»

"

"

"VOLUME VI, NO. 11 1

THE MONTHLY NEWSLETTER ON EXPERT SYSTEMS, OOP, CASE, NEURAL NETWORKS & NATURAL LANGUAGE

Editor: Paul Harmon

AUTOMATING KNOWLEDGE FLOW AT THE HELP DESKAsreaders know, we rarely run articles by

anyone other than your editors. In this case, weare making an exception because of the high levelof interest we thoughtreaders would have in this ar-ticlepreparedfor ISS by Avron Barr. Avron Barris the President ofAldo Ventures, as well as an in-dependentconsultant, writer, andpublic speaker onknowledge systems in business. He co-edited thefour-volume Handbook of Artificial Intelligenceand is writing Knowledge Distribution Systems tobe published in 1991 by Addison-Wesley. Mr.Barr's current half-day seminar, "TheEvolution ofExpert Systems in MIS," is generating interestfromMIS managers worldwide. Aldo Venturesoffers a

planning service to help deskmanagers looking attechnology alternativesfor improving their opera-tions. Thefollowing article by Avron provides anoverviewof the current stateof expertsystems andhelp desks. —Paul Harmon

"Information is not knowledge until and unlessit is applied effectively."

Dan Burrus, Futureview, 1989As anyone who has spent an hour vainly looking

through a manual for the answer to a questionwould testify, information is notknowledge.Modemknowledge systems, applying expert sys-tems in conjunction with other software and com-munications technologies, are helping manycompanies turn information into knowledge.

IN THIS ISSUE

AUTOMATING KNOWLEDGE FLOW AT THEHELP DESK — By Avron Barr

The need for automation at the helpdesk, the role ofexpert systems technology in systems solutions,and some currently availableproducts 1

WHAT'S HAPPENING IN EUROPE — PART IA generaloverviewof theexpert systems and OOP

activities in Europe 9MERLIN: A FACTORY KNOWLEDGE-BASED

SYSTEMFOR FORECASTING PRODUCTDEMAND

An interesting new applicationthat helps H-Pmanagers forecast product demand 12

TRENDS AND NEWS BRIEFSShort notesand comments about significant recent

announcements 14

While expert systems were originally introducedcommercially as large, problem-solving programs,theproper nurturing and cultivation for thistechnology's development at this time is found inautomating current efforts to distribute corporateknowledge. The natural home for this technologyis in the support ofexisting knowledge distributionactivities like training, employeepolicy handbooks,regulatory compliance guidelines, system and productdocumentation, troubleshooting guides, new productbulletins, procedures manuals, and help desks.

In recent years, expert systems tools and techniqueshaveproven effective in the automation ofcustomersupport as well as internal, e.g., data center, helpdeskactivities. This reflects a maturation of expertsystems technology and the real benefits achievedby thosepioneering companies that identified and

*I

*

"

"

"11

2

NOVEMBER 1990

pursued help-desk applications in their search foruses for the technology. In the last year, severalvendors have come forth with task-specific pro-gramming products designed for this market niche,and there are more to come. This article reviewsthe need for automation at the help desk, the role ofexpert systems technology in systems solutions, andwhat kinds ofproducts are available now.

KNOWLEDGE FLOW AT THE HELP DESKEvery help deskoperation has different

problems, for different reasons. For some, the highturnoverof staff and enormous time and cost oftraining new customersupport representatives(CSRs) results in a nearly disabling productivityproblem and often low-quality advice. For others,the size of theproduct line they support, frequencyof changes and additions to theproducts, and com-plexity of problems and interactions in thefieldmake the jobdifficult even for well-trained,ex-perienced personnel.

Some of the typical help deskproblems (e.g.technical product support that meetsdetailed cus-tomerspecifications), are best dealt with by conven-tional information systems. But many of thedifficulties that help desk operations face are in-herentlyknowledge problems. For example, a newCSR cannot use the information available to him orher inmanuals, notebooks and meetings withoutex-tensive training on the products, documentation,and databases. But even experienced and talentedCSRs have trouble integrating into their telephonetroubleshooting the new problems that customerspresent (amorphously) every day and the fixes thatengineering figures out.

Similarly,troubleshooting guidance is some-times needed tokeep up with new products,releases, and repair procedures. This is especiallyimportant when the CSR is trying to scheduleproperly equipped field service visits that will fixthe customer's problem withoutrequiring a secondtrip. And in addition to the often troublesome taskof figuring outexactly what thecustomer is talkingabout, the CSR must conform to the company's

emergencyresponse guidelinesor product returnpolicy.

KNOWLEDGE FLOW ANALYSISMuch of the value ofbuilding computer systems

solutions to business problems lies in the analysisnecessary to specify the problem and design an ade-quate solution. Forknowledge systems, this invol-ves understanding what people need to know to dothe jobright. I call thisknowledge flow analysis,resulting in an identification ofknowledgeproblems, some of which may thenbe addressedbyknowledge systems solutions.

Knowledge flow analysis is like dataflowanalysis in that it examines the flow of somethingaround an organization. But unlike data or informa-tion,knowledge is intangible. Despite the commonusage of the word, knowledge doesnotexist inbooks, manuals, databases, orknowledge bases.Knowledge is used to describe people's behavior.Knowledge flow analysis involves looking atpeople's experiences and behavior and inferringwhat goes on in their heads. The issue is not whatinformation is or should be available,but ratherwho can answer the questions, today, underthesespecific circumstances, given thecustomers' par-ticular, confused presentations of their situations.With theproviso that everyhelp desk operation isdifferent, let's look at some of theelements of aknowledge flow analysis.

The nodes inknowledge flow are people facedwith situations. Do they act knowledgeably? Op-timally? The CSRs for instance, when dealing withincoming calls, must take into account theircustomers' products, operational environments,technical capabilities, and attitudes, in order tobe-haveproperly. They must also be familiar with theentire product line, including thecommon glitchesand their various manifestations. Often this invol-ves some acquaintance with products from othervendors with which thecustomer might be interfac-ing. They mustrecall, given that they have iden-tified the problem, what the possible fixes are, ifany. They must adhere to theirown company's

VOLUME VI, NO

I

"

"

"3VOLUME VI, NO. 11

NOVEMBER 1990

policies and priorities, like the things you never say("the carpet isn'treally stain-proof), therelativeimportance of different customers, and the criticalissues which must be addressed immediately, for ex-ample, safety equipment needed for emergencyresponse. They also mustknow about each others'knowledge, about what specialists to call in for dif-ferentkinds of unsolved problems, and to whomone refers irate customers.

Add to this theirknowledge ofhow to use existingdatabase and textbasetools, theirknowledge of thedocumentation on their desks, and theirknowledgeabout inter-human communication protocols, and it isno wonder that training time for new CSRs is solong.

A SYSTEMS PERSPECTIVEBut the story doesn'tend there. The analysis in-

volves all thepeople in the knowledge chain. Thecustomers have knowledge about theirproblem,working environment, and about the products inquestion. Engineering hasknowledge about designdecisions, other vendors' interfaced products,designchanges, new products andreleases, whatcauses problems, and whycertain fixes work some-times and not others. And management, marketing,documentation, distributors, and field servicepeople also have relevant knowledge orknowledgeneeds. To maximize customer satisfaction, all ofthisknowledge may have to flow through the helpdesk.

Generally, where there areknowledge problemsthere have been prior attempts at solutions:manuals,bulletins,documentation, troubleshootingguides, CSR notebooks, CSR and customertrain-ing, weekly information exchange meetings, and ac-cess to the experts. When the operation has beenautomated, theremay be call-directing, database,and textbase technology in place. These effortsmust also be examined in theknowledge flowanalysis to see what works, what doesn't, and why.The output of the knowledge flow analysis is adetermination of who could do a better jobif theycould getbetter access to relevant information,andwhat is thebest way to getthat information there.

Once thecritical knowledge problems areproperlyidentified, the bestknowledge distribution strategycan be determined. The effort may take the form ofnew documentation, communication or training, orit may involve aknowledge system. The solutionmay be targeted for the CSRs or field service tech-nicians, given (or sold) to distributors or customers,or even integrated into future products. It may bebuilt by the customersupport department, engineer-ing, documentation, or by all of them.

Keeping an open mind about how informationshould be disseminated will facilitate your gettingthebiggest impact Besides the help deskoperation,yourdocumentation, marketing communications,and trainingorganizations should also becomefamiliar with the technology and its implications totheir functions. Ultimately, for most organizations,a full-time knowledge base maintenance capabilitywill be important tothe success of this technology,as was the case with database technology.

CONVENTIONAL AUTOMATION AT THEHELP DESK

Many service and support organizations have al-ready invested in some kind of automation, in mostcases call forwarding and database and textbase sys-tems. Several new products have been introducedinthe last year that employ intelligent database andtextretrieval technologies,rather than expert sys-tems, to address help desk problems. A recent ex-tensive survey of service and support managementsystems by AIM Consulting and CrescentProjectManagementexamines 90 software products aimedat this market The Help Desk Institute has also surveyed the current support systems in this area andhas proposed some new directions.

Theseproducts offer automationfor thefollow-ingkinds ofactivities:

1)Call Management: Answeringand queuing incoming calls for theright CSR. Sometimes this iscompletely automated, so thatthe call is answeredby a prerecorded voice and forwarded by thephonetreescharacteristic of automaticcall directing(ACD) technology. The information given to the

"

"

"

"4

NOVEMBER 1990

phone system (by pressing the touch tonebuttons inresponse to prerecorded questions) should bepassed on to the CSR.

2) Customer tracking: Whatequipment does acustomerhave? Have they called before about thisproblem or otherproblems? Access control is alsoprovided, since many companies sell their customerservices requiring the CSRs to identify the callers.Some systems even automate updating the billingdatabase by CSR "connecttime."

3) Field service management: Scheduling andtracking the dispatchand status of field service rep-resentatives. Advanced systems help determinewhich FSR should go and what tools and spares sheshould take along. Some even keep track of thespares inventory.

4) Help desk management: Keeping track ofwhat kinds of calls are coming in from whom,which ones get answered when,and how manycallers each CSR handles per hour.

5) Productproblem tracking: Which products aremost troublesome andwhy? Whatare the unresolvedcustomerproblems, who is working on them, andwhich customers are waiting for answers?

6) CSR notebooks: Sharing problem/symptom/solutionknowledge among CSRs by allowing themto enter descriptive text, laterretrieved by keywordsearches. This is the most sophisticated kind of sup-port currently available,and thekind most requestedby CSRs.

Some of these systems are quite sophisticated,well beyond the typical database application, espe-cially in the area of the CSR's shared notebook.Answer Computer's Apriori, for example, usesgraphicuser interfaces and sophisticated indexingand retrieval schemes to allow CSRs to share infor-mation about current problems, how customers

describe them, and what the fix is. (Table 1 detailssome Help Desk application/tools mentioned in thisarticle.) The system actually learns, real-time, asCSRs are dealing with a call, and makes the infor-

mation available,via networked UNIX worksta-tions, to every CSR immediately.

Another sophisticated system is Lysis Corp.'sSupport Information System. In this system, theCSRs do not enter information as free form text,

but rather use "semantic primitives," reminiscent ofRoger Schank's early-AI conceptual dependencytheory, to developa canonical underlying schemato uniquely capture problems and their descriptions.Companies consideringbuilding their own cus-tomized versions of "intelligent textbase systems"might look at the new groupware and textbase toolslike Lotus's Notes and Verity's Topic.

WHY ARE EXPERT SYSTEMS NEEDED?With all this technology available to bring to bear

at the help desk, where do expert systems fit in?First of all, they better fit in. That is, the expert sys-tem component should either duplicate convention-al facilities for automating helpdesk record keeping,information exchange, and integration with databasesand textbases,or should (preferably) be fullyin-tegrated with otherdatabase products on the market.

I have seen the following additional benefits,beyond database and textbase solutions, ofknowledge-based systems (observed on help desksin, for example, emergency response centers, datacenters,product complaint hot lines, and sales andtechnical support centers). They can:

1) Automate,completely, theresolution ofrepeti-tive kinds of calls, often handling 30-60% of the in-coming call volume without human intervention.This involves integration with thephone system'sACD capability, so that for "simple" problems thecaller simply answers a series ofpre-recorded ques-

tions by typing the appropriate numberon hertelephone. In those remaining situations when thesystem can't solve thecaller's problem, it serves asa triage analyst, gatheringall the information thatthe CSR will eventually want to see on the screen.

2) Support novice CSRs who simply can't findtheirway around the documentation or the databaseortextbase tools. These systems may also dramatically

VOLUME VI, NO. 11

fl

* _■_

__■»

"

#

"VOLUME VI, NO. 11 5

NOVEMBER 1990

Table 1 . Some Commercial Help Desk Applications & ToolsName of Product(Vendor)

Price Hardware (Operating System) Comments

Adept $695 IBM

PC,

PS/2 and compatibles(DOS); runs under Windows 3.0.

Proceduralexpertsystem tool using decision treestructures.(Symbologic)

Apriori $20,000 -$30,000

Sun 3 & 4,

SPARCstations;

16 MbRAM and 100 Mbhard drive

Customersupport/internal help desk system.Presents info to user in a "mostfrequently" handledmanner, automatically adjusting according tofrequency of access; maintainstext, graphics,applicationcode, and data files.

(Answer Computer)recommended(Unix).

CAIS $50,000 -$150,000

IBM PC/AT with minimum 640 XRAM (DOS).

Troubleshootingenvironment for service organizationsprovides diagnostic advice, electronic service manuals,product history data,etc.

(ROSH IntelligentSystems)

Enterprise/Dx $42,000 -$300 X

Widerange of platforms,including:

PCs,

minis,Intendedfor high-tech customer service andsupport;for developingand embeddingdiagnosticapplications.

(Coherent Thought)workstations, and mainframes.

Help Desk Assistant(Advantage KBS)

$35,000 Operates standaloneon IBM PCorLAN configs. with MF access;implemented using ADS/OS/2.

Problem management facility includingKBs to solveIBM networkand PC hardware/softwareproblems.Knowledge editorallows users to add newknowledge.

KLUE IBM PC/AT (DOS); 2 Mb RAMrecommended for Windows 3.0.Also requires KAPPA/PC.

Allowsnon-programmersto developdiagnostic/helpdeskKB applicationsfor sharing and maintaininginformation. Development/maintenance is bymeansof interactivedecisiongraphs.

(IntelliCorp)

LogicTree $495 IBM PC and compatibles (DOS). Rule-basedexpert system developmenttoolemployingdecision trees. Allowsaccess ofinformation in textualand graphic form.

(CAM Software)

Mahogany HelpDesk $10,000 IBM

PC/AT,

Windows; Mac with1 Mb RAM.

Development and deploymenttoolfor diagnostic/service applications. Uses decision-treestructures.(Emerald Intelligence)

Notes $62,500 IBM 286/386

PC;

requiresWindows 3.0 or PresentationManager. Runs under DOS orOS/2 for client side; OS/2 forserver.

Group commun. system for creating/distributinginfoon PCs andLANs. Applications include customertracking, statusreporting, and project management,etc. Allows interconnectionof DOS and OS/2 work-stationson LAN; supportsEthernet, Token Ring,Novell,

3Com,

and IBM networking systems.

(Lotus Corp.)

Path Builder $695/995 IBM

PC,

PS/2 and compatibles(DOS) (OS/2), IBM

mainframes;

implemented with ADS.

Task-specific tool utilizing diagnosticrepresentationindecisionnetworkform. Aids non-programmersindeveloping help desk/clientservice applications.

(Aion)

Support Information $3,000 -$50,000

IBM PC ornetwork (DOS),VAX(VMS).

For such applicationsas call tracking, answerretrievalmanagement, andreport generation.System(Lysis, Inc.)

TechSys $36,000 386-based systems with digitizedvoice hardware.

Automatedtech. support system combiningvoicetechnologywith a rule-basedexpertsystem. Allowscallers to do their own problem diagnosis.Diagnostic sessions maybe savedfor future use.

(IntelliSystems, Inc.)

Topic $25,000 -$30,000

DecVAX and RISC-basedplatforms (VMS/Ultrix),Sun, HP 9000, Mac.

Documentretrieval system for networked/hetero-genous environments. Ranks documents byrelevance accordingto query. Integrateswith Word-perfect, Word, Interleafand other tech.publishingsystems, and with

SQL-based

RDBMS.

(Verity)

© 1990 Harmon Associates. All Rights Reserved.

$5,000

" A.

"

#

"._____________■_■■-

NOVEMBER 1990

reduce the amount of training needed before a CSRbecomes effective on the job. When the problem isthat the CSR doesn't know what questions to ask,then an expert system'sQ&A style interaction,withexplanation of the line ofreasoning, may bring some-thing to game. In other situations, the use of an ex-pert system as a frontend to a textbase,capturingthe knowledge more experienced users have aboutwhat information is where, can be of greatassistance.

knowledge representations specifically designedtocapture diagnosticknowledge, allows information tobe added incrementally and automatically indexedand organized into troubleshooting procedures bythe system.

A SURVEY OF THE KINDS OF EXPERTSYSTEMS PRODUCTS ON THE MARKET

As vendors see these opportunities, they areresponding with new products. This involves mold-ing and extending expert systems technology to thesystems developers and users in this market niche.It alsorequires integrating the relatively small ex-pert systems component with the user interface,textbase, and database software that is already inuse. I will attempt to illustrate here some of thekinds of help desksystems incorporatingknowledgebase techniques appearing on the market. Althoughmost run on several standard IBM and DEC plat-forms, some only run on UNIX, Macintoshes, oreven Apollo workstations. And the price range isall over the map.

3) Implement standard diagnosticprocedures toprovide uniformityand consistency, orsupportCSRs who must deal with many pieces of complexequipment which heretofore only experienced en-gineerscould troubleshoot. Sometimes these sys-tems are used to help non-technical CSRs ask theright questions so that theycan schedule field ser-vice visits with the right repair person carrying theright tools and right spare parts on the truck, thusreducing the annoyance and cost ofrepeat visits.

4) Package information in a more useful form fordistributors or customers. For some companies,these software "advisors," which can provide serviceguidelines, step through diagnostic procedures orindex into on-line documentation, are a new sourceofrevenue. For others, the troubleshootingknowledge base is incorporated into the product it-self. The effect on the help-desk, of course, is toreduce the volume of "simple" calls thatcome in.

INTEGRATED WITH ACDPerhaps the most novelcustomer support system

product involving expert systems is IntelliSystems'TechSys. TechSys is designed to integrate withACD technology to automate,completely, the "callscreening and directing" task. In addition, the mostcommon "nuisance"problems can be diagnosedand handled by the system, over the phone, withoutCSR intervention. In the more difficult cases, thesystem helps thecaller organize all the required in-formation before passing thecall on to a CSR.

5) Enforce corporate policy about how thingsmust be phrased and sequenced over thephone.For example, in an emergencyresponse center hotline, there is no room for individual variation in theadvice given. Among TechSys 's installed base was the system

which handled my distress call about Windows 3.0this week. (The alternative to workingwith theautomated system, I was told in its first pre-recorded message, was to wait approximately 30minutes for thefirst available CSR.)

6)Facilitate maintenance of the diagnosticpro-cedures and other knowledge. This is sometimesaccomplished by offering decision-tree-like pro-gramming tools to the CSRs themselves. The goalis to reduce current knowledge updatingcosts andkeep everybody better informed. In situationswhere there are tens of thousands of possibleproblems, the diagnosticprocedures can be too complex to sort out in any other medium. Themodularity of theknowledge base, often using

TechSys' knowledge base representation andtotal integration withACD technology, voice mail,databases, billing systems, and the human technicalsupport operation show it to be a well-thought-outsystem.

" " ' VOLUME VI, NO. 11

i

X

_.

#

"

#

VOLUMEVI, NO. 11 7

NOVEMBER 1990

SPECIAL PURPOSE SHELLS ANDOVERLAYS

Another feature ofTechSys, which is commonto all of the expert systems entries in the help deskarena, is that the knowledge representation schemeis specificallydesigned for customer support tasks.TechSys 'sknowledge base, for instance, is or-ganized around products, symptoms, and problems.A number of so called task-specific shells, designedfor modeling support-related tasks like checklistsand equipment diagnosis, are now appearing. Someare designedfor organizations already into expertsystems, some for very complex diagnostic situa-tions, while some use decision-tree interfaces tofacilitate "end-user programming" ofdiagnosticprocedures by the CSRs themselves.

One such tool is the Mahogany HelpDeskfromEmerald Intelligence. Emerald's MahoganyProtool was justanother entry in the low-end shellgame for PCs and Macs, with some novel featuresfor an inexpensive tool like point and click user in-terface, simple-to-useknowledge base editingfacilities,automatic integration of rule parameterswith an object-oriented database, and pattern-match-ing rules over that database. But the MahoganyHelpDesk product goes well beyond that. Its Help-Builder module is designed to support end-userdevelopment, that is, programming theknowledgebase by the CSRs themselves,by offering adecision-tree-like developer's interface.

GRAPHIC USER INTERFACEMahogany's HelpStationruntime module allows

full integration with graphics,scanned photographs,and even video or animation. As Pam McGahademonstrated many years ago in a help desk ap-plication shebuilt at Texas Instruments, it is oftenthe case in supporting CSRs that the "user inter-face," the visual screens that the CSR sees, mustbecarefullydesigned to allow the CSR to take into ac-count the levelof sophistication of the caller. Inother words, a traditional expert systems dialogue,which asks all the questions in the appropriate ordermodeling the expert's diagnostic thinking, may be

tooconstraining for use by the CSR. Graphics andpictures, from which the CSR can phrase her ownquestions, can sometimes be veryeffective.

BEYOND END USER PROGRAMMINGThere are limits as to how much programming to

expect from the end users. And expert systems areprograms. The simple decision-tree developmentenvironment offered by the Mahogany HelpDeskand similar tools like Symbologic's Adept and CAMSoftware's LogicTree, will only go so far. If thetrees get too big, or if they need to be changed toooften, or if they must be maintained by several dif-ferent CSRs, theycan become more of a nuisancethan an aid. Furthermore, if data needs to be drawnin from adatabase, a programmer must get involved.

TheMahogany HelpDesk, like other low-endproducts, will be offering a Customizeroption toallow programmers to integrate the decision-tree-created knowledge bases with other code, user inter-faces, and databases. Another overlay tooldesignedfor end users in conjunction with profes-sional programmers is the new Path Builder productfrom Aion, now in beta-testing. This task-specificoverlay for Aion'sbest-selling ADS mainframeshell allows convenient representation of diagnosticdecision trees. Path Builder can produce an ADSexportfile, which can thenbe integrated with other,programmer-produced ADS code using ADS's Ver-sion Maintenance feature.

Advantage KBS has gone one step further withits Help Desk Assistant (see 755, Vol. VI, No. 7).This task-specific help desk tool, again based onAion's ADS, is designedfor data center supportoperations. The system offers problem manage-mentfacilities and integrates with IBM's Info/Manproduct and NetView. What is novel about this of-fering is that you can buy some "knowledgemodules" along with the tool. Knowledge Editorlets youupdate and maintain knowledge bases asyour hardware, software, or procedures change.This enables the user to modify and add new know-ledge using point-and-click interfaces. Program-ming knowledge of the underlying development

\

.i ___r»

"

"

i

"VOLUME VI, NO 118

NOVEMBER 1990

tool is not required. Currently modules for networkand PC problems are available. Obviously everydatacenter will have its own knowledge about itsown networks and PCs, but theseknowledgeproducts might be tremendous acceleratorsfor or-ganizations juststarting to automate the help deskwithknowledge systems.

HEAVYWEIGHT ENTRIESIt remains to be seen howfar one can go with

end user programming for the help desk. In addi-tion to the inherentcomplexity of large decisiontrees, there are attitudinal issues involving CSRs asprogrammers. And there are technical issues: mostimportantly, how do you integrate a bunch ofdecisiontrees built by specialist CSRs on networked PCsinto a logically consistent knowledge system offer-ing all the relevant advice for every situation?

Whatever the eventual solutionto this technicalproblem is, the end user programming model willnot cover all situations. In particular, whenoffering

service aid for suites of complex equipment withtens of thousands of possible problems andrigorousdiagnostic procedures, we are back in the realm ofprofessionalknowledge engineering. Severalentries —for example, ROSH Intelligent Systems'CAIS, IntelliCorp'sKLUE, andCoherent Thought'sEnterprise/Dx (seeESS, Vol. 6, No. 6for more onEnterprise/Dx)— now offer workstation-based,task-specific tools for knowledge engineers facedwith this level of diagnosticcomplexity.

SOLD WITH CONSULTINGSeveral general-purposeshell vendors and sys-

tems-integration consulting firms now have serviceofferings specificallydesigned for help-desk ap-plications, based on their experience, and oftencode, from other client's help-desk applications.The vendors specifically addressinghelp deskswhom I met at the recent AAAI meeting in Bostoninclude IBM, AlCorp, Bell Atlantic,SoftwareArtistry, IntelliCorp, AndersenConsulting, andTechnology Applications.

SOURCES

CONSULTANTS AlCorp, Waltham, Mas-. _. _,

■

sachusetts;(6l7)B9l-Daniel Burrus, Burrus 6500Research Associates,P.O. Box 26413, Mil- Bell Atlantic, Princeton,waukee, Wisconsin 53226- New Jersey; (609) 275--0413; (414)774-7790. 4545.

Keith Telle, AIM Consult- Software Artistry, In-ing, Pleasant Hill, Califor- dianapolis,Indiana;(317)nia; (41 5) 944-4749. 876-3042.Jacky Hood , Crescent TechnologyApplications,Project Management, Palo Jacksonville,Florida;Alto,

California;

(41 5) 328- (904) 296-1 685.1720. . . „Andersen Consulting,The Help Desk Institute, Chicago, Illinois;(312)ColoradoSprings, 580-0033.Colorado; (800) 248-5667. nn„n, 1/wsv ' PRODUCTSPam McGaha, Texas In- . . . . _struments, Piano, Texas; AP"°". An!wer Co"-(214) 575-4982 P uter> ,nc- Sunnyvale,

V£i-*;o/o

*. California; (408)739-6130.IBM, Palo Alto,California;(415)855-4076.

Support Information Sys- LogicTree, CAMtern, Lysis, Inc., Decatur, Software, Provo, Utah;Georgia; (404) 373-3359. (801 ) 373-4080.Topic, Verity, Mountain Path Builder, Aion, PaloView, California; (41 5) Alto, California; (41 5) 328--960-7600. 9595.Notes, Lotus Corp., Help DeskAssistant, Ad-Cambridge, Massachusetts; vantageKBS, Edison,(617) 577-8500. New Jersey; (201 ) 287-

-2236.TechSys, IntelliSystems,Inc., Chatsworth, Califor- Enterprise/Ox, Coherentnia; (818) 341-7000. Thought, Palo Alto, Califor

v . n _ nia; (41 5) 493-1555.Mahogany HelpDesk,Emerald Intelligence,Ann CAIS Adviser, ROSH In-Arbor, Michigan; (313) telligent Systems, Need--663-8757. ham,Massachusetts;

„_..„_. (617)449-0049.Adept, Symbologic, Red-mond, Washington; (800) KLUE, IntelliCorp, Moun--448-9292. tain View, California; (41 5)

965-5500.

t

'_■_ m

"

*9VOLUME VI, NO. 11

NOVEMBER 1990

CONCLUSIONJustas database software toolschanged people's

ideas about computers in the 19505, this technol-ogy for embedding and debuggingknowledge insoftware will change the way businesses use com-puters. Networkedknowledge systems will emergeas a new communications medium. And, just aspayroll applications years ago showed that dataprocessing technology was non-optional for busi-ness organizations, so help desk applications willblaze the trail for networked knowledge systems.Ultimately, the medium will find use outside of thecorporate world, in political arenas where the shar-ing ofknowledge is most difficult.

Knowledge, the AI theorists tell us, is an"ascribed" property. We attribute it to people,

WHAT'S HAPPENING IN EUROPEOVERVIEW

European companies are very interested in Ex-pert Systems. From the perspective of an American,it's about 1987 in Europe, with some significant dif-ferences. European companies have discovered ex-pert systems and have produced a wide variety ofsophisticated applications, enough to convince cor-porations that there really is something worth in-vesting in. On the other hand, the mainframe toolsare justbeginning to show up in Europe. That inturn, means thatcorporate MIS departments don'tcurrently have any effective way to developknowledge-basedsystems, which theycan integratewith their existing systems.

It's not that the U.S. mainframe vendors aren'tthere and making some sales,but theEuropeansstill seem to be primarily focused on PC andworkstation-based tools. They are also still inter-ested inLisp-based tools. Indeed, the only thingkeeping the U.S. Lisp vendors alive, at this point, istheirEuropean sales.

Unfortunately, it's like 1967 in the U.S. in thesense thatmany of the expert system tool vendorsare in trouble. 1991 will probably see a shakeout that

programs, or robots, depending on how they be-have. These researchers then go on to confuse theissue by talking about "representing"knowledge indata structures, but that is another story (see Barr,TheRepresentation Hypothesis, 1979). In movingexpert systems technology into the networked sys-tems or groupware arena, we will see the real na-ture and power ofknowledge, as a social enterprise,transforming the way we work and the way wethink. If we can successfullyautomateknowledgeflow at the help desk, we can change the way com-panies and people know.

For further information,contact Avron Barr atAlto Ventures, 261 Hamilton Aye.,Suite 214, PaloAlto,CA 94301; (415) 322-2233, Fax: (415) 322--2918.

PARTIwill eliminate lots of the small Lisp and Prolog-based tool vendors and usher in a much greaterem-phasis onC and mainframe-based tools. Ignoringtherecession which Europe may or may not ex-perience, we expect the shakeout to be over quickly,since the C and mainframe tools that corporationswill increasingly demand are already available —from the same U.S. vendors that have emergedfromthe transition in the U.S. We predict that 1991 willbe the year of the mainframe expert system tool inEurope.

Thereare more, popular domain- and problem-specific tools in Europe than in the U.S. Moreover,although Texas Instrument's Testßench andGensym's G2 are both popular, mostof the interest-ing domain-specific tools are of European origin.Some of the best of these tools, including Design++ and Maintex, are already being marketed in theU.S. and we will be surprised ifwe don't seeseveral more European domain/problem-specifictools being sold in the U.S. in 1991 and 1992.

Theother thing that impresses me is the numberofmethodologies andknowledge acquisition toolsavailable in Europe. Clearly theEuropeans are in-terested in learning how to systematically develop