 2000 steps ahead on all pos-sible moves of the opponent, before it made its move.The machine also used the playing style and tacticsof the champion, Kasparov, from his several previousgames as inputs. One can conclude that a team of hu-man specialists augmented by powerful computers andprograms can out-perform a single human expert.

One of the assumptions we make is that a com-puter cannot be smarter than its designers—who arehuman beings. A person with computer resources canbe ‘superior’ to a person without such support. A teamof humans properly supported can perform better thana single person with a machine. Ultimately, groups ofteams could be more efficient than a single team. These


are just empirical observations and it is good remem-ber the context and the implicit assumptions that wemade. In essence, we can hypothesize empirically fromthese observations the following:

Machine < Human < (Human+Machine) < Teamsof (Human+Machine)where the symbol “<” implies “inferior to” or ‘notas good as’ relation, and the symbol “+” refers to“logical or,” or, “either one or both” relation.

In a simplistic way, the Turing Test (1950) sought toestablish whether a device possesses human-like intel-ligence or not. If the test results on both the machineand human show that one cannot the distinguish tworesults-the results from the machine matches to thoseof the human-then one can conclude the machine pos-sesses human-like intelligence. Turing provided little orno details on how such a test can be conducted and theresults are to be evaluated. H. Dreyfus, from a phe-nomenological basis and John Searle from a linguisticand cultural basis (Chinese Room Argument) questionthe fundamental premise of the test.

11.1 Team Types

Team type depends on the main purpose for which theteam is formed. It defines the organization and the na-ture of dependency amongst the team members. Thecriteria one can use in classification are the team’s pur-pose the nature of its collaboration, and its organiza-tional and control philosophies. These govern the na-ture of interaction and dependency amongst its mem-bers We shall only discuss some specific but importanttypes of teams used in the service functions namely,the collaborative, co-operative, consensus seeking, hi-erarchical and virtual teams.

Collaborative team members work together insmall groups, very closely, i.e., cohesively, to achievethe best outcome for their intended goal. Generally thegoal of the task is well defined, and the team membersare chosen because of their motivation with the objec-tive, perceived potential and their competence in thatarea. Collaboration creates a synergy such that theirtotal contribution is greater than the sum of their in-dividual contributions. The team members are tightlycoupled who work together very intensively and striveto achieve the very best outcome. A parallel conceptfrom software engineering of this is cohesiveness, whichimplies that the components (members) of the team arevery dependent on each other, and work together veryclosely and effectively.

Cooperative team members support, and interactwith each other in a loose or a loosely coupled fash-ion. Their participation may be sporadic or may worktogether on an ‘as needed’ basis because of their spe-cialized expertise. They may be located at differentplaces. Two collaborating teams may be cooperating

and therefore will be interacting with each other ina loosely coupled fashion. Example of a cooperatingteam would be a group of consultants, who work witheach other in some specific phase or problem of a largeproject because of its expertise in that activity.

Consensus developing teams are organized to cre-ate an acceptable or a negotiated solution for the crit-ical issues. When a team is tackling a large number oftrade-off issues, the team members may not all agreeon any one of proposed solutions. To break the dead-lock, it is usual, for the team to agree only to the leastcontentious but most agreeable compromise solutions.This may not be the best, but it would be acceptableall around. For the sake of time and expense, all or amajority of the team members may go for this ‘leastcommon denominator’ or ‘politically correct’ solution.Practices such as the quality circle method of evaluat-ing the quality of software follow this principle becauseof numerous trade-off issues. Consensus teams containmembers from a diversity of interests, and always striveto find the best workable common ground. Anothersimple example is a team of jurors in a legal trial.

Virtual teams may be considered as simulated rep-resentation of real teams. Political pollsters use thismethod to gauge the public support for several officeseekers. They provide the freedom to experiment andstudy the results about activities of people and ma-chines by computer simulation. Irreversible and actionswhich either take a long time or too dangerous for re-alistic evaluation can be studied in this virtual worldscenario. Virtual teams may consist of software agents,representing specialists (expert systems), or monitorsand observers who record critical events during simula-tions. Currently the phrase ‘virtual team’ also is usedto denote a geographically dispersed group of peopleforming a team.

Hierarchically organized team has a command andcontrol type of relationship amongst its members. Amember in a lower hierarchy reports to or obeys theorders of those in the next higher level. A common formof the model is an inverted tree, the root node is thecommander and the leaf nodes are soldiers. The nodesin the intermediate level represent officers. Hierarchicalteams are tightly coupled. Examples include militarycommand and control systems, gangster teams, drugcartels and the common master-slave systems.

We shall next comment on other types of teams,whose dependency objectives may be useful for servicetype operations.

Reciprocal dependency teams understand eachother’s goals and try to eliminate conflict, competi-tion or confrontation by working around them. Theconflict may be about their need to share some com-mon resources and therefore requiring them to rear-range their activity schedules to avoid conflicts. In an-other type of dependency, called mutual dependency,each team needs the other’s help to achieve its desired


goal. Benevolent dependency-based teams try to helpthe overall society and encourage others to do likewise.

Altruistic teams sacrifice their goals for the com-mon good of the society without any direct benefit forthemselves. This type of behavior is observed amongstanimals that sacrifice themselves to save their herd.This typifies the moral: ‘All for one and one for all.’

11.2 Cohesion and Coupling in Teams

Similar to the terminology used in software engineering,cohesion is the property that attracts, or binds two en-tities together to produce a new one with different char-acteristics. It is a dependency relationship and in ourcase, produces tight collaborative interaction betweenteam members. This tight relationship overshadowstheir individual behaviors. It is a synergy producingrelationship or fusion. Cohesion exists between humansand their computers. A team of people working in anassembly line all participating in a single task exhibitcohesion. On the other hand, coupling is a relationship,which preserves, in some way the individual propertiesof the constituents, like a team of consultants all work-ing on different aspects in a loosely coupled fashion. Itis a dependency relationship between larger aggregates.An analogous, often-cited concept is from chemistry.The atoms form molecules and molecules combine tobecome compounds. We shall give an oft-cited exam-ple. Sodium (Na), an explosive element combines withChlorine (Cl), a highly toxic element to form commonsalt (NaCl), which is neither explosive nor toxic. NaClis a molecule or a simple compound. This illustratesthe important property that molecules formed from co-hesive combination of atoms may not exhibit the prop-erties of their constituent elements. Compounds pro-duced from the interaction of molecules may retain thecharacteristics of their constituent molecules. Cooper-ating teams show loose coupling whereas collaboratingteams show tight coupling, or cohesion.

11.3 Evolution and Control of Man-Machine Teams

In this subsection, we shall model the evolution ofteams particularly those where the team members areheavily computer supported as in the service industries.The four models use simple directed graph structures.

1) In the first model, each node represents a team,and is independent and isolated. There are no linksor arc emanating from the nodes. Nodes have lit-tle or no communication capability, that is eachnode is isolated from others. Each node makes itsown decisions. (They can be ripe merger or take-over candidates). We shall label this as a totallyisolated team model. The nodes are like islandsof isolation developing themselves over time-verymuch like the ancient clans who have their own

Fig. 10 Team types.

languages, religions, and cultures which are localand specific, uninfluenced by their neighbors. Thisis shown in Fig. 10(a). Here the cost of communi-cation across teams is zero, and there will be nofear of virus invasion or security breaches.

2) The second model centralized model which por-trays a top down hierarchy, in which one node, theroot node of the tree acts as a controller or direc-tor receiving information from its subordinates andgiving directives to its subordinates. This modelsa centralized enterprise, where the head office re-ceives status information from the branch officesand sends out directives or commands to the subor-dinate nodes to carry out. Military organizations,which use the command and control regimen, andstreet gangs are examples of the centralized, tightlycoupled hierarchical organization. This model isillustrated in Fig. 10(b). This model is suitablewhere the communication costs are high and there-fore communication is kept to a minimum. Thisphilosophy is also used in the obedience trainingschools for dogs. This system works well when thenumber of branch offices (dogs) is few, and there isno need for real time updating of the informationat several sites. The down side of this organizationis there is no interaction amongst the nodes, thehead office can micromanage the operation of itsbranches.

3) The next model (partially decentralized model) isillustrated in Fig. 10(c). The sub-trees under thenodes B and C are organized like the centralizedmodel. The nodes (teams) B and C report to thecentralized node A. We can consider nodes B andC as regional head offices. The enterprise is stillcontrolled by a single headquarters node A. By par-tially decentralizing the system, the computationalburden on the central node is lessened. This modelprovides the next step towards expansion or evo-lution of the centralized model described above.The communication requirements can be high de-pending on the size (number of nodes), and thestructure of the system.

4) The fourth model is the fully decentralized modeland is shown in Fig. 10(d). This model assumesfull connectivity between the teams or nodes andthe inter-node communication is assumed to be


Fig. 11 The effect of decreasing cost of communications andincreasing decentralization of control.

instantaneous with negligible (e.g., the Internet)cost. Ideally, the teams can have total access toeach other’s information databases at all times.Each node makes functional and administrativedecisions rapidly and collaboratively with othersbased upon a set of agreed-upon rules, protocolsand principles. The designers of the system mustcarefully check and validate these rules and roles toavoid conflicts, inconsistencies, and ambiguities tothwart any security breaches and catastrophes thatcould occur during faults, malfunctions or virus in-vasions on the system.

We envision that a realistic system may portray anyone of the above organization models or combinationsthereof. A flexible or a reconfigurable system can switchbetween these configurations to adapt to the changes inthe application, communication environment and otherconditions.

We infer through our classification and discussions,that as the communication and computational costs godown, the system evolves from an unlinked, isolatedand independent enterprise units (Fig. 11(a)) to a cen-tralized or hierarchical configuration (Fig. 11(b)), thento a partially decentralized configuration (Fig. 11(c))and finally to a fully decentralized configuration(Fig. 11(d)). This is very similar to the cultural evolu-tion through globalization that we see around us. Oncenot very long ago, the world was littered with king-doms, nationalities and irreconcilable differences due tonon-interaction and lack of communication. But now itis giving place to a democratic and cooperative world—towards a fully decentralized system—which is madepossible by the advances in education, information andcommunication technologies.

12. A Finite State Machine Model of the Sys-tem and an Actor Model of the Component

In this section, we model the organization and control

Fig. 12 Finite state machine model of air craft carrier controland organization.

of man-machine teams. We show how this organiza-tion and its control structure can be changed by dif-ferent circumstances. Every state in this model is ameta-state-an ensemble of states. A finite state ma-chine (FSM) model provides a conceptual frameworkfor the rest of the discussion. The behavior of the teamis represented by the finite state machine. The entity orteam member is modeled as becomes an actor, who as-sumes a given role during that state and acts accordingto a given set of rules of behavior. The team member isa service provider. When the system is in that state theentity follows the rules of behavior assigned to its partor role. The play is personified by the state. When thefinite state machine of the system moves to a differentstate, the play changes. The actors, the components,entities or team members assume different roles in a dif-ferent play and act their part accordingly. To reiterate,when the system state changes (a new play is initiated)the team member assumes a new role and acts undera different set of rules. When the system encounters adifferent situation, the system state changes (the sys-tem reconfigures), the team members assume new rolesand the system changes its organizational and controlstructure.

We shall provide an example first and then createa finite state model from it. The example is taken fromPoole [9]. A navy aircraft carrier has many sailors andofficers in it. We will use the sailor as the model ofan actor. We represent the states of the sailor witha finite state diagram as shown in Fig. 12. State A isone where the sailor does normal routine operations as-signed to him. These are based on standardized proce-dures. In this state the sailor follows the command andcontrol protocols and procedures. The command andcontrol procedures and protocols prevail. The organiza-tion is hierarchical, i.e.; the command orders propagatethrough a chain of command in a top down fashion.The responsibilities and the duties are as prescribed inthe command and control rulebook. Ranks and author-ity are recognized and strictly enforced. The primaryconcern here is performance efficiency. The State B


Fig. 13 A model of financial institution.

in Fig. 12 exemplifies the aircraft launch and recoveryoperations. These operations are most stressful to thepersonnel and sometimes quite dangerous. Every manand machine have well-defined functions and the sailorsand officers are trained for this. The primary concernhere is the safety of the individuals, the aircraft andthe ship. Ranks and commands can be ignored for thesake of safety. If the sailor sees that a plane’s landinggear is not in proper position during its landing, he canforget the protocol and wave off the plane not to landor abort the landing.

The third state, State E, represents an emergencyor an unanticipated situation on the aircraft carrier.This could be due to a fire on the deck; a missile hit onthe ship etc. Even though the personnel of the carriermay have been drilled for several types of emergency,most of the accidents cannot be anticipated. In suchsituations, each sailor acts for himself as well as for thetotal benefit for all in the carrier. The emphasis, insuch instances, is one should not panic but one shouldexhibit an altruistic, predictable and orderly behavior.Collective safety and altruistic spirit takes precedenceover one’s own personal safety, rank and protocol. Forsimplicity, state E may also include subsequent repairand reconfiguration of the damaged system.

We have shown above is an idealized version ofthe states the aircraft carrier and the role of one itscomponents, a sailor, as they go through their life cycle.A finite state system models the states of the carrier.The actor model fits the component, the sailor. We willnot discuss the state transition conditions since they areapplication dependent.

12.1 An Example of Evolution of a Financial Institu-tion

This is a legacy application, in which initially the fi-nancial institutions X, Y, and Z were in geographicallydifferent locations with little or no communication be-tween them (Fig. 13(a)). The institutions merged theirenterprises, with Office Z becoming the head-office.The central data files for the whole enterprise, X, Yand Z are located in Office Z. There is data (transac-tion) update activity between X, Y and Z. The head-quarters node Z controls the system. It has access tothe data files at the branches. The headquarters node Z

has the leadership role. This is an example of a central-ized (hierarchical) system. When more branch officesare created the enterprise establishes several central-ized systems each controlled by a regional head-office asshown in Fig. 13(b). The regional head offices are againas a tree with a single root node serving as the enter-prise head-office. This organization the old hierarchicalstructure is maintained and therefore old proceduresused in the previous system can be re-used. This is theexample of a partially decentralized system. When thecommunications costs go way down and become negligi-ble, all the files are accessible from any node or office ofthe enterprise (Fig. 13(c)). Any high-level decision canbe made and administered at any node or office at anytime. In essence, each node can make decisions with theinstantaneous participation of other nodes. Each nodeacts in conformance with the rules and policies govern-ing the whole network. This represents what we calledearlier, the democratic or decentralized way of control.It is also an example of decentralized autonomous con-trol that empowers the teams in an industry to performtheir very best efficiently in rendering services to theircustomers.

12.2 Comments on the Models

These models are implementable and simulatable foran arbitrary system. The actors or components areman-machine teams. For example, the team membersare tightly coupled and hierarchical in one state, butchange as the circumstances warrant. They may be-come partially or fully decentralized as the situationunfolds. Within a team the members are collabora-tive, whereas the teams within a system state may beloosely coupled. The behavioral rules of the actorswithin teams and the rules that govern inter-team in-teraction must be consistent, and complete but alsonon-conflicting and unambiguous. We shall concludethis section with an analog taken from digital design.Consider the design of a time-multiplexed subsystem offour flip-flops. We wish to reduce the hardware by time-sharing them in many computer functions. We can de-sign an incrementing counter, a right shifting registerand a complementation register using the same fourflip-flops. The designer designs a system that when ashift command signal is given, the subsystem performsone position open-ended right shift and later when anincrement command comes in, it increments the con-tents of the four flip flop register by one, and so on. Inother words, during a shift command the flip-flops be-have as a shift register. Under the increment commandthe subsystem behaves like a counter. The flip-flopschange their role according to the command they haveto execute. This is very much like the same actor per-forming the part of Hamlet in one play, and the roleof Othello in another but still working for the sameplayhouse.


13. Conclusions

Service functions have matured, and have reached themost dominant position by entwining themselves withthe advances of information and business technologies.Software engineering has been the technology catalyst,a behind-the-scenes player but the major enabling dis-cipline in all service sectors. Recent data from the Bu-reau of Labor Statistics indicate that more than two-third of all jobs in US involves service functions. Inthis paper, we have traced the evolution of the ser-vice industry. We have overviewed the distinguishingfeatures of its functions. These vary with the appli-cation and the industry. Generally they involve inter-actions amongst teams of professionals and machines.Faults and errors inadvertently or intentionally intro-duced during the performance of service functions cancreate failures and breakdowns in critical systems.

Transfer of knowledge and technology into thebuilding of commercial products and services (alsoknown as technology transfer) is an important servicefunction. We decompose it into phases, which includethe development of appropriate technology from scien-tific knowledge, the creation of tools (mostly software),and then, designing, simulating, prototyping, assem-bling and implementing the product.

As we trace the evolution of innovative technolo-gies over the last two centuries, we note that automatedmachinery and processes are replacing the manuallyintensive activities. These automated tools and ma-chines have benefited very greatly by advances in infor-mation technology and artificial intelligence. Automa-tion always improved the productivity and reliability ofproducts, and services. It also eliminates a large por-tion of manual and mental interaction errors as well,which currently account for 80% of all failures in crit-ical systems. Such errors are the result of improperuser/operator’s interactions with the machines. We ar-gue that by proper humanization and personalization ofmachines and computers, the failure incidence can begreatly reduced. The use of human teams with com-puter support had been found to be very effective aswell in reducing errors in design and operation. Interac-tion amongst team members help their rapid learning,and towards sharing and creation of knowledge.

The type, organization, and control of man/mach-ine teams amongst other things depend on the applica-tion. The Finite State Machine and Actor models fitwell in portraying an important class of teams. Dur-ing a particular state of the system, the teams (compo-nents) or actors follow a specified set of rules of behaviorand operation. During the next state, the actors followa different set of rules of operation and behavior as ifthey are doing in a new role in a different play. Thestate transitions depend on the application events andcircumstances. These models are helpful in the simula-

tion and analyses complex man-machine systems.We consider four distinct types of team organiza-

tions, namely, isolated, centralized or hierarchialized,partially decentralized and fully decentralized. Whenthe communication and computer processing costs arenegligible, fully decentralized disciplines appear to bemost congenial to the rapidly evolving technologies andeconomies.

There exist lots of research issues worth lookinginto. We shall touch upon a few. The performanceevaluation of service functions and activities dependson customer satisfaction. We need to establish the cri-teria, the measurement metrics to detect trends of whatcustomers desire to develop suitable data-mining ap-proaches. Currently we are studying representationsof service activity work elements, like the one we haveused in this paper. We have combined the ideas remi-niscent work of Gilbert and Taylor of the time and mo-tion study and operations research fame and the ideasof computer programming to represent a service activ-ity as a program entity and perform an entropy analysissimilar to the ways Shannon did for information theory.These will be subjects of future papers.

Acknowledgements

The author owes a deep debt of gratitude to Prof.G. Kozmetsky for his invaluable advice and encourage-ment. Also Prof. R.T.Yeh provided valuable insightswhile Profs. Kane Kim and Kinji Mori helped in clari-fying several ideas discussed in this paper. Dr. R. Paulof the Dept of Defense provided valuable commentson surety engineering ideas. Last but not the leastDr. T. Kwon, his colleague at the University of Cali-fornia, Berkeley provided invaluable advice and com-ments in the overall organization. The writer expresseshis deepest appreciation to these friends.

References

[1] D.S. Davis, “J. Boskin, ‘The coming of knowledge based busi-ness’,” Harvard Business Review, Sept.–Oct. 1994.

[2] K. Weick, “Ross Ashby’s principle of requisite variety,” inThe Social Psychology of Organizing, Addison Wesley, Read-ing, MA, 1979. (also, R. Cooper, “Debunking the myths ofnew product development,” Research Technology Manage-ment, July/Aug. 1994, 1999.)

[3] A. Grove, “Only the paranoid survive,” Doubleday, April1999.

[4] T. Dellin, “Confidence by design,” Sandia Technology, San-dia National Labs, vol.1, no.1. Winter 1999. Also M.C. Paulk,B. Curtis and M.B. Chrisis, “Capability maturity modelfor software Version 1.1,” Software Engineering InstituteCMU/SEI-93-TR, Feb. 24, 1993.

[5] G. Held, Data Communication Networking Devices, JohnWiley & Sons, 1986.

[6] T. Petzinger, Jr., “A model for the nature of business. It isalive,” Wall Street Journal, Feb. 2nd, 1999.

[7] P.M. Senge, The Fifth Discipline: The Art and Practice ofLearning Organization, Bantam Doubleday Dell Publishing


Group, 1990.[8] R. Poole, “When failure is not an option,” IEEE Engineering

Management Review, Spring 1999.[9] D.M. Amidon, Innovation Strategy for the Knowledge Econ-

omy, The Ken Awakening, Butterwoth-Heinemann, 1997.

Chitoor V. Ramamoorthy earned his two undergraduatedegrees in physics and technology from the University of Madras,India, two graduate degrees in mechanical engineering from theUniversity of California, Berkeley and a Master’s and Ph.D. fromHarvard University in Applied Mathematics (Computer Science)in 1964. His education was supported by Honeywell Inc.’s Com-puter Division, Waltham, MA with whom he was associated till1967, last as Senior Staff Scientist. He then joined the Univer-sity of Texas, Austin as a Professor in Electrical Engineering andComputer Sciences. After serving as Chairman of the ComputeScience Dept., he joined the University of California, Berkeleyas Professor of Electrical Engineering and Computer Sciences in1972, a position that he still holds. He has supervised 73 Ph.D.students, who include Vice Chancellor of University of Texas Sys-tem, Deans, Dept Chairmen, Chair Professors, the CEO of Lu-cent Technology’s largest subsidiary and including, most recently,the President Elect of the IEEE Computer Society. He has heldthe Control Data Distinguished Professorship at the University ofMinnesota and the Grace Hopper Chair at the U.S. Naval Post-graduate School, Monterey, California. He was also a VisitingProfessor at the Northwestern University and Visiting ResearchProfessor at the University of Illinois, Urbana-Champaign. He isa Senior Research Fellow at the ICC Institute of the Universityof Texas, Austin. He has received the IEEE Computer Society’sGroup Award in Education, the Taylor Booth Award for Edu-cation, the Richard Merwin Award for Outstanding ProfessionalContributions, the Golden Core Award and the IEEE CentennialMedal. He is a Fellow of IEEE and of the Society of Design andProcess Sciences, from which he received he R.T. YEH Distin-guished Achievement Award in 1997. He also received a BestPaper Award from the IEEE Computer Society in 1987. Threeinternational conferences were organized in his honor as well oneUC Berkeley Graduate Student Research Award and two Inter-national Conference/Society Awards have been established in hisname. He served as the Editor in Chief of the IEEE Transactionsof Software Engineering, and the founding Editor in Chief of theIEEE Transactions of Knowledge and Data Engineering. He isalso founding Co-Editor in Chief of the International Journal ofSystems Integration and of the Journal of the Society of De-sign and Process Sciences. He served in several capacities in theIEEE Computer Society including the First Vice President andthe Governing Board Member. He served on several AdvisoryBoards of the Federal Government and Academia including U.S.Army, Navy, Air Force, the DOE’S Los Alamos Lab, the Univer-sities of Texas, California, Toronto, and State University Systemof Florida. He is one of the founding directors of the InternationalInstitute of Systems Integration in Campinas, Brazil, supportedby the Federal Government of Brazil and for several years, amember of International Board of Advisors, of the Institute ofSystems Science of the National University of Singapore. He haspublished more than 150 papers and co-edited three books. Hehas worked on and holds patents in computer architecture, soft-ware engineering, computer testing and diagnosis and data bases.He is currently involved in the research on models and methodsto assess the evolutionary trends in information technology.

A Study of the Service Industry—Functions, … Schumpeter tracedthegrowthandtheimpactoftechnologiesover thelasttwocenturies.(Figs.3,4).Theﬁguresshowthe ... (TheEconomist,Dec.31,

Documents